hash stringlengths 32 32 | doc_id stringlengths 7 13 | section stringlengths 3 121 | content stringlengths 0 2.2M |
|---|---|---|---|
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 2 References | |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 2.1 Normative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found in the ETSI docbox. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long term validity. The following referenced documents are necessary for the application of the present document. [1] ETSI TS 104 050: "Securing Artificial Intelligence (SAI); AI Threat Ontology and definitions". [2] ISO/IEC 22989: "Information technology - Artificial intelligence - Artificial intelligence concepts and terminology". |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 2.2 Informative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long term validity. The following referenced documents are not necessary for the application of the present document but they assist the user with regard to a particular subject area. [i.1] ETSI TR 104 221: "Securing Artificial Intelligence (SAI); Problem Statement". NOTE: An earlier version of the above document is available as ETSI GR SAI 004. [i.2] ETSI TR 104 048: "Securing Artificial Intelligence (SAI); Data Supply Chain Security". NOTE: An earlier version of the above document is available as ETSI GR SAI 002. [i.3] ETSI GR NFV-SEC 003: "Network Functions Virtualisation (NFV); NFV Security; Security and Trust Guidance". [i.4] Auguste Kerckhoffs: "La cryptographie militaire" Journal des sciences militaires, vol. IX, pp. 5-83, January 1883, pp. 161-191, February 1883. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 6 [i.5] Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act). [i.6] DARPA: "XAI: Explainable Artificial Intelligence". [i.7] Margaret Mitchell, Simone Wu, Andrew Zaldivar, Parker Barnes, Lucy Vasserman, Ben Hutchinson, Elena Spitzer, Inioluwa Deborah Raji, Timnit Gebru. Conference on Fairness, Accountability, and Transparency: "Model Cards for Model Reporting", 29 January 2019, Atlanta, GA, USA. ACM, New York, NY, USA. [i.8] Samek W., Montavon G., Vedaldi A., Hansen L. K. and Müller K. R. (eds.) (2019): "Explainable AI: Interpreting, Explaining and Visualizing Deep Learning". Cham, Springer. [i.9] Timnit Gebru, Jamie Morgenstern, Briana Vecchione, Jennifer Wortman Vaughan, Hanna Wallach, Hal Daumé III and Kate Crawford: "Datasheets for Datasets", Communications of the ACM, Volume 64, Issue 12, pp. 89-92, November 2021. [i.10] Lapuschkin S., Wäldchen S., Binder A., Montavon G., Samek W. and Müller K. R. (2019): "Unmasking Clever Hans predictors and assessing what machines really learn". Nat. Commun. 10, doi: 10.1038/s41467-019-08987-4. [i.11] Molnar C.: "Interpretable Machine Learning-A Guide for Making Black Box Models Explainable". [i.12] Samek W., Montavon G., Binder A., Lapuschkin S. and Müller K. R. (2016): "Interpreting the predictions of complex ML models by layer-wise relevance propagation", arXiv abs/1611.08191. [i.13] ETSI TR 104 102: "Cyber Security (CYBER); Encrypted Traffic Integration (ETI); ZT-Kipling methodology". |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 3 Definition of terms, symbols and abbreviations | |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 3.1 Terms | For the purposes of the present document, the terms given in ETSI TS 104 050 [1] and ISO/IEC 22989 [2] and the following apply: AI system: engineered system that generates outputs such as content, forecasts, recommendations or decisions for a given set of human-defined objectives NOTE: Definition from ISO/IEC 22989 [2]. explainability: property of an AI system to express important factors influencing the AI system results in a way that humans can understand NOTE: Definition from ISO/IEC 22989 [2]. explicability: property of an action to be able to be accounted for or understood transparency: property of an action to be open to inspection with no hidden properties |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 3.2 Symbols | Void. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 7 |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 3.3 Abbreviations | For the purposes of the present document, the following abbreviations apply: AI Artificial Intelligence BTT Build-Train-Test DARPA Defence Advanced Research Projects Agency LRP Layer-wise Relevance Propagation ML Machine Learning OECD Organization for Economic Cooperation and Development RTE Run Time Explicability TA Trust Association XAI eXplainable AI |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 4 Explicability and transparency | The SAI problem statement, ETSI TR 104 221 [i.1], identifies explicability as being a contributor in establishing trust in AI systems as one element of achieving transparency. However, in computer science the concept of transparency is somewhat at odds with explicability and can be interpreted as "functioning without the user being aware of its presence" when referring to a process. The term transparent (and its associated noun form, transparency) when applied to AI is, for the purposes of the present document, the core concept of being open to examination, or having no part hidden. The term explicability is, in very crude terms, being able to show how any result was achieved ("show your working"), which when combined with transparency gives assurance that nothing is hidden. NOTE 1: In ETSI TR 104 221 [i.1] and in ISO/IEC 22989 [2] the term explainability is used whereas in the present document the more common term in UK English, explicability, is used. NOTE 2: It is recognized that many processes are protected from disclosure by mechanisms that protect the intellectual property that the processes contain and such protections are not intended to be impacted by the requirement to maintain attributes of transparency and explicability. The outcome of applying constraints of explicability and transparency to systems is that trust can be conferred as a system attribute that is open to examination and verification by third parties. It is recognized that in many systems, such as in telecommunications, the role of AI is often at a component level. The role of most applications is not to explicitly design or develop intelligence as a primary goal. Trust should not be attributed where purpose is not clear. One purpose of transparency and, particularly, explicability is to prevent the AI components of a system from denying that they took part in an action, and to prevent the AI component denying they were the recipient of the output of an action from any other part of the system. NOTE 3: The description above is very close to the common definition of non-repudiation but there is a subtly different intent in the scope of explicability and transparency, hence for the present document this is not referred to as non-repudiation. In ETSI TS 104 050 [1], it is stated that there are a number of characteristics associated to intelligence the key elements of which are given below, and in the context of transparency and explicability it is expected that each of these characteristics, if they are present in the AI component or system, is described: • reasoning: the application of learned strategies in order to solve puzzles, and make judgments where there is uncertainty in either the input or the expected outcome; • learning: the means by which reasoning and other behaviour evolves over time to address new input; • communicating: in natural language (to human third parties), in particular when within the bounds of the system it is unable to process data to a known state. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 8 In terms of explicability it should be clear where reasoning takes place, and on what data and algorithm, such reasoning is based. Similarly the scope of explicability and transparency addresses the means by which the system learns. Finally, in the context of the key characteristics above, the means by which the system's purpose is communicated should be in natural language where the intended recipient should be considered as a lay person (i.e. having no knowledge of any specialized language of AI/ML or of the programming techniques of AI/ML). Many concerns raised regarding AI/ML (see ETSI TR 104 221 [i.1]) and addressed as "Design challenges and unintentional factors" can be made visible through the application of specific explicability techniques. An example is the concern of bias (confirmation bias and selection bias in particular) where, by the application of simple checklists (see clauses 5 and 6) the system deployment should be able to answer questions of the form "why was this data source selected?". EXAMPLE: An AI can be biased by design if the purpose of the AI is to filter candidates for a job based on some personal characteristic (i.e. as opposed to a meritocratic selection engine, the AI acts as a characteristic selection engine). In such a case the explicability and transparency requirements will be able to identify that negative, or trait-based, filtering is at the root of the reasoning engine of the AI. It is reasonable to suggest that bias in inputs will be reinforced in the output, hence in clause 5 it is stressed that explicability addresses the purpose of data. If data is preselected to achieve a particular result that could be seen to be consistent with selection bias and that would need to be explained as part of the system purpose (as in the example) or removed by design. |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 5 Static explicability analysis | |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 5.1 Summary of the role of static explicability analysis | The role of static explicability is closely related to giving detailed system documentation. The purpose of explicability is to allow a lay person (i.e. not a professional programmer or system analyst) to gain a reasonable understanding of the main data flows and processing steps in the program. EXAMPLE: A data set of images is used as training data and routinely classified as images of, say, "Cat", "Dog", "Fox", "Badger" where the purpose is to enable a camera observing a suburban garden to record movements of particular animals at night, thus being able to say that a badger crossed the garden lawn at a particular time of the night. In a simple scenario such as in the example above the purpose is clear (identify which animal is in the capture range of the camera), it is clear where the training data comes from (the set of images), and it is reasonable to expect a layperson to understand the purpose, the role of data and components in the system, and to make reasonable attempts to verify the veracity of the system (e.g. by getting a dog to pass in front of the camera and be recognized as a dog, or for a deer to pass in front of the camera and not to be recognized as one of the animals it has been trained to recognize). As more components are added to the system to improve the system's ability in recognition, say by adding gait analysis (dogs and cats move quite differently) static explicability shall be maintained (i.e. at all times static explicability shall be a characteristic of the current system). The components identified in table 1 shall be clearly identifiable in the system documentation. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 9 Table 1: System documentation elements in static explicability analysis Documentation Element Element Mandatory Short description 1 Statement of system purpose Yes This element of the system documentation is intended to allow a layperson to clearly understand the purpose of the system and to explicitly identify the role of AI in achieving that purpose. 2a Identification of data source(s) Yes Where the data comes from and how the authenticity of the data source is verified. 2b Purpose of data source(s) (in support of system purpose) Yes The role of the particular data source in the system (e.g. training data containing images of dogs to train the system in recognizing a dog from an image). 2c Method(s) used to determine data quality Strongly recommended Methods and processes used in determining if the input data is a fair and accurate representation of the desired input. This should address how bias or preference is identified and corrected in the data input. 3 Identity of liable party Yes For each processing or data element a means to identify liability for correction of errors or for maintenance of the element. 5.2 Requirements for documenting the statement of system purpose The statement of system purpose is critical in allowing a layperson to clearly understand the intent of the system and the role of AI in achieving that purpose or intent. EXAMPLE 1: AI used in a voice-recognition personal assistant. The purpose of the system is to allow the user to issue spoken commands in natural language and to translate those into machine commands for purposes including machine control, and internet-based information search and retrieval. The AI in the system provides a number of functions in order to achieve its purpose including: AI to enable speech recognition; AI to assist in parsing of recognized speech to commands; AI to drive voice responses to spoken commands; AI to parse and relay the results of search commands into natural language. NOTE 1: In the above example multiple AI capabilities are identified even if the perception of the user is of a single AI being applied. EXAMPLE 2: AI used in adaptive cruise control in road vehicles. The primary purpose is to ensure that whilst the driver can set a target speed to be maintained it is recognized that strict adherence to the target speed can be unsafe. The role of the AI in this system is to maintain a safe distance between vehicles whilst maximizing the time spent at the target speed. The system therefore adaptively modifies the vehicle speed (not exceeding the target speed) by maintaining a "safe" distance from other vehicles through selective braking and acceleration where data on the presence and actions of other vehicles are obtained from system sensors and driver input. The statement of system purpose should be written in natural language and be concise as well as precise (i.e. not open to variations in interpretation). The following characteristics shall be identifiable in the statement of system purpose: • Unambiguous: it should be impossible to interpret the system purpose in more than one way. • Complete: the system purpose should contain all the information necessary to understand it without requiring reference to other documents. NOTE 2: The above requirement may be seen to contradict best practice in standards development where referencing is used to ensure succinctness, whereas in the statement of system purpose a little more verbosity may be beneficial. • Precise: the system purpose should be worded clearly and exactly, without unnecessary detail that might confuse the reader. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 10 • Well-structured: any individual elements of the system purpose should be included in an appropriate and easy-to-read manner. The present document provides a template for the documenting of the system purpose in Annex D. 5.3 Methods in documenting the identification, purpose and quality of data sources As outlined in table 1 where data is used in AI the liable party should ensure that answers are documented for the following questions (this is also addressed in the ZT-Kipling method defined in ETSI TR 104 102 [i.13] and in Annex A): • Where does the data come from? - As the purpose of data has been indicated earlier this clarifies explicitly the source of the data. This can include statements such as the following for the example of adaptive cruise control: "the range-data indicating the distance to surrounding vehicles and environmental objects is sourced from a radar array positioned at the front left, centre and right of the vehicle". • How is the authenticity of the data source verified? - The aim here is to ensure that only trusted data (data sources) are used in the system. • What is the role of the particular data source in the system? (e.g. training data containing images of dogs to train the system in recognizing a dog from an image). • What methods and processes are used in determining if the input data is a fair and accurate representation of the desired input? • What steps have been taken to determine if the input data has bias? - It can be argued that all data is biased and that all designers will have some degree of selection bias in the data chosen to train and run their systems. However it is essential that designers be as objective as possible when documenting their sources. If similar data sources were available it may be necessary for the designer to show why one source was selected over any alternatives (e.g. for reasons of cost, or trust in the source as opposed to the content). • What steps have been taken to compensate for any bias in the input? - As has been noted bias can be a design decision. In many instances it may not. Bias can be compensated in a number of ways including modification of data ranking or direct modification of the source to remove inherent bias. Any steps taken to compensate for bias should be documented in clear, concise, and precise natural language. The use of Model Cards outlined in [i.7] performs much of the above role and where in [i.7] it is stated that there are no standardized documentation procedures to communicate the performance characteristics of trained Machine Learning (ML) and Artificial Intelligence (AI) models the approaches outlined in the present document and those in [i.7] are part of closing that gap in standardization. In addition, the use of datasheets as outlined in [i.9] provides a means to facilitate communication between dataset creators and consumers that is consistent with the intentions of the present document. |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 5.4 Identifying who is the liable party | In undertaking analysis and in providing the necessary documentation it should be made clear who is responsible for the AI system, and the system of which it forms a component. This should be consistent with any other obligations when placing products on the market. NOTE: This is addressed in part in the AI Act [i.5] as part of the transparency requirements in Article 13. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 11 |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6 Run time explicability | |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6.1 Summary of service | When an AI system is running it applies its AI to data to achieve its purpose. The goal of run time explicability is to ensure that the system developer, and other stakeholders in the supply chain, can identify the role of active processes, and data, in achieving the system purpose. Static explicability is a pre-requisite to run-time explicability. Run Time Explicability (RTE) is defined in the present document as an explicit service of a running system. The goal of the explicability service is to collect, maintain, make available and validate irrefutable evidence concerning the purpose of, and data contributing to, an action of the machine in order to assist in determining the validity of the action at the time it was taken. NOTE: The explicability service is closely related to conventional non-repudiation services but with the intent of explaining actions rather than for solving disputes (see also clause 4). |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6.2 Abstraction of AI system | An abstract model of an AI processing system is given in ETSI TR 104 221 [i.1] from which figure 1 is taken to represent stages in the ML lifecycle. Figure 1: Typical machine learning lifecycle (Source: ETSI TR 104 221 [i.1]) Explicability applies to the Build-Train-Test (BTT) cycle during model design, and to the role of the update cycle during deployment that supplements the BTT cycle. |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6.3 Evidence requirements for explicability | The requirements for static explicability, outlined in clause 5, apply as a pre-requisite to providing evidence for run-time explicability. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 12 As indicated above, explicability (and transparency as a pre-requisite) aims to prevent the AI components of a system from denying that they took part in an action, and to prevent the AI component denying they were the recipient of the output of an action from any other part of the system. The RTE service expands on the set of questions outlined in clause 5.3 and summarized below: • What process does data undergo between acquisition and curation? - The lifecycle shown in figure 1 identifies data acquisition and curation used in development of the model that is used in implementation (following a BTT cycle), and also in the active deployment phase where results are used in feedback to refine the implemented model. It is reasonable to filter data between acquisition (say where multiple data sources are used) and its curation (say by removing fields from data sources where those fields are not relevant to the model). • What are the metrics that determine change in the learning/weighting of data? - Notwithstanding any intention by the designers to open intellectual property embedded in the feedback and feedforward learning process it should be made clear to the user of the system what is involved in the learning process. |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6.4 Performance considerations | |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6.4.1 General requirement | An AI/ML system can make decisions at a rate that, if a detailed evidential record was to be created, and retained securely, has potential to overload the system. Rather than take a detailed evidential record for every decision the goal of explicability and transparency is to ensure that the rationale for a decision is clear. The system documentation shall clearly identify those events that are logged for future analysis. The content of the log record shall be sufficient to identify the trigger conditions of the event and should include the following: • The time that the event occurred (including the metric and basis of the time). EXAMPLE: Time may be system clock time, e.g. UTC time, or may be relative to some datum (e.g. clock cycles from a known datum point). • The software versions of the processes involved or impacted in the event. • The hardware elements involved or impacted in the event. • The data, and its supply chain, involved or impacted in the event. In addition the liable party for resolving the impact of the event should be clearly identified. NOTE: In the AI Act [i.5] it is stated that high risk systems have the capability to allow for automatic recording of events over the lifetime of the AI System and this is consistent with the requirements outlined in the present document. |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6.4.2 Precision and recall metrics | In addition to issues related to performance from audit and logging as above, the designer shall define and publish the expected accuracy of the system. This should be achieved by explicitly identifying the measure of precision and of recall against both static data and live data. • Precision, the measure of positive predictive value, measures the correctness of the decision every time the model made a positive decision. Precision can only be reliably measured against a known input (the number of relevant elements in any sample is known). Precision = Number of true positives / (number of true positives + number of false positives) • Recall is the measure of overall success at identifying relevant elements. As for precision, recall can only be reliably measured against a known input. Recall = Number of true positives / (number of true positives + number of false negatives) ETSI ETSI TS 104 224 V1.1.1 (2025-03) 13 EXAMPLE 1: An AI system is designed to recognize dogs in an image (dogs are the relevant elements). If the system is presented with an image that contains ten cats and twelve dogs (i.e. there are 22 identifiable animals in the image), and the system identifies eight dogs, of the eight elements identified as dogs, only five actually are dogs (true positives), while the other three are cats (false positives). Seven dogs were missed (false negatives), and seven cats were correctly excluded (true negatives). The program's precision is then 5/8 (true positives/selected elements) while its recall is 5/12 (true positives/relevant elements), i.e. precision of 62,5 % and recall of 42 %. EXAMPLE 2: An AI system is designed to grant people access to a secure building using facial recognition. The system recognizes 150 people and grants access to 100 of them with a precision of 98 % meaning that of 100 people granted access, 2 were not supposed to enter the building. However if in the 150 people recognized there were in fact 120 that should have been granted access the recall rate is 98/120 or only 82 %. There are many other ways of measuring the system performance using other statistical measures but the key point is that the system documentation shall clearly indicate the measure by which the system claims to be accurate and the level of accuracy to be met by the system. A run-time measure of accuracy should be considered to be developed and implemented as part of an AI system's design. NOTE: Accuracy can be used as a component in developing trust, see also Annex A. |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 6.5 Application of XAI approaches | Complementing the approaches presented above, academic research on more complex technical methods for gaining insights into the behaviour and decisions of AI models is performed in the field of eXplainable AI (XAI). Depending on the use case, different methods can be used. An overview of the different approaches is given in [i.8]. When making predictions from structured data, probabilistic methods are generally considered promising [i.11], whereas applications from computer vision rely on more advanced methods such as Layer-wise Relevance Propagation (LRP) [i.12]. Some XAI methods provide global explanations, while others explain individual (local) model decisions. One useful application of XAI methods has been to uncover spurious patterns in data sets learned by AI models and leading to wrong decisions [i.10]. A number of projects have been created under the DARPA XAI [i.6] leadership to address the following aspects of AI as applied to ML: • produce more explainable models, while maintaining a high level of learning performance (prediction accuracy); and • enable human users to understand, appropriately trust, and effectively manage the emerging generation of artificially intelligent partners. Whilst the XAI programme is not complete and does not directly produce standards the goals are aligned to both the static explicability analysis (clause 5) and the RTE service (clause 6) of the present document. It is noted that the XAI program is focused on the development of multiple systems by addressing challenging problems in two areas: 1) ML problems to classify events of interest in heterogeneous, multimedia data; and 2) ML problems to construct decision policies for an autonomous system to perform a variety of simulated missions. These have been chosen to represent the intersection of classification and reinforcement learning, and also address the intersection of gathered data analysis and autonomous systems. A third major element of the XAI project is to gain a better understanding of the psychology of explanation which reinforces the intent of the present document to provide the user with greater understanding of the role and scope of AI in systems. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 14 |
1a78e9090133dfd7ab8c19db79ccb9ca | 104 224 | 7 Data transparency | ETSI TR 104 048 [i.2] identifies the role of understanding the data supply chain as a link in integrity and availability assurance. As stated in clause 4 transparency when applied to AI is related to being "open to examination". The value of integrity checks, e.g. using cryptographic hashes, in transparency is that they are able to indicate unauthorized change between sender and receiver. Thus a general requirement for data transparency with respect to integrity is as follows: • The recipient of data should be able to determine if the data has been manipulated by a 3rd party before receipt (i.e. in the period from the sender releasing data to the recipient receiving it). The general requirement shall be extended to the data lifecycle, consisting of the following phases, as follows: 1) Data in transit, i.e. Connectivity, data transported through space, mathematical integrity shall be assured. 2) Data at rest, i.e. Storage, data transported through time, mathematical integrity shall be assured. 3) Data in process, i.e. Compute, data acted upon. In this case as the data is likely to be modified the integrity of the processing should be assured. In addition to determining data integrity the recipient, in support of transparency, needs to determine the source of the data. This in turn requires that at least one of the following additional technical measures is addressed: • The recipient of data should be able to identify the source of data. NOTE 1: A recipient process may receive data from a process operating in the same environment with the same owner. In such cases the identity of the data source would be the process. • The recipient of data should be able to verify the identity of the liable entity for the source of data. NOTE 2: The intent of this measure is to give assurance that the data recipient not only knows the data source but also that that data source has liability for the content of the data received. • The recipient of data should be able to verify that the data source has authority to share data with the recipient. NOTE 3: The intention of this measure is to limit the risk of data that has been unlawfully obtained (e.g. a copyright infringement) is used in the AI system. Data transparency in ML systems applies in particular to the Data Acquisition and Data Curation phases, i.e. where the data comes from. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 15 Annex A (normative): Trust in AI for transparency and explicability In the context of AI the model of trust that is offered by the AI is part of the overall relationship of the AI and its dependent users. How AI entities build trust is complex and can differ from the trust measures used in simpler, non-AI, systems. In practice a number of security assurance elements are combined to determine an overall trust level. Such elements include identity, attribution, attestation and non-repudiation. In the context of AI a number of objectives for trust apply, alongside transparency and explicability. The assignment of trust in conventional discourse is the decision that an entity A should trust entity B in one or more particular contexts. Key criteria for assigning trust are: • The identity of the entity to be trusted. • The contexts within which the trust should be constrained. The security relationships of an AI, in addition to countering risks and attacks on the system, are used to reinforce trust relationships. A number of trust models are commonly used in technology: • Delegated trust: - entity A is unable to evaluate the appropriate level of trust for a relationship with another entity B, thus entity A can choose to delegate the decision to another entity C. • Collaborative trust: - two entities (entities A and C) work together to decide whether to trust another (entity B) - the final goal can be for both entity A and entity C to have a trust relationship with entity B. • Transitive trust: - entity A trusts entity B because entity C trusts it. A more complete description of the role of trust in networks is found in ETSI GR NFV-SEC 003 [i.3]. In the context of an AI the role of trust is somewhat complex as there is not a single root of trust, rather there has to be trust in the process of learning, of data sources, and of the actions taken. The relying party, that is the party dependent on the AI output, should be able to build a trust model of the AI system. There are therefore a number of Trust Associations (TA) in the AI/ML system each with an independent quantitative (and qualitative) assessment of their Trust Value. The metrics for determining the trust value are for further study, but it is considered that the Trust Value assigned to the overall system is given as the (vector) sum of the set of Trust Values of each TA in the system. = ∑ . In addition trust can be associated to accuracy (e.g. the combination of precision and recall), or to other metrics associated to the processing. It should be assumed that a zero-trust model applies and that every TA is verified using the ZT-Kipling model from ETSI TR 104 102 [i.13] and applied as below in order to reinforce both transparency and explicability and the common security principles of: • Minimize the attack surface. • Impose a principle of least privilege to allow the use of any asset. • Impose a principle of least persistence for the use of any asset. The ZT-Kipling method is characterized as a security strategy designed to prevent breaches by eliminating implicit trust in the digital world while constantly verifying all users, devices, and applications across all locations in order that trust becomes explicit. As AI and ML are means of realizing a digital world the method applies to AI and ML as to any other realization of the digital world. The ZT-Kipling method consists of five (5) iterative (and recursive) steps each of which poses the Kipling questions (or criteria). The steps are: 1) Define the protected surface - identify what shall be protected. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 16 2) Map the transaction flows - how does the traffic flow to, through, and from the protected surface. 3) Build a Zero Trust Architecture (ZTA) - based on the protected surface and the transaction flows, what should ZTA look like? What are its security components and mechanisms? 4) Create Zero Trust security policy - follow Kipling methodology to define the Zero Trust security policy, which adheres to the defined ZTA. 5) Monitor and maintain - maintain and monitor the protected surface. Figure A.1: ZT-Kipling Methodology Steps The ZT-Kipling method a number of questions are answered in order to allocate trust to any relationship in the system (see table A.1). Table A.1: ZT-Kipling method root question set Question Example for asset existence Example for asset access What What is the asset? What is the entity accessing the asset? Why Why is that asset in the system? Why is that entity accessing the asset? When When is the asset meant to be available (e.g. is it ephemeral or persistent, if ephemeral how is it invoked and so forth)? When is the asset being accessed (is it being accessed at a reasonable time)? How How is the asset operated (e.g. what does it require in order to operate)? How does the asset know and verify that access is permitted? Where Where is the asset (logically and geographically)? Where is the entity with relation to the asset (local or remote)? Who Who owns the asset? Who is the entity accessing the asset? ETSI ETSI TS 104 224 V1.1.1 (2025-03) 17 Annex B (informative): Threats arising from explicability and transparency B.1 Overview There is a legitimate concern that by making processes more open by adopting measures that make a system more explicable or more transparent that it also makes those systems more vulnerable to attack. The principle from crypto-security described by Auguste Kerckhoffs "A cryptosystem should be secure even if everything about the system, except the key, is public knowledge" [i.4] can be extended to AI systems. In applying Kerckhoffs' principle to AI the aim is that the purpose of algorithms, data and the intelligence model, when they are public do not impact on system security, where system security includes the ability to demonstrate and prove the explicability and transparency of the system. B.2 Model extraction In ETSI TR 104 221 [i.1] it is inferred that most AI systems are opaque, where the systems accept inputs, and generate outputs without ever revealing the internal logic, algorithms or parameters. In addition, training data sets, which effectively contain all the knowledge of the trained system, are also usually kept confidential. The role of transparency and explicability however challenges the inference of [i.1]. If opacity is removed in favour of transparency it can reasonably be asked: How transparent? The short answer is that it depends on context and some examples below can assist in determining to what extent an AI system can remain opaque, or its data sets remain confidential. EXAMPLE: An AI system that is categorized as High Risk under the EU's AI Act [i.5] can be required to undergo compliance testing against certain mandatory requirements and an ex-ante conformity assessment. In such cases it would be reasonable to expect the AI system to be fully open, at least to the assessors. NOTE: An open system does not infer an insecure or unsafe system. Rather by adopting Kerckhoffs' principle [i.4] the AI system is expected to be designed in such a way that it is secure and safe, and its secrets secret, whilst also being open. If the AI system is transparent and explicable it should not infer that it can be easily extracted. The intention is therefore to encourage transparency and explicability whilst at the same time offering assurance to developers that the model itself will not be open to abuse (e.g. by theft). Methods to achieve this are still under study and development. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 18 Annex C (informative): Data quality in AI/ML Many of the commonly perceived threats in AI/ML systems can be classified as arising from data quality issues. The aim of transparency and explicability as outlined in the present document is part of the quality metric of the system. The provisions recommended and identified in ETSI TR 104 048 [i.2] apply in support of element 2c of the static explicability analysis (see clause 5 of the present document). Table C.1 Documentation Element Element Short description 2c Method(s) used to determine data quality Methods and processes used in determining if the input data is a fair and accurate representation of the desired input. This should address how bias or preference is identified and corrected in the data input Common methods of data quality assessment include table C.2, where the AI/ML concern is noted. Table C.2 Metric Definition Role in AI/ML Accuracy Measures the number (and type) of errors in a dataset. Typically measured as a percentage of errors across all the records. Completeness Checks if all elements in a data record are complete. Consistency Measured across datasets to determine if the same data is presented in the same way. Timeliness Determines if the data is fresh (for the context it is consumed in). Uniqueness Tracks duplicate data with a view to eliminating duplicates. Whilst often a necessary constraint in relational databases there is often a different view in statistical analysis where a cleaned data source may actually give misleading results (there is some value in ensuring that complete records are not duplicated within single datasets but care has to be taken to validate duplication versus repetition). Validity The ISO 8000 series of standards also address data quality as identified by their titles below with the most relevant elements for transparency and explicability highlighted in bold type (these are not cited as explicit references but are listed in the bibliography): • ISO/TS 8000-1:2011: "Data quality - Part 1: Overview" • ISO 8000-2:2017: "Data quality - Part 2: Vocabulary" • ISO 8000-8:2015: "Data quality - Part 8: Information and data quality: Concepts and measuring" • ISO 8000-61:2016: "Data quality - Part 61: Data quality management: Process reference model" • ISO 8000-63:2019: "Data quality - Part 63: Data quality management: Process measurement" • ISO 8000-100:2016: "Data quality - Part 100: Master data: Exchange of characteristic data: Overview" • ISO 8000-102:2009: "Data quality - Part 102: Master data: Exchange of characteristic data: Vocabulary" (Withdrawn) ETSI ETSI TS 104 224 V1.1.1 (2025-03) 19 • ISO 8000-110:2009: "Data quality - Part 110: Master data: Exchange of characteristic data: Syntax, semantic encoding, and conformance to data specification" • ISO 8000-115:2017: "Data quality - Part 115: Master data: Exchange of quality identifiers: Syntactic, semantic and resolution requirements" • ISO 8000-120:2016: "Data quality - Part 120: Master data: Exchange of characteristic data: Provenance" • ISO 8000-130:2016: "Data quality - Part 130: Master data: Exchange of characteristic data: Accuracy" • ISO 8000-140:2016: "Data quality - Part 140: Master data: Exchange of characteristic data: Completeness" • ISO/TS 8000-150:2011: "Data quality - Part 150: Master data: Quality management framework" • ISO/TS 8000-311:2012: "Data quality - Part 311: Guidance for the application of product data quality for shape (PDQ-S)" It is suggested in ETSI TR 104 048 [i.2] that poisoning as an attack can be determined by identifying data values significantly outside of the norm for base data. However it is also known that influencing opinion, e.g. on social media and in news articles, does not require significant modification of data, but that data is stressed differently. Thus the methods of data quality assessment in ETSI TR 104 048 [i.2] may not always be practical if such filtering also misidentifies long term, or short term, actual variation in data. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 20 Annex D (informative): Document template for explicability and transparency D.1 Static Explicability template The following template proforma is taken from table 1 in clause 5.1 with the guidance given in clause 5 applying. Table D.1: Static explicability statement template Documentation Element Element Short description 1 Statement of system purpose 2a Identification of data source(s) 2b Purpose of data source(s) (in support of system purpose) 2c Method(s) used to determine data quality 3 Identity of liable party As stated in clause 5 the purpose of explicability is to allow a lay person (i.e. not a professional programmer or system analyst) to gain a reasonable understanding of the main data flows and processing steps in the program. D.2 Run-time Explicability template The following template proforma builds from the definitions and expectations described in clause 6. Table D.2: run time explicability statement template Documentation Element Element Short description RTE-1 Static explicability statement RTE-2 What process does data undergo between acquisition and curation? RTE-3 What are the metrics that determine change in the learning/weighting of data? RTE-4 Identification of events to be logged RTE-5 Identification of performance target and associated metrics RTE-6 Identification of liable party (if different from that identified in the static explicability documentation) D.3 Data transparency template The template for documenting data transparency is taken from clause 7. ETSI ETSI TS 104 224 V1.1.1 (2025-03) 21 Table D.3: Transparency statement template Documentation Element Element Short description T-1 Static explicability statement T-2 Run time explicability statement For each data source T-3a Verified identification of source of data T-3b Verified proof of liability of data source T-3c Verified proof of consent to use ETSI ETSI TS 104 224 V1.1.1 (2025-03) 22 Annex E (informative): Bibliography E.1 Data Quality • OECD: "Quality Framework and Guidelines for OECD Statistical Activities", Version 2011/1. • ISO/TS 8000-1:2011: "Data quality - Part 1: Overview". • ISO 8000-2:2017: "Data quality - Part 2: Vocabulary". • ISO 8000-8:2015: "Data quality - Part 8: Information and data quality: Concepts and measuring". • ISO 8000-61:2016: "Data quality - Part 61: Data quality management: Process reference model". • ISO 8000-63:2019: "Data quality - Part 63: Data quality management: Process measurement". • ISO 8000-100:2016: "Data quality - Part 100: Master data: Exchange of characteristic data: Overview". • ISO 8000-102:2009: "Data quality - Part 102: Master data: Exchange of characteristic data: Vocabulary" (Withdrawn). • ISO 8000-110:2009: "Data quality - Part 110: Master data: Exchange of characteristic data: Syntax, semantic encoding, and conformance to data specification". • ISO 8000-115:2017: "Data quality - Part 115: Master data: Exchange of quality identifiers: Syntactic, semantic and resolution requirements". • ISO 8000-120:2016: "Data quality - Part 120: Master data: Exchange of characteristic data: Provenance". • ISO 8000-130:2016: "Data quality - Part 130: Master data: Exchange of characteristic data: Accuracy". • ISO 8000-140:2016: "Data quality - Part 140: Master data: Exchange of characteristic data: Completeness". • ISO/TS 8000-150:2011: "Data quality - Part 150: Master data: Quality management framework". • ISO/TS 8000-311:2012: "Data quality - Part 311: Guidance for the application of product data quality for shape (PDQ-S)". • ETSI TR 104 222: "Securing Artificial Intelligence (SAI); Mitigation Strategy Report". • ISO/IEC TR 24028: "Information technology - Artificial intelligence - Overview of trustworthiness in artificial intelligence". • ISO/IEC 22989: "Artificial intelligence concepts and terminology". ETSI ETSI TS 104 224 V1.1.1 (2025-03) 23 History Document history V1.1.1 March 2023 Publication as ETSI GR SAI 007 V1.1.1 March 2025 Publication |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 1 Scope | The present document describes an ensemble of cyber security specifications and other materials, especially the ETSI Critical Security Controls in ETSI TR 103 305-1 [i.9] that can be applied to support NIS2 Directive [i.1] requirements by EU Member States and affected essential and important entities. The present document also considers, and makes reference to, the work being done by ETSI ESI on Trust Services. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 2 References | |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 2.1 Normative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found at https://docbox.etsi.org/Reference. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long-term validity. The following referenced documents are necessary for the application of the present document. Not applicable. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 2.2 Informative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. NOTE 1: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long term validity. The following referenced documents are not necessary for the application of the present document but they assist the user with regard to a particular subject area. [i.1] Directive (EU) 2022/2555 of the European Parliament and of the Council of 14 December 2022 on measures for a high common level of cybersecurity across the Union, amending Regulation (EU) No 910/2014 and Directive (EU) 2018/1972, and repealing Directive (EU) 2016/1148 (NIS 2 Directive) (Text with EEA relevance). [i.2] Regulation (EU) No. 910/2014 of the European Parliament and of the Council of 23 July 2014 on electronic identification and trust services for electronic transactions in the internal market and repealing Directive 1999/93/EC. [i.3] Directive (EU) 2016/1148 of The European Parliament and of The Council of 6 July 2016 concerning measures for a high common level of security of network and information systems across the Union. [i.4] Resolution (EC) 13084/1/20, Council Resolution on Encryption - Security through encryption and security despite encryption. [i.5] Recommendation 2003/361/EC, Commission Recommendation of 6 May 2003 concerning the definition of micro, small and medium-sized enterprises (Text with EEA relevance). [i.6] 2020/0365 (COD), COM(2020) 829 Final: "Proposal for a directive of the European Parliament and of the Council on the resilience of critical entities". ETSI ETSI TS 103 992 V1.1.1 (2024-05) 8 [i.7] Directive (EU) 2018/1972 of the European Parliament and of the Council of 11 December 2018 establishing the European Electronic Communications Code (Recast) (Text with EEA relevance) Text with EEA relevance. [i.8] ETSI TR 103 456: "CYBER; Implementation of the Network and Information Security (NIS) Directive". [i.9] ETSI TR 103 305-1: "Cyber Security (CYBER); Critical Security Controls for Effective Cyber Defence; Part 1: The Critical Security Controls". [i.10] ETSI TR 103 305-3: "Cyber Security (CYBER); Critical Security Controls for Effective Cyber Defence; Part 3: Internet of Things Sector". [i.11] ETSI TR 103 305-4: "Cyber Security (CYBER); Critical Security Controls for Effective Cyber Defence; Part 4: Facilitation Mechanisms". [i.12] ETSI TR 103 305-5: "Cyber Security (CYBER); Critical Security Controls for Effective Cyber Defence; Part 5: Privacy and personal data protection enhancement". [i.13] ETSI TR 103 331: "Cyber Security (CYBER); Structured threat information sharing". [i.14] Recommendation ITU-T X.1500, Amd. 12: "Overview of cybersecurity information exchange" (03/2018). [i.15] Regulation (EU) 2019/1150 of the European Parliament and of the Council of 20 June 2019 on promoting fairness and transparency for business users of online intermediation services (Text with EEA relevance). [i.16] ETSI EN 319 401: "Electronic Signatures and Infrastructures (ESI); General Policy Requirements for Trust Service Providers". [i.17] ETSI EN 319 403: "Electronic Signatures and Infrastructures (ESI); Trust Service Provider Conformity Assessment; Part 1: Requirements for conformity assessment bodies assessing Trust Service Providers". [i.18] MISP Threat Sharing. [i.19] ENISA: "Orchestration of CSIRT Tools", December 2019. [i.20] ETSI Security Conference 2022: "H2020 Project MEDINA". [i.21] NIST: "OSCAL: the Open Security Controls Assessment Language". [i.22] ETSI GR ETI 006: "Encrypted Traffic Integration (ETI); Implementation of the EU Council Resolution on Encryption". [i.23] OASIS CACAO Security Playbooks Version 1.0. [i.24] FIRST: "CSIRT Services Framework". [i.25] FIRST: "Traffic Light Protocol (TLP)". [i.26] FIRST: "Guidelines and Practices for Multi-Party Vulnerability Coordination and Disclosure". [i.27] FIRST: "Common Vulnerability Scoring System (CVSS) v3.1". [i.28] ETSI TR 103 838: "Cyber Security; Guide to Coordinated Vulnerability Disclosure". [i.29] ISO/IEC 29147: "Information technology -- Security techniques -- Vulnerability disclosure". [i.30] ISO/IEC 30111: "Information technology -- Security techniques -- Vulnerability handling processes". [i.31] ISO/IEC TR 5895: "Cybersecurity -- Multi-party coordinated vulnerability disclosure and handling". ETSI ETSI TS 103 992 V1.1.1 (2024-05) 9 [i.32] 2022/0272 (COD), COM(2022) 454 final: "Proposal for a Regulation of the European Parliament and of the Council on horizontal cybersecurity requirements for products with digital elements and amending Regulation (EU) 2019/1020". [i.33] OASIS Common Alerting Protocol Version 1.2. [i.34] NIST Special Publication 800-207: "Zero Trust Architecture", August 2020. [i.35] National Security Agency, PP-21-0191: "Embracing a Zero Trust Security Model", February 2021. [i.36] 2020/0266 (COD), COM(2020) 595 final: "Proposal for a Regulation of the European Parliament and of the Council on digital operational resilience for the financial sector and amending Regulations (EC) No 1060/2009, (EU) No 648/2012, (EU) No 600/2014 and (EU) No 909/2014". [i.37] Regulation (EU) 2022/2065 of the European Parliament and of the Council of 19 October 2022 on a Single Market For Digital Services and amending Directive 2000/31/EC (Digital Services Act) (Text with EEA relevance). [i.39] Regulation (EU) No 1025/2012 of the European Parliament and of the Council of 25 October 2012 on European standardisation, amending Council Directives 89/686/EEC and 93/15/EEC and Directives 94/9/EC, 94/25/EC, 95/16/EC, 97/23/EC, 98/34/EC, 2004/22/EC, 2007/23/EC, 2009/23/EC and 2009/105/EC of the European Parliament and of the Council and repealing Council Decision 87/95/EEC and Decision No 1673/2006/EC of the European Parliament and of the Council (Text with EEA relevance). [i.40] 2022/0021 (COD), COM(2022) 32 final: "Proposal for a Regulation of the European Parliament and of the Council amending Regulation (EU) No 1025/2012 as regards the decisions of European standardisation organisations concerning European standards and European standardisation deliverables". [i.42] Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, and repealing Directive 95/46/EC (General Data Protection Regulation) (Text with EEA relevance). [i.44] Regulation (EU) 2022/1925 of the European Parliament and of the Council of 14 September 2022 on contestable and fair markets in the digital sector and amending Directives (EU) 2019/1937 and (EU) 2020/1828 (Digital Markets Act) (Text with EEA relevance). [i.45] MITRE ATT&CK®. [i.46] EU Council Resolution on Encryption Security through encryption and security despite encryption, adopted 14 December 2020. [i.47] ETSI TR 103 954: "Cyber Security (CYBER); Critical Security Controls for Effective Cyber Defence; Mobile Communications Sector". [i.48] ETSI TR 103 959: "Cyber Security (CYBER); Critical Security Controls for Effective Cyber Defence; Cloud Sector". [i.49] NIST Special Publication 800-160, Volume 2, Revision 1: "Developing Cyber-Resilient Systems: A Systems Security Engineering Approach", November 2022. [i.50] IEC 62443 series: "Industrial communication networks - Network and system security". [i.51] ISO/IEC 27000 family: "Information security management". [i.52] ETSI GR ETI 002: "Encrypted Traffic Integration (ETI); Requirements definition and analysis". [i.53] ANSSI: "Le modèle Zero Trust". [i.54] BSI: "Bundesinnenministerium: Zero-Trust-Architektur wird angestrebt". ETSI ETSI TS 103 992 V1.1.1 (2024-05) 10 |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 3 Definition of terms, symbols and abbreviations | |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 3.1 Terms | For the purposes of the present document, the terms given in the NIS2 Directive [i.1] and the following apply: NOTE: It should be noted that the NIS2 created definitions of terms that are nuanced and may vary with those used elsewhere. Because the present document concerns the implementation of the NIS2 Directive, it builds on the definitions provided by NIS2 on key terms. controls workbench: tool to inform users and enterprises exactly which Control Safeguards to implement to achieve desired or required levels of security and risk NOTE: As defined in [i.11]. critical security control: prioritized set of actions to protect information assets from threats, using technical or procedural Safeguards NOTE: As defined in [i.9]. cybersecurity: activities necessary to protect network and information systems, the users of such systems, and other persons affected by cyber threats NOTE: As defined in [i.1]. cyber threat: any potential circumstance, event or action that could damage, disrupt or otherwise adversely impact network and information systems, the users of such systems and other persons NOTE: As defined in [i.1]. impact: harm that may be suffered when a threat compromises an information asset incident: any event compromising the availability, authenticity, integrity or confidentiality of stored, transmitted or processed data or of the services offered by, or accessible via, network and information systems NOTE: As defined in [i.1]. incident handling: all actions and procedures aiming at prevention, detection, analysis, and containment of, response to, and recovery from an incident NOTE: As defined in [i.1]. large-scale cybersecurity incident: incident whose disruption exceeds a Member State's capacity to respond to it or with a significant impact on at least two Member States NOTE: As defined in [i.1]. near miss: event that could have compromised the availability, authenticity, integrity or confidentiality of stored, transmitted or processed data or of the services offered by, or accessible via, network and information systems, but was successfully prevented from transpiring or did not materialize NOTE: As defined in [i.1]. risk: potential for loss or disruption caused by an incident and is to be expressed as a combination of the magnitude of such loss or disruption and the likelihood of occurrence of that incident NOTE: As defined in [i.1]. risk analysis: process of estimating the likelihood that an event will create an impact and includes as necessary components, the foreseeability of a threat, the expected effectiveness of Control Safeguards, and an evaluated result risk assessment: comprehensive project that evaluates the potential for harm to occur within a scope of information assets, controls, and threats ETSI ETSI TS 103 992 V1.1.1 (2024-05) 11 risk management: process for analysing, mitigating, overseeing, and reducing risk safeguard: technical or procedural protections that prevent or detect threats against information assets that are implementations of a Critical Security Control NOTE: As defined in [i.9]. security of network and information systems: ability of network and information systems to resist, at a given level of confidence, any event that may compromise the availability, authenticity, integrity or confidentiality of stored or transmitted or processed data or of the services offered by, or accessible via, those network and information systems NOTE: As defined in [i.1]. security playbook: workflow for security orchestration containing a set of steps to be performed based on a logical process and may be triggered by an automated or manual event or observation, and provides guidance on how to address a certain security event, incident, problem, attack, or compromise NOTE: As defined in [i.23]. significant cyber threat: cyber threat which, based on its technical characteristics, can be assumed to have the potential to severely impact the network and information systems of an entity or its users by causing considerable material or non-material losses NOTE: As defined in [i.1]. vulnerability: weakness, susceptibility or flaw of ICT products or ICT services that can be exploited by a cyber threat NOTE: As defined in [i.1]. zero trust security model: security model consisting of a set of system design principles, and a coordinated cybersecurity and system management strategy based on an acknowledgement that threats exist both inside and outside traditional network boundaries and eliminates implicit trust in any one element, node, or service and instead requires continuous verification of the operational picture via real-time information fed from multiple sources to determine access and other system responses NOTE: As defined in [i.35]. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 3.2 Symbols | Void. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 3.3 Abbreviations | For the purposes of the present document, the following abbreviations apply: AI Artificial Intelligence ANSSI Agence Nationale de la Sécurité des Systèmes d'Information (National Agency for the Security of Information Systems) (France) Art. Article ATT&CK Adversarial Tactics, Techniques, and Common Knowledge BSI Bundesamt für Sicherheit in der Informationstechnik (Federal Office for Information Security) (Germany) CAP Common Alerting Protocol CERT Computer Emergency Response Team CISA Cybersecurity and Infrastructure Security Agency (USA) CSC Critical Security Controls CSIRT Computer Security Incident Response Team CTI Cyber Threat Intelligence CVD Coordinated Vulnerability Disclosure CVSS Common Vulnerability Scoring System CyCLONe Cyber Crisis Liaison Organization Network DMARC Domain-based Message Authentication, Reporting & Conformance EEA European Economic Area ETSI ETSI TS 103 992 V1.1.1 (2024-05) 12 EEC European Economic Community eIDAS electronic Identification, Authentication and Trust Services ENISA European union agency for Network and Information Security ESI Electronic Signatures and Infrastructures ETI Encrypted Traffic Integration EU European Union EUCS European Cybersecurity Certification Scheme for Cloud Service FIRST Forum of Incident Response and Security Teams GDPR General Data Protection Regulation [i.42] ICT Information and Communication Technology IEC International Electrotechnical Commission ISAC Information Sharing and Analysis Centre ISG Industry Specification Group ISO International Organization for Standardization ITU-T International Telecommunication Union Telecommunications Standardization Sector MANRS Mutually Agreed Norms for Routing Security MISP Malware Information Sharing Platform NCSC National Cyber Security Centre (UK) NIS Network and Information Security NIST National Institute of Standards and Technology OASIS Organization for the Advancement of Structured Information Standards OSCAL Open Security Controls Assessment Language para. paragraph PSIRT Product Security Incident Response Team SME Small and Medium Enterprise TF-CSIRT Task Force on Computer Security Incident Response Teams ZT Zero Trust |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4 Measures common to Directive requirements | |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.1 Description | The NIS2 Directive [i.1] requires the implementation of cybersecurity measures by different types of stakeholders that are articulated in different contexts in the individual articles and annexes that are identified in clause 5, below. It should be noted that the entities included in the annexes are subject to change under existing and subsequent EU legislative instruments (see [i.15], [i.37] and [i.44]). These measures can be distilled into ten categories that are described below, together with the means for effectively implementing them - largely through the Control Safeguards of the Critical Security Controls and facilitating mechanisms (see [i.9], [i.10] and [i.11]). Several sector-specific implementations are also available (see [i.10], [i.47] and [i.48]). A mapping of the NIS2 Directive [i.1] requirements identified in clause 4 and the Critical Security Control Safeguards [i.9] is provided in Annex A. For information sharing and reporting requirements, as well as reporting vulnerabilities, additional capabilities are needed, for which guidance is provided. 4.2 Exchanging best practices, information sharing, and reporting The treatment of exchanging best practices, information sharing, and reporting is threaded though most of the Directive provisions and expressly found in the Cooperation Group (Art. 14), and information sharing (Arts. 29 and 30) [i.1]. Exchanging best practices, information sharing, and reporting can be addressed through the use of Critical Security Control Safeguards which establish a prioritized set of actions to protect the assets through the use of technical and procedural Safeguards that are measurable and can establish risk levels. Specific Safeguards for exchanging best practices, information sharing and reporting include: - Control 4 - Secure Configuration of Enterprise Assets and Software; - Control 9 - Email and Web Browser Protections; - Control 14 - Security Awareness and Skills Training; ETSI ETSI TS 103 992 V1.1.1 (2024-05) 13 - Control 16 - Application Software Security; - Control 17 - Incident Response Management; and - Control 18 - Penetration Testing. The implementation group levels include IG1 (SME), IG2 (medium enterprise) and IG3 (large and critical enterprises) [i.9]. The proposed Control safeguards do not address how essential and important entities can find and get in touch with relevant communities to perform information sharing. Forum such as FIRST and TF-CSIRT, national platforms (depending on national policies), and sector specific ISACs can all provide information exchange, be it in the form of physical and online meetings, discussion spaces, and online tools, e.g. for Cyber Threat Intelligence (CTI). Control 4 consists of twelve Safeguard subsets that describe the procedures and tools at three implementation group security levels to establish and maintain the secure configuration of enterprise assets (end-user devices, including portable and mobile; network devices; non-computing/IoT devices; and servers) and software (operating systems and applications). The different subsets include best practices, information sharing and reporting specifics that will vary by implementation group level, assumed risk, and industry/regulatory requirements. Control 9 consists of seven Safeguard subsets that improve protections and detections of threats from email and web vectors, as these are opportunities for attackers to manipulate human behaviour through direct engagement. The different Safeguard subsets include best practices, information sharing and reporting specifics that will vary by implementation group level, assumed risk, and industry/regulatory requirements. Also recommended in conjunction with this control is initiating Domain-based Message Authentication, Reporting and Conformance (DMARC) which helps reduce spam and phishing activities. Installing an encryption tool to secure email and communications adds another layer of user and network-based security. Control 14 consists of nine Safeguard subsets that establish and maintain a security awareness program to influence behaviour among the workforce to be security conscious and properly skilled to reduce cybersecurity risks to the enterprise. The different Safeguard subsets include best practices that are largely the same except for advanced skill sets appropriate for higher implementation group levels, assumed risk, and industry/regulatory requirements. Control Safeguard Subsets 14.6 and 14.7 that involve training workforce members to recognize potential incidents and verify software patch installations are especially relevant. Control 16 consists of 14 Safeguard subsets that manage the security life cycle of in-house developed, hosted, or acquired software to prevent, detect, and remediate security weaknesses before they can impact the enterprise. The different Safeguard subsets include best practices, information sharing and reporting specifics that will vary by implementation group level, assumed risk, and industry/regulatory requirements. Control 17 consists of nine Safeguard subsets that establish a program to develop and maintain an incident response capability (e.g. policies, plans, procedures, defined roles, training and communications) to prepare, detect, and quickly respond to an attack. The different Safeguard subsets include best practices, information sharing and reporting specifics that will vary by implementation group level, assumed risk, and industry/regulatory requirements. Control 18 consists of five Safeguard subsets that test the effectiveness and resiliency of enterprise assets through identifying and exploiting weaknesses in controls (people, processes, and technology), and simulating the objectives and actions of an attacker. The different Safeguard subsets include best practices, information sharing and reporting specifics that will vary by implementation group level, assumed risk, and industry/regulatory requirements. In addition to the above Control Safeguards, an array of important, continuously-evolving relevant tools are found in the Control Facilitation Mechanisms, and include: the Community Defense Model, Risk Assessment Methods, playbooks, and applications to specific provider communities and SMEs [i.10] and [i.11]. The Privacy enhancements for each Control Safeguard are also valuable [i.12]. Notably, the Community Defense Model includes the use of playbooks which have emerged as a set of structured steps to be performed based on a logical process and may be triggered by an automated or manual event or observation that provides guidance on how to address a certain security event, incident, problem, attack, or compromise [i.23]. The implementation of NIS2 through cybersecurity controls includes an enormous amount of continuously exchanged structured information, The specific interfaces and protocols used worldwide for the actual structuring and exchanges of the information vary. The principle recommended platforms that are widely used by industry and government for exchanging can be found in ETSI TR 103 331 [i.13]. Of special note for structured threat information sharing is the emergence of MISP as the leading Open Source Threat Intelligence Platform, including an open standard for powering intelligence and information exchange, sharing and modelling among a number of different fields [i.18]. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 14 Those fields include cybersecurity intelligence, threat intelligence, financial fraud, vulnerability information, digital forensic and incident response, among others. ENISA promotes the use of MISP and its tools for CSIRT orchestration [i.19]. Therefore, the use of MISP [i.18] is recommended. The evolution and implementation of CSIRT structured Information sharing practices and standards as well as the exchange of information has been significantly advanced by the Forum of Incident Response and Security Teams (FIRST) and its CSIRT Services Framework, which includes Product Security Incident Response Teams (PSIRTs) in the form of an All Services Frameworks [i.24]. FIRST also maintains the Traffic Light Protocol, which is used as an indicator of confidentiality level in many information sharing scenarios [i.25]. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.3 Addressing cyber threats | The treatment of cyber threats in the NIS2 Directive [i.1] can be found in the following articles: - Scope (Art. 2); - national strategies (Art. 7); - CSIRT tasking (Arts. 10, 11); - national cooperation (Art. 13); - Cooperation Group (Art. 14); - CSIRTs network (Art. 15); - CyCLONE (Art. 16); - Risk Management (Art. 21); - critical supply chains (Art. 22); - reporting obligations (Art. 23); - information sharing (Art. 29); and - voluntary notification (Art. 30). According to [i.1], a cybersecurity threat refers to any potential circumstance, event or action that could damage, disrupt or otherwise adversely impact network and information systems, the users of such systems and other persons. Threats can be external or internal, and of human, environmental or technological origin. Human threats can be malicious or non malicious, as well as accidental or intentional. Threat effects can include: 1) destruction, corruption, disclosure, or illegal usage of information; 2) denial of use; and/or 3) elevation of information system privileges. Cyber threats can be addressed through the use of Critical Security Control Safeguards [i.9], which establish a prioritized set of actions to protect the assets through the use of technical and procedural Safeguards that are measurable and can establish risk levels. Specific threat-related Safeguards are found in ten Critical Security Controls in [i.9] and include: - Control 5 - Account Management; - Control 7 - Continuous Vulnerability Management; - Control 8 - Audit Log Management; - Control 9 - Email and Web Browser Protections; - Control 10 - Malware Defences; - Control 11 - Data Recovery; ETSI ETSI TS 103 992 V1.1.1 (2024-05) 15 - Control 13 - Network Monitoring and Defence; - Control 14 - Security Awareness and Skills Training; - Control 16 - Application Software Security; and - Control 17 - Incident Response Management. In addition to the above Critical Security Control Safeguards, an array of important, continuously-evolving, relevant tools are specified in the Control Facilitation Mechanisms, and include: the Community Defense Model, Risk Assessment Methods, playbooks, and applications to specific provider communities and SMEs [i.5], [i.10] and [i.11]. For cloud data centre implementations, hardened images identified in the Facilitation Mechanisms [i.11] are especially important. The Privacy enhancements for each Control Safeguard [i.12] are also valuable. The defence of increasingly dispersed and complex networks from sophisticated cyber threats to secure sensitive data, systems, and services can be significantly enhanced by embracing a Zero Trust Security Model necessary to deploy and operate a system engineered according to Zero Trust principles. A breach is assumed to be inevitable or likely already occurred, so it constantly limits access to only what is needed and looks for anomalous or malicious activity. Zero Trust embeds comprehensive security monitoring; granular risk-based access controls; and system security automation in a coordinated manner throughout all aspects of the infrastructure in order to focus on protecting critical assets (resources - including data - and services) in real-time within a dynamic threat environment. This data-centric security model allows the concept of least-privileged access to be applied for every access decision, allowing or denying access to resources based on the combination of several contextual factors. Determining anomalous, out-of-ordinary or malicious activity versus expected, modelled behaviour can further be used to lower the risk of (unintentional) data exfiltration or leaking of sensitive data. To be fully effective to minimize risk and enable robust and timely responses, Zero Trust principles and concepts need to permeate most aspects of the network and its operations ecosystem. To enable real-time risk determination and deliver ongoing protection in evolving organizations, contextual data is analysed and frequently re-evaluated with machine learning algorithms. Guidance from national and industry security authorities is essential, per [i.3], [i.4] and [i.35] (see annex B). Therefore, the implementation of a Zero Trust Security Model is highly recommended. NOTE: Clause 4.9 of the present document stresses the importance of encryption. Encrypted traffic can still be categorized as valid or possibly malicious and subject to remediation (per [i.3] and [i.4]). |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.4 Addressing incidents and near misses | The treatment of incidents and near misses in the NIS2 Directive [i.1] can be found in the following articles: - Scope (Art. 2); - national strategies (Art. 7); - national cooperation (Art. 3); - Cooperation Group (Art. 14); - CSIRTs network (Art. 15); - reporting obligations (Art. 23); and - voluntary notification (Art. 30). According to [i.1], a cybersecurity incident refers to any event compromising the availability, authenticity, integrity or confidentiality of stored, transmitted or processed data or of the related services offered by, or accessible via, network and information systems. Because near misses are part of voluntary notification arrangements under Art. 30 of [i.1], entities that are uncertain about the criteria for the characterization of near misses could ask the national CSIRT for guidance. Cyber incidents can be addressed through the use of Critical Security Control Safeguards [i.9], which establish a prioritized set of actions. Specific incident related Safeguards are found in ten Critical Security Controls in [i.9] and include: - Control 1 - Inventory and Control of Enterprise Assets; ETSI ETSI TS 103 992 V1.1.1 (2024-05) 16 - Control 3 - Data Protection; - Control 4 - Secure Configuration of Enterprise Assets and Software; - Control 8 - Audit Log Management; - Control 10 - Malware Defences; - Control 11 - Data Recovery; - Control 13 - Network Monitoring and Defence; - Control 14 - Security Awareness and Skills Training; - Control 15 - Service Provider Management; and - Control 17 - Incident Response Management. In addition to the above Critical Security Control Safeguards, an array of important, continuously evolving relevant tools are found in the Control Facilitation Mechanisms, and include: the Community Defense Model, Risk Assessment Methods, playbooks, and applications to specific provider communities and SMEs (per [i.10] and [i.11]). The Privacy enhancements for each Control Safeguard [i.12] are also valuable. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.5 Addressing vulnerabilities | The treatment of vulnerabilities in the NIS2 Directive [i.1] can be found in the following articles: - national strategies (Art. 7); - coordinated vulnerability disclosure (Art. 6); - CSIRT tasking (Arts. 9, 10); - Cooperation Group (Art. 14); - CSIRT's network (Art. 15); - risk management (Art. 21); - critical supply chains (Art. 22); and - information sharing (Art. 29). Vulnerabilities can be addressed through the use of Critical Security Control Safeguards [i.9], which establish a prioritized set of actions and include: - Control 1 - Inventory and Control of Enterprise Assets; - Control 3 - Data Protection; - Control 7 - Continuous Vulnerability Management; - Control 8 - Audit Log Management; - Control 10 - Malware Defences; - Control 14 - Security Awareness and Skills Training; - Control 16 - Application Software Security; - Control 17 - Incident Response Management; and - Control 18 - Penetration Testing. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 17 In addition to the above Critical Security Control Safeguards, an array of important, continuously evolving relevant tools are found in the Control Facilitation Mechanisms, and include: the Community Defense Model, Risk Assessment Methods, playbooks, and applications to specific provider communities and SMEs (per [i.10] and [i.11]). For cloud data centre implementations, hardened images identified in the Facilitation Mechanisms [i.11] are especially important. The Privacy enhancements for each Control Safeguard [i.12] are also valuable. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.6 Instituting risk management measures | Risk management in the NIS2 directive [i.1] encompasses a broad array of subjects that include: - risk analysis; - incident handling; - business continuity; - supply chain security; - systems acquisition, development, and maintenance security; - certification, testing and auditing; - cyber hygiene and training; - use of cryptography and encryption; and - authentication and secured communications. The instituting of risk management measures in the NIS2 Directive [i.1] can be found in: - Subject Matter (Art. 1); - Scope (Art. 2); - peer review (Art. 16); - governance (Art. 17); - risk management measures (Art. 21); - critical supply chains (Art. 22); and - supervision and enforcement for essential and important entities (Arts. 29 and 30). Risk management measures can be addressed through the use of Critical Security Control Safeguards [i.9], which establish a prioritized set of actions and include: - Control 1 - Inventory and Control of Enterprise Assets; - Control 2 - Inventory and Control of Software Assets; - Control 4 - Secure Configuration of Enterprise Assets and Software; - Control 5 - Account Management; - Control 6 - Access Control Management; - Control 7 - Continuous Vulnerability Management; - Control 9 - Email and Web Browser Protections; - Control 10 - Malware Defences; - Control 11 - Data Recovery; - Control 12 - Network Infrastructure Management; ETSI ETSI TS 103 992 V1.1.1 (2024-05) 18 - Control 13 - Network Monitoring and Defence; - Control 14 - Security Awareness and Skills Training; - Control 15 - Service Provider Management; - Control 16 - Application Software Security; and - Control 18 - Penetration Testing, Advanced or Multifactor Authentication, and Secured Communications. In addition to the above Control Safeguards, an array of important, continuously evolving relevant tools are found in the Control Facilitation Mechanisms, and include: the Community Defense Model, Risk Assessment Methods, playbooks, and applications to specific provider communities and SMEs (per [i.10] and [i.11]). For cloud data centre implementations, hardened images identified in the Facilitation Mechanisms [i.11] are especially important. For managing “basic computer hygiene practice” (as stated in Article 18 of the NIS2 Directive [i.1]), the Critical Security Control Malware Defences Safeguard [i.9] provides a good foundation and support. However, considering the need for effective protection against ever evolving zero-day exploits by sophisticated threat actors, additional Control Safeguards supported by AI based or non-signature-based malware protection is required. Such predictive malware protection has advantages over signature-based protection in that its protection is typically more current and requires fewer updates. The Privacy enhancements for each Control Safeguard [i.12] are also valuable. 4.7 Capacity, awareness-raising initiatives, trainings, exercises and skills The treatment of capacity, awareness-raising initiatives, trainings, exercises and skills in the NIS2 Directive [i.1] can be found in the following articles: - national strategy (Art. 7); - crisis management (Art. 7); - CSIRT tasking (Arts. 10, 11); - Cooperation Group (Art. 14); - CSIRT’s network (Art. 15); - CyCLONe (Art. 16); - governance (Art. 17); - standardization (Art. 22); and - information sharing (Art. 29). Capacity, awareness-raising initiatives, trainings, exercises and skills can be addressed through the use of Critical Security Control Safeguards which establish a prioritized set of actions and include: - Control 14 - Security Awareness and Skills Training; - Control 17 - Incident Response Management; and - Control 18 - Penetration Testing [i.9]. In addition to the above Control Safeguards, an array of important, continuously evolving relevant tools are found in the Control Facilitation Mechanisms, and include: the Community Defense Model, Risk Assessment Methods, playbooks, and applications to specific provider communities and SMEs (per [i.10] and [i.11]). The Privacy enhancements for each Control Safeguard [i.12] are also valuable. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 19 |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.8 Standards and technical specifications | The treatment of standards and technical specifications in the NIS2 Directive [i.1] can be found in the following articles: - Cooperation Group (Art. 14); - CSIRT's network (Art. 15); - risk management (Art. 21); - critical supply chains (Art. 22); and - standardization (Art. 25). Standards and technical specifications can be addressed through the use of Critical Security Control Safeguards in their entirety. The ETSI Critical Security Controls are also acknowledged by Recommendation ITU-T X.1500 [i.14], ITU-T's basic global intergovernmental standard for cybersecurity information exchange, where it states Critical Security Controls are a global technique "to detect, prevent, respond, and mitigate damage from the most common to the most advanced of cyber attacks…reflect[ing] the combined knowledge of actual attacks and effective defences". In addition to the Critical Security Control Safeguards, an array of important, continuously-evolving, relevant tools are found in the Control Facilitation Mechanisms, and include: the Community Defense Model, Risk Assessment Methods, playbooks, and applications to specific provider communities and SMEs (per [i.10] and [i.11]). The Privacy enhancements for each Critical Security Control Safeguard [i.12] should be provided. Especially significant among the Facilitation Mechanisms are the Mappings to every other commonly used frameworks, controls, and standards use by different industry sectors and nations. The Critical Security Controls also implement: 1) the MITRE Adversarial Tactics, Techniques, and Common Knowledge (ATT&CK®) framework [i.45] which enables expression of any attack type as a set of attack techniques, known as attack patterns, and more definitive risk management measures; and 2) the Open Security Controls Assessment Language (OSCAL) [i.21] which facilitates interoperability among Control frameworks and Safeguards. The creation of the Controls Workbench as a means to deal with the complexity is notable as it enables each implementation to select all the relevant jurisdiction, sector/national, context and risk variables to facilitate the NIS2 Directive [i.1] requirements and identify specific Critical Security Control Safeguards. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.9 Trust Services measures | The treatment of Trust Services and their provision in the NIS2 Directive [i.1] can be found in the following articles: - Scope (Art. 2); - Definitions (Art. 4); - risk management (Art. 21); - reporting (Art. 23); - certification schemes (Art. 21); and - Digital Infrastructure sectors of high criticality (Annex I, Sector 8). Trust services providers should take appropriate and proportionate measures to manage the risks posed to their services, including in relation to customers and reliant third parties, and to report security incidents under the NIS2 Directive [i.1] pursuant to ETSI EN 319 401 [i.16] and ETSI EN 319 403 [i.17] that implement eIDAS Regulation [i.2] and [i.37]. The NIS2 Directive [i.1] provisions provide for additional actions by Member States, the Commission, and CSIRTs: - The requirements found in the NIS2 Directive [i.1] for trust services risk management, including privacy and security, are already extensively treated by eIDAS and supporting ETSI EN 319 401 [i.16] and [i.42]. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 20 - NIS2 Directive [i.1] provisions relating to incident reporting, management, and training requirements are also found in eIDAS and extensively treated in the related ETSI EN 319 401 [i.16]. The NIS2 certification requirements are similarly found in ETSI EN 319 403 [i.17]. Any new NIS2 Directive [i.1] requirements should be addressed in conformance with the existing ETSI EN 319 401 [i.16]. This approach includes any certification framework established for the NIS2 Directive [i.1]. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 4.10 Encryption measures | The NIS2 Directive [i.1] includes an array of new actions and requirements related to encryption that is consonant with [i.46]. ETSI's ISG on Encrypted Traffic Integration (ETI) has published ETSI GR ETI 006 [i.22] specifically relating to matters relating to the [i.46] and the NIS2 Directive [i.1]. ETSI ISG ETI has also published ETSI GR ETI 002 [i.52], which identifies requirements for allowing encrypted traffic integration across an abstracted network architecture, informed in part by a Zero Trust Security Model. In addition, the NIS2 Directive [i.1] sets the risk management measures for regulated entities (see Article 18 of [i.1]) and promotes or mandates where necessary the deployment and use of end-to-end encryption as a possible risk mitigation, mindful of the potential adverse effects. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 5 Implementation of NIS2 Directive | |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 5.1 Description | The present clause suggests how requirements of the NIS2 Directive [i.1] can be effectively met using ETSI Critical Control Safeguards [i.9], [i.10], [i.11] and [i.12] together with associated facilitation mechanism, privacy enhancements, and sector guides. NOTE: Articles of the NIS2 Directive that treat legislative process, jurisdiction, enforcement, and other purely legal matters are not addressed in the present document. EXAMPLE: The classification of very large online platforms can be treated as essential entities under the Digital Services Act [i.37]. Additionally, the treatment of personal data under the GDPR [i.42], which applies to multiple NIS2 capability requirements, can alter their implementation. The Critical Security Control Safeguard groups identified in [i.9], [i.10], [i.11] and [i.12] are as follows: 1) Inventory and Control of Enterprise Assets 2) Inventory and Control of Software Assets 3) Data Protection 4) Secure Configuration of Enterprise Assets and Software 5) Account Management 6) Access Control Management 7) Continuous Vulnerability Management 8) Audit Log Management 9) Email and Web Browser Protections 10) Malware Defences 11) Data Recovery 12) Network Infrastructure Management 13) Network Monitoring and Defence ETSI ETSI TS 103 992 V1.1.1 (2024-05) 21 14) Security Awareness and Skills Training 15) Service Provider Management 16) Application Software Security 17) Incident Response Management 18) Penetration Testing NOTE: The selection of specific controls from each group is contextual and are further described below. The use of MISP Project [i.18] tools, community collaboration, instances, and standards are important to effective implementation of these capabilities. The MISP Project [i.18] is substantially EU based, supported by ENISA, and deployed in EU Member States, making it relevant for NIS2 Directive [i.1] purposes. Similarly, OSCAL use for meeting NIS2 Directive [i.1] requirements: 1) has been significantly advanced through EU funded research and development; 2) provides for open availability and interoperability among relevant regulatory requirements and standards; and 3) enables use automated tools for implementing tailored requirements for both initial certification and continuous monitoring that are especially important for SME and Micro Enterprises. The application of OSCAL [i.21] for implementing multiple NIS2 Directive [i.1] provisions can be used where affected providers need to meet an array of different, constantly changing requirements in diverse contexts. Implementation of a Zero Trust Security Model (see [i.34], [i.35] and relevant entries in Annex B) necessitates multiple NIS2 Directive [i.1] related actions. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 5.2 National cybersecurity strategy (Art. 7) | Article 7 of the NIS2 Directive [i.1] requires each EU Member State to adopt a national cybersecurity strategy defining the strategic objectives and appropriate policy and regulatory measures, with a view to achieving and maintaining a high level of cybersecurity. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 5.3 Coordinated vulnerability disclosure (Art. 12) | Article 6 of the NIS2 Directive [i.1] requires each EU Member State to designate one of its CSIRTs as a coordinator for the purpose of Coordinated Vulnerability Disclosure (CVD). It additionally requires ENISA to develop and maintain a European vulnerability database, in consultation with the Cooperation Group. Implementation of the relevant Critical Security Control Safeguards together with the relevant Control Facilitation Mechanisms and Privacy can be used for effective implementation of coordinated vulnerability disclosure requirements. In practice, Critical Security Control Safeguards 16.2, 16.3, and 16.6 cover the discovery and management of vulnerabilities, but this is from an application security perspective. CVD practices and specifications exist in publications of multiple organizations that provide standards and guidance documents with greater details into to CVD processes. Exchanged information should be labelled using the Traffic Light Protocol, and for triage and expression of criticality, the CVSS [i.27] should be employed. General, single organization CVD requirements pursuant to NIS2 Directive [i.1], Art. 12, para. 1 can be met using the following: - FIRST All Service Frameworks CSIRTs and PSIRTs [i.24] - ETSI TR 103 838 (guide to CVD) [i.28] - ISO/IEC 29147 [i.29] - ISO/IEC 30111 [i.30] ETSI ETSI TS 103 992 V1.1.1 (2024-05) 22 General, multi-party CVD requirements can be met using the following: - FIRST Guide to Multi Party CVD [i.26] - ISO/IEC TR 5895 [i.31] 5.4 Crisis management, CSIRT, cooperation, peer review (Arts. 9 to 16, 19) Art. 9 of the NIS2 Directive [i.1] requires each Member State designate one or more competent authorities responsible for the management of large-scale cybersecurity incidents and crises to identify capabilities, assets and procedures that can be deployed in case of a crisis and adopt a national cybersecurity incident and crisis response plan. Arts. 10 - 16 describe the implementation of CSIRTs, their requirements and tasks, their cooperation at the national level, creation of an EU Cooperation Group (EU CSIRT), an EU CSIRTs network, and the European cyber crises liaison organization network (EU - CyCLONe). Art. 19 requires the Cooperation Group establish a peer learning system. Implementation of the relevant Critical Security Control Safeguards together with the relevant Control Facilitation Mechanisms and Privacy can be used for effective implementation of crisis management, CSIRT, cooperation, and peer review requirements. Use of the MISP Project [i.18] tools, community collaboration, instances, and standards are important for effective implementation of some of these requirements. The MISP Project [i.18] is already in wide use today in ENISA and various other national cybersecurity agencies around the world for automated cyber-threat intelligence information sharing and visualization. MISP can be used for NIS2 Directive requirements [i.1] pertaining to crisis management, CSIRT, cooperation, and peer review capabilities. Therefore, the use of MISP [i.18] is recommended. Article 9, paragraph 4 of the NIS2 Directive [i.1] requires each Member State to adopt a national cybersecurity incident and crisis response plan that encompasses 'cyber crisis management procedures, including their integration into the general national crisis management framework and information exchange channels', 'the relevant public and private stakeholders and infrastructure involved,' and 'national procedures and arrangements between relevant national authorities and bodies to ensure the Member State's effective participation in and support of the coordinated management of large-scale cybersecurity incidents and crises at Union level'. These requirements infer that fast and secure alerting is needed to a plethora of entities and endpoints, of whom can be located in different regions and countries, in order to collect information and connect people for situational awareness. Furthermore, the dissemination of critical security information might need to be followed by immediate action and/or coordinating activities. To facilitate this, personnel might need to acknowledge receipt of the critical security information. The OASIS Common Alerting Protocol (CAP) [i.33] is specifically designed for enabling large-scale crisis management. OASIS CAP [i.33] provides a general format for exchanging emergency alerts and public warnings over any kind of network, enabling consistent warning messages to be disseminated simultaneously over a multitude of warning systems. In particular, OASIS CAP [i.33] offers the following features and functionalities that may be useful for some NIS2 Directive requirements [i.1] and possibly other EU resilience and risk management requirements (e.g. [i.6] and [i.36]): - flexible geographic targeting using latitude/longitude shapes and other geospatial representations in three dimensions; - multilingual and multi-audience messaging; - phased and delayed effective times and expirations; - enhanced message update, acknowledgement, and cancellation features; - template support for framing complete and effective warning messages; - compatible with digital signature capability; and - facility for digital images and audio. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 23 5.5 Cybersecurity risk management and critical supply chain (Arts. 20 to 22) Art. 20, para. 1 of the NIS2 Directive [i.1] requires Member State management bodies of essential and important entities approve the cybersecurity risk management measures taken by those entities, oversee its implementation and holding them accountable for the non-compliance. Art. 20, para. 2 requires Member States to ensure that the management bodies of essential and important entities are required to be trained on risk-management measures and offer it to their employees. Art. 21 requires Member States to ensure that essential and important entities take appropriate and proportionate technical and organizational measures to manage the risks posed to the security of network and information systems which those entities use in the provision of their services. Art. 22 states that the Cooperation Group, in cooperation with the Commission and ENISA, can carry out coordinated security risk assessments of specific critical ICT services, systems, or product supply chains, taking into account technical and where relevant, non-technical risk factors. In addition, Art. 22 requires the Commission, after consulting with the Cooperation Group and ENISA, identify the specific critical ICT services, systems or products that may be subject to the coordinated risk assessment. Implementation of the relevant Critical Security Control Safeguards together with the relevant Control Facilitation Mechanisms and Privacy are all essential for effective implementation of risk management and critical supply chain requirements. Use of the MISP Project [i.18] tools, community collaboration, instances, and standards are important to effective implementation of these capabilities. The application of OSCAL for implementing multiple NIS2 provisions can be appropriate where affected providers need to meet an array of different, constantly changing requirements in diverse contexts. Art. 24, paras. 1 and 2 deal with the use of certified products by essential and important entities. Art. 14, para. 3 deals with the definition of a candidate scheme. Providers and vendors are likely to be faced with innumerable different regulatory and industry certification requirements in different jurisdictions. OSCAL was created to facilitate compliance with such complex environments. NOTE: The adoption of the proposed European Commission draft Cyber Resilience Act (CRA) [i.32]- which significantly expands on the NIS2 Directive risk management and critical supply chain requirements – could potentially affect the implementation of NIS2 Directive Articles 20-22. ETSI GR ETI 006 [i.22] provides useful guidance for implementing Art. 21, para. 2(h) requirements. ETSI EN 319 401 [i.16] should be implemented as it is essential to the implementation of the NIS2 Directive's [i.1] risk management requirements for trust services and providers. Implementation of a Zero Trust Security Model (see [i.34], [i.35] and Annex B) should be considered in the context of risk management and critical supply chains. Potential compromise of devices and applications in the supply chain is assumed even if certified. Privileges and access to data are controlled, minimized, and monitored; segmentation is enforced by policy; and analytics are used to monitor for anomalous activity. Complementary to the Zero Trust Security Model, implementing a Moving Target Defence methodology (which includes deception, dynamic positioning, and non-persistence) can help provide a proactive defence, increasing a system's resilience and protection against zero-day attacks (see [i.49]). The goal is to prevent unauthorized access to resources and services coupled with making the access control enforcement as granular as possible. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 5.6 European cybersecurity certification scheme (Art. 24) | Art. 24 of the NIS2 Directive [i.1] states that Member States may require essential and important entities to use particular ICT products, services and processes certified under specific European cybersecurity certification schemes. It further notes that the Commission may adopt implementing acts specifying which categories of essential or important entities will be required to use certain certified ICT products, services, and processes or obtain a certificate under a European cybersecurity certification scheme. The Commission can request ENISA to prepare a candidate scheme or to review an existing European cybersecurity certification scheme in cases where no appropriate European cybersecurity certification scheme is available [i.39]. The application of OSCAL [i.21] for implementing multiple NIS2 Directive [i.1] provisions may be necessary where affected essential and important entities need to meet an array of different, constantly changing requirements in diverse contexts. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 24 NOTE: The MEDINA Project [i.20] is an EU-funded research project developing a framework to achieve continuous audit-based certification in compliance with the EU Cybersecurity Certification Scheme for Cloud Services (EUCS). It is investigating the usage of OSCAL [i.21] for compliance automation in support of relevant regulations including the NIS2 Directive [i.1]. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 5.7 Standardization (Art. 25) | Art. 25 of the NIS2 Directive [i.1] requires, with respect to risk management measures, that Member States encourage the use of European or internationally accepted standards and specifications relevant to the security of network and information systems. Furthermore, that ENISA, in collaboration with Member States, draw up advice and guidelines regarding the technical areas to be considered as well as regarding already existing standards, including Member States' national standards, which would allow for those areas to be covered [i.39] and [i.40]. Implementation of the relevant Critical Security Control Safeguards ([i.9]) together with the relevant Control Facilitation Mechanisms and Privacy can be used for implementation of standardization requirements. Use of MISP Project tools, community collaboration, instances, and standards are important to effective implementation of these capabilities. Use of a Zero Trust Security Model (see [i.34], [i.35] and Annex B), government security, and industry standards and guidelines should be considered to enhance cybersecurity postures. |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 5.8 Reporting and information-sharing (Arts. 23, 29, 30) | Art. 23 of the NIS2 Directive [i.1] requires Member States establish an extensive array of reporting and sharing obligations concerning any incident having a significant impact on the provision of services and mitigation measures. Art. 29 requires Member States ensure that essential and important entities may exchange on a voluntary basis relevant cybersecurity information among themselves including information relating to cyber threats, near misses, vulnerabilities, indicators of compromise, tactics, techniques and procedures, cybersecurity alerts and configuration tools. Art. 30 requires Member States that essential and important entities may notify, on a voluntary basis, to the competent authorities or the CSIRTs any relevant incidents, cyber threats or near misses. Special requirements apply to trust service providers [i.33] and the treatment of personal data under the GDPR [i.42]. Implementation of the relevant Critical Security Control Safeguards ([i.9]) together with the relevant Control Facilitation Mechanisms and Privacy can be used for implementation of reporting and information-sharing requirements. Use of MISP Project [i.18] tools, community collaboration, instances, and standards are important to effective implementation of these capabilities. The information sharing may include routing threat and mitigation information using MANRS (see Annex B). Art. 23 implementations, however, requires an array of potentially difficult-to-implement information sharing and reporting capabilities and best practices described in clause 4.1 of the present document. For example, Art. 23, para. 1 of the NIS2 Directive [i.1] requires the reporting of incidents that have a significant impact on the provision of an entity's services ("significant incident"). Art. 23, para. 3 provides some guidance on classifying an incident as significant, however, additional guidance is needed to characterize certain stated terms e.g. severe operational disruption, (severe) financial losses, considerable material losses. An additional example involves the Art. 23, para. 4 and Art. 30 notification processes where stakeholders need to know what channels have been established by the national CSIRTs or national authorities. Important and essential entities that are newcomers to the NIS2 Directive [i.1] need to be informed of the technical and operational interfaces with the national CSIRTs. Similarly, Art. 29 of the NIS2 Directive [i.1] also concerns information sharing. MISP [i.18] and other CTI protocols are relevant to implement this requirement at a technical level. Yet, the objective of the article is that EU Member States ensure that there are communities in place where information can be exchanged in digital or physical form, as per Art. 29, para. 2. These capabilities can consist of the national platforms and communities put in place by the CERT/CSIRTs. The implementation of Art. 29 requires the effective identification and participation of essential and important entities in the communities. Implementation of ETSI EN 319 401 [i.16] is essential to the implementation of reporting requirements for trust services and providers according to the NIS2 Directive [i.1]. Enhancing continuous reporting and information-sharing capabilities to support implementations of a Zero Trust Security Model (see [i.34], [i.35] and Annex B) should be considered to enhance cybersecurity postures. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 25 |
e9ad103446e7301c9322c48d5ddebb4b | 103 992 | 6 Recommendations | The Critical Security Control Safeguards profiled to be part of Implementation Group 1 as defined in ETSI TR 103 305- 1 [i.9] shall be implemented by entities. All other Critical Security Control safeguards from ETSI TR 103 305-1 [i.9] and privacy enhancements specified in ETSI TR 103 305-5 [i.12], as detailed in the subclauses of 4 and 5 of the present document, should also be implemented. To help train a workforce to be cybersecurity aware, including such activities as verifying installations of software patches, entities should follow guidance in one or more of ETSI TR 103 305-1 [i.9], IEC 62443 series [i.50], and ISO 27000 family [i.51]. To help reduce the risk of spam and phishing activities, entities should implement Domain-based Message Authentication, Reporting and Conformance (DMARC). To further secure email and other communications, entities should also utilize encryption tools and protocols. ETSI GR ETI 006 [i.22] should be used for guidance. To help address cyber threats within their networks and depending on national policy, entities shall implement the Zero Trust-related aspects of Art. 21, para. 2 of the NIS2 Directive [i.1], which include continuous authentication solutions; secured voice, video, and text communications; and secured emergency communication systems. Entities should implement other components of the Zero Trust Security Model, following guidelines provided by the entity's national authority (e.g. see [i.53] and [i.54]) or [i.34]. Entities should include a Moving Target Defence methodology (including deception, dynamic positioning, and non-persistence) in order to provide a proactive, resilient defence, protecting against zero-day attacks. AI-based or non-signature-based antimalware software to detect the most current malware should also be utilized by entities to support this. To help address cyber threats in cloud data centre implementations, entities should utilize hardened images as identified in ETSI TR 103 305-4 [i.11]. For effective implementation and interoperability of crisis management, CSIRT, and cooperation requirements between Member States, entities shall use the MISP Project [i.18] tools, community collaboration, instances, and standards. To enable large crisis management, consideration should be given to utilizing OASIS CAP [i.33]. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 26 Annex A (informative): Mapping between the NIS2 Directive and the Critical Security Control Safeguards This Annex provides a mapping between the articles of the EU NIS2 Directive [i.1] (excluding those that are informational i.e. Arts. 1 to 6) and the Critical Security Control Safeguards specified in ETSI TR 103 305-1 [i.9]. ETSI ETSI TS 103 992 V1.1.1 (2024-05) 27 Table A.1: Mapping between the EU NIS2 Directive [i.1] and the Critical Security Control Safeguards [i.9] Art. # Art. Title TS clause NIS2 Directive Group Critical Cybersecurity Control Safeguards 4.1 4.2 4.3 4.4 4.5 4.6 4.7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 7 National cybersecurity strategy 5.1 X X X X X X X X X X X X X X X X X X X X X X 8 Competent authorities and single points of contact 9 National cybersecurity crisis management frameworks 5.3 X X X 10 Computer security incident response teams (CSIRTs) 5.3 X X X X X X X X X X X X X X X 11 Requirements, technical capabilities and tasks of CSIRTs 5.3 X X X X X X X X X X X X X X X 12 Coordinated vulnerability disclosure and a European vulnerability database 5.2 X X X X X X X X X X X X X X X X X X X 13 Cooperation at national level 5.3 X X X X X X X X X X X X X X X X 14 Cooperation Group 5.3 X X X X X X X X X X X X X X X X X X X X X X X X 15 CSIRTs network 5.3 X X X X X X X X X X X X X X X X X X X 16 European cyber crises liaison organization network (EU-CyCLONe) 5.3 X X X X X X X X X X X X X 17 International cooperation 18 Report on the state of cybersecurity in the Union 19 Peer-reviews 5.3 X X X X X X X X X X X X X X X 20 Governance 5.3, 5.4 X X X X X X X X X X X X X X X X X 21 Cybersecurity risk management measures 5.4 X X X X X X X X X X X X X X X X X X X X X X 22 Union level coordinated security risk assessments of critical supply chains 5.4 X X X X X X X X X X X X X X X X X X X X X X 23 Reporting obligations 5.7 X X X X X X X X X X X X X X X 24 Use of European cybersecurity certification schemes 5.5 25 Standardization 5.6 X X X X X X X X X X X X X X X X X X X X 26 Jurisdiction and territoriality 27 Registry of entities 28 Database of domain name registration data 29 Cybersecurity information-sharing arrangements 5.7 X X X X X X X X X X X X X X X X X X 30 Voluntary notification of relevant information 5.7 X X X X X X X X X X X X X X X ETSI ETSI TS 103 992 V1.1.1 (2024-05) 28 Art. # Art. Title TS clause NIS2 Directive Group Critical Cybersecurity Control Safeguards 4.1 4.2 4.3 4.4 4.5 4.6 4.7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 31 General aspects concerning supervision and enforcement 32 Supervision and enforcement for essential entities (Annex I) 33 Supervision and enforcement for important entities (Annex II) 34 General conditions for imposing administrative fines on essential and important entities 35 Infringements entailing a personal data breach ETSI ETSI TS 103 992 V1.1.1 (2024-05) 29 Annex B (informative): Bibliography • ETSI Doc. CYBER(16)006023r1: "Deconstructing the EU NIS Directive: model, architecture, interfaces, expressions", 8th February 2016. • ETSI Doc. CYBER(21)27b003: "Deconstruction and Implementation of the Revised Network and Information Security (NIS2) Directive Proposal", 7th December 2021. • ETSI Security Week 2017, Session 1: "Standards and Legislation". NOTE: The above is available only to those with an ETSI Online Account. • NIST, Open Security Controls Assessment Language (OSCAL). • NIST, OSCAL Concepts, Layers and Models. • OSCAL community resources. • MANRS Primer: "CSIRTs". • ENISA Telecom Security Forum: "The MANRS Project". • NCSC (UK): "Zero trust architecture design principles". • CISA: "Zero Trust Maturity Model" April 2023, Version 2.0. • 3GPP TR 33.894: "Technical Specification Group Services and System Aspects; Study on applicability of the Zero Trust Security principles in mobile networks (Release 18)". ETSI ETSI TS 103 992 V1.1.1 (2024-05) 30 Annex C (informative): Change history Date Version Information about changes May 2024 1.1.1 First publication ETSI ETSI TS 103 992 V1.1.1 (2024-05) 31 History Document history V1.1.1 May 2024 Publication |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 1 Scope | The present document specifies security protocols as to be used for O-RAN compliant implementation. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 2 References | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 2.1 Normative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found in the ETSI docbox. NOTE 1: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long-term validity. The following referenced documents are necessary for the application of the present document. NOTE 2: All specifications issued by IETF referenced in the present document are valid in the latest version which the IETF declares as valid. Any updates to valid RFCs have to be considered and implemented, if applicable. [1] Void. [2] IETF RFC 4252: "The Secure Shell (SSH) Authentication Protocol". [3] Void. [4] Void. [5] IANA: "Secure Shell (SSH) Protocol Parameters". [6] O-RAN ALLIANCE TS: "O-RAN Management Plane Specification". [7] Void. [8] IANA: "Transport Layer Security (TLS) Parameters", Retrieved 2021-01-21. [9] Void. [10] O-RAN-WG1.O1-Interface-v04.00: "O-RAN Operations and Maintenance Interface Specification v04.00". [11] IETF RFC 6668: "SHA-2 Data Integrity Verification for the Secure Shell (SSH) Transport Layer Protocol", July 2012. [12] IETF RFC 8268: "More Modular Exponentiation (MODP) Diffie-Hellman (DH) Key Exchange (KEX) Groups for Secure Shell (SSH)", December 2017. [13] IETF RFC 8308: "Extension Negotiation in the Secure Shell (SSH) Protocol", March 2018. [14] IETF RFC 8332: "Use of RSA Keys with SHA-256 and SHA-512 in the Secure Shell (SSH) Protocol", March 2018. [15] IETF RFC 8709: "Ed25519 and Ed448 Public Key Algorithms for the Secure Shell (SSH) Protocol", February 2020. [16] Void. [17] Void. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 7 [18] IETF RFC 8446: "The Transport Layer Security (TLS) Protocol Version 1.3". [19] Void. [20] ETSI TS 133 210: "Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; 5G; Network Domain Security (NDS); IP network layer security (3GPP TS 33.210)". [21] Void. [22] IETF RFC 6066: "Transport Layer Security (TLS) Extensions: Extension Definitions". [23] IETF RFC 9147: "The Datagram Transport Layer Security (DTLS) Protocol Version 1.3". [24] IETF RFC 6083: "Datagram Transport Layer Security (DTLS) for Stream Control Transmission Protocol (SCTP)". [25] Void. [26] ETSI TS 133 310: "Universal Mobile Telecommunications System (UMTS); LTE; 5G; Network Domain Security (NDS); Authentication Framework (AF) (3GPP TS 33.310)". [27] IETF RFC 4301: "Security Architecture for the Internet Protocol". [28] 3GPP TS 33.401 (V16.0.0): "3GPP System Architecture Evolution (SAE); Security architecture (Release 15)". [29] ETSI TS 133 501 (V16.6.0): "5G; Security architecture and procedures for 5G System (3GPP TS 33.501 version 16.6.0 Release 16)". [30] Void. [31] IETF RFC 4303: "IP Encapsulating Security Payload (ESP)". [32] Void. [33] IETF RFC 4306: "Internet Key Exchange (IKEv2) Protocol". [34] IETF RFC 7296: "Internet Key Exchange Protocol Version 2 (IKEv2)". [35] IETF RFC 6749: "The OAuth 2.0 Authorization framework". [36] IETF RFC 7519: "JSON Web Token (JWT)". [37] IETF RFC 7515: "JSON Web Signature (JWS)". [38] IETF RFC 4210: "Internet X.509 Public Key Infrastructure Certificate Management Protocol (CMP)". [39] IETF RFC 4211: "Internet X.509 Public Key Infrastructure Certificate Request Message Format (CRMF)". [40] NIST: "Cryptographic Standards and Guidelines". [41] Void. [42] FIPS 186 series: "Digital Signature Standard (DSS)", National Institute of Standards and Technology. [43] Void. [44] FIPS 197: "Advanced Encryption Standard (AES)", National Institute of Standards and Technology. [45] Void. [46] Void. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 8 [47] FIPS 202: "NIST Permutation-Based Hash and Extendable-Output Function". [48] NIST SP 800-131A: "Transitioning the Use of Cryptographic Algorithms and Key Lengths". [49] Void. [50] IETF RFC 8017: "PKCS #1: RSA Cryptography Specifications Version 2.2". [51] Void. [52] IANA: "Transport Layer Security (TLS) Parameters", Accessed May 7, 2022. [53] Void. [54] IETF RFC 6125: "Representation and Verification of Domain-Based Application Service Identity within Internet Public Key Infrastructure Using X.509 (PKIX) Certificates in the Context of Transport Layer Security (TLS)". [55] IETF RFC 7633: "X.509v3 Transport Layer Security (TLS) Feature Extension". [56] CA-Browser-Forum-BR-1.8.0: "Baseline Requirements for the Issuance and Management of Publicly‐Trusted Certificates", August 2021. [57] Void. [58] Void. [59] Void. [60] IETF RFC 7093: "Additional Methods for Generating Key Identifiers Values". [61] Void. [62] IETF RFC 6960: "X.509 Internet Public Key Infrastructure Online Certificate Status Protocol - OCSP". [63] IETF RFC 3647: "Internet X.509 Public Key Infrastructure Certificate Policy and Certification Practices Framework". [64] IETF RFC 5280: "Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile". [65] SFTPv3: "SFTP Industry Best Practice". [66] IETF RFC 4217: "Securing FTP with TLS". [67] ETSI TS 128 532: "5G; Management and orchestration; Generic management services (3GPP TS 28.532)". [68] ETSI TS 128 537: "5G; Management and orchestration; Management capabilities (3GPP TS 28.537)". [69] IETF RFC 9110: "HTTP Semantics". [70] IETF RFC 4253: "The Secure Shell (SSH) Transport Layer Protocol". [71] IETF RFC 9325: "Recommendations for Secure Use of Transport Layer Security (TLS) and Datagram Transport Layer Security (DTLS)". [72] IETF RFC 3279: "Algorithms and Identifiers for the Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile". [73] FIPS 180-4: "Secure Hash Standard (SHS)", National Institute of Standards and Technology. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 9 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 2.2 Informative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long-term validity. The following referenced documents may be useful in implementing an ETSI deliverable or add to the reader's understanding, but are not required for conformance to the present document. [i.1] ETSI TR 121 905: "Digital cellular telecommunications system (Phase 2+) (GSM); Universal Mobile Telecommunications System (UMTS); LTE; 5G; Vocabulary for 3GPP Specifications (3GPP TR 21.905)". [i.2] Mozilla Wiki: "Security/Server Side TLS", 2 January 2020. . [i.3] OWASP Cheat Sheet Series: "TLS Cipher String Cheat Sheet", 2020. [i.4] NIST Special Publication 800-52.2: "Guidelines for the Selection, Configuration, and Use of Transport Layer Security (TLS) Implementations", Kerry McKay (NIST), David Cooper (NIST), August 2019. . [i.5] 3GPP TR 33.821 (V9.0.0): "Rationale and track of security decisions in Long Term Evolved (LTE) RAN / 3GPP System Architecture Evolution (SAE) (Release 9)". [i.6] BSI TR-02102-1: "BSI Technical Guideline - Designation: Cryptographic Mechanisms: Recommendations and Key Lengths", 2022-01. [i.7] NIST SP 800-38 series: "Recommendation for Block Cipher Modes of Operation". [i.8] NIST SP 800-186: "Recommendations for Discrete Logarithm-based Cryptography: Elliptic Curve Domain Parameters". [i.9] NIST SP 800-57: "Recommendation for Key management". [i.10] Recommendation ITU-T X.509: "Information technology - Open Systems Interconnection - The Directory: Public-key and attribute certificate frameworks". [i.11] Void. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 3 Definition of terms, symbols and abbreviations | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 3.1 Terms | Void. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 3.2 Symbols | Void. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 10 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 3.3 Abbreviations | For the purposes of the present document, the abbreviations given in ETSI TR 121 905 [i.1] and the following apply. NOTE: An abbreviation defined in the present document takes precedence over the definition of the same abbreviation, if any, in ETSI TR 121 905 [i.1]. CA Certification Authority CMP Certificate Management Protocol DH Diffie–Hellman DPD Dead Peer Detection DTLS Datagram Transport Layer Security ESP Encapsulating Security Payload FTPES Explicit SSL File Transfer Protocol HTTPS Hypertext Transfer Protocol Secure IKE Internet Key Exchange IPsec Internet Protocol security NAT Network Address Translation NE Network Element NETCONF Network Configuration protocol PKI Public Key Infrastructure RA Registration Authority SA Security Association SFTP SSH File Transfer Protocol SPD Security Policy Database SSH Secure Shell TLS Transport Layer Security 4 Security protocols specifications for O-RAN compliant implementation |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1 SSH | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1.1 General requirements | O-RAN and 3GPP interfaces that implement authentication, confidentiality and integrity using SSH shall: • Support SSHv2 [1] and [70]. • Disable by default cryptographically insecure ciphers as specified in clauses 4.1.2.1, 4.1.2.3 and 4.1.2.4 of the present document. • Enable an O-RAN deployer to configure SSH to offer less secure ciphers using standard SSH configurations to enable backward compatibility with older SSH implementations. • Enable remote shell access only if required by the interface. If remote shell access is enabled, disable remote login as root or equivalent highest privileged user. Such access shall be limited to the local system console only. Root user shall not be allowed to login to the system remotely. Entities providing O-RAN components that support SSH for authentication, confidentiality or integrity shall: • Stay current with SSH [5]. • Provide an upgrade path for changes to the SSH protocol and ciphers [5]. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 11 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1.2 Required ciphers | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1.2.0 Introduction | O-RAN requires the following ciphers when using SSH. For more information see [2], [5], [11], [12], [13], [14] and [15]. See the Security chapter of the O-RAN Working Group 4 Management Plane Specification for the M-plane mandated SSH ciphers [6]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1.2.1 Key agreement | Note that the present document uses the IANA cipher naming convention [5]. • Required a. ecdsa-sha2-nistp256 b. ecdsa-sha2-nistp384 c. ecdsa-sha2-nistp521 d. ssh-ed25519 • Optional a. ssh-ed448 • Shall not be implemented a. ssh-rsa b. ssh-dss |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1.2.2 Symmetric algorithms for encrypting transferred data | • Required a. aes256-gcm b. aes128-gcm c. aes256-ctr d. aes192-ctr e. aes128-ctr |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1.2.3 Key exchange algorithms (KexAlgorithms) | • Required a. ecdh-sha2-nistp521 b. ecdh-sha2-nistp384 c. ecdh-sha2-nistp256 d. diffie-hellman-group-exchange-sha256 e. curve25519-sha256 • Shall not be implemented a. Diffie-hellman-group1-sha1 ETSI ETSI TS 104 107 V9.0.0 (2025-05) 12 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.1.2.4 Message Authentication Codes (MACs) | • Required a. hmac-sha2-512-etm b. hmac-sha2-512 c. hmac-sha2-256-etm d. hmac-sha2-256 e. umac-128 • Shall not be implemented a. hmac-sha1 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.2 TLS | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.2.1 General requirements | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.2.1.0 Introduction | O-RAN interfaces that implement authentication, confidentiality and integrity using Transport Layer Security (TLS) shall: • Support TLS 1.2 [18]. • Configure the TLS 1.2 Intermediate server ciphers as specified in [i.2] and [8]. • Support TLS 1.3 [18] and [i.4]. • Enable an O-RAN deployer to configure TLS 1.2 to offer less secure ciphers using standard TLS configurations to enable backward compatibility with weaker TLS ciphers. • Disable by default cryptographically insecure ciphers identified in [8] and [i.3]. Entities providing O-RAN components that support TLS for authentication, confidentiality or integrity shall: • Stay current with the latest release of the TLS software used to implement the protocol, such as OpenSSL. • Provide an upgrade path for new software releases. Any version of SSL and any version of TLS below 1.2, shall not be supported. Entities providing O-RAN components that support TLS shall support mutual certificate-based authentication, i.e. client authentication shall be supported in TLS communications, also referred as mutual TLS authentication. • A TLS server supporting TLS 1.2 shall request a certificate from the client as specified in [18] clause 7.4.4 and validate the certificate that is being returned from the client using the structure defined in [18] clause 7.4.2. • A TLS server supporting TLS 1.3 shall request a certificate from the client as specified in [18] clause 4.3.2 and validate the certificate that is being returned from the client. See the Security chapter of the O-RAN Working Group 4 Management Plane Specification for the M-plane mandated TLS ciphers [6]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.2.1.1 Specific requirements | In addition, one-way TLS authentication may be supported with server certificate exchanged as specified in [18] clause 7.4.2 (TLS 1.2) or [18] clause 4.4.2 (TLS 1.3). ETSI ETSI TS 104 107 V9.0.0 (2025-05) 13 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.2.2 TLS Protocol profiles specifications | TLS 1.2 used on all interfaces except the Open Fronthaul interfaces shall support the TLS 1.2 profiles as defined by ETSI TS 133 210 [20] clause 6.2.3. See Chapter 5.4 Security in [6] for the M-plane mandated TLS ciphers [6]. TLS 1.3 used on all interfaces except the Open Fronthaul interfaces shall support the TLS 1.3 profiles as defined by ETSI TS 133 210 [20] clause 6.2.2. See Chapter 5.4 Security in [6] for the M-plane mandated TLS ciphers [6]. Use of a cipher suite in TLS shall be configurable. Broken ciphers should be removed from the list of negotiable ciphers. O-RAN specified protocols using TLS may support additional TLS ciphers recommended by IANA [52]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.2.3 Certificate Profile for TLS Entity | The present clause addresses the certificate profile requirements for the TLS entities that may behave either as client, server, or both. The structure of the TLS certificate profile described in this clause under Table 4-1 provides requirements based on the ETSI TS 133 310 [26], CA-browser forum [56] and NIST Special Publication 800-52.2 [i.4]. The TLS entity certificate profile shall be applied to all nodes and interfaces that use the TLS protocol for secured communication in the O-RAN network except the Open Fronthaul interfaces: • The TLS entity certificates shall adhere to the certificate profile outlined in Table 4-1, based on their respective roles. • The CA responsible for handling the certificate signing request shall ensure compliance in accordance with the certificate profile specified in clause 4.2.3. • The common rules for all certificates defined in ETSI TS 133 310 [26], clause 6.1.1, clause 6.1.3.a shall apply. • The CA-browser forum [56] has the following requirement for the certificate operational and key usage periods. Certificates issued on or after 1 September 2020 shall not have a Validity Period greater than 397 days. • Certificate Policies: Certificate policies shall be crafted using the guidelines defined in IETF RFC 3647 [63]. • subjectAltName shall (for TLS server certificates) contain at least one FQDN(Hostname) and shall not contain only IP Address as described in IETF RFC 6066 [22]. As per IETF RFC 6125 [54], an application service shall be identified by a name or names carried in the subject field (i.e. a CN-ID) and/or in one of the following identifier types within subjectAltName entries (DNS-ID, SRV-ID. URI-ID). • Optional and non-critical TLS Feature Extension: IETF RFC 7633 [55] defines a certificate extension that indicates that clients expect stapled OCSP responses with a value of "status_request (5)" for the certificate and aborts the handshake ("hard-fail") if such a response is not available. As per IETF RFC 9325 [71] TLS servers should support the following as a best practice: OCSP IETF RFC 6960 [62] and OCSP stapling using the status_request extension defined in IETF RFC 6066 [22]. The exact mechanism for embedding the status_request extension differs between TLS 1.2 and 1.3. As a matter of local policy, server operators may request that CAs issue must-staple IETF RFC 7633 [55] certificates for the server and/or for client authentication. • Extensions: Mandatory Critical Key Usage: - digitalSignature for both TLS client and Server certificates. This applies for RSA signature certificate, ECDSA signature certificate, or DSA signature certificate. - When using RSA IETF RFC 3279 [72] with TLS 1.2, the keyEncipherment shall be set. - When using DH [72] or ECDH [72] with TLS 1.2, the keyAgreement shall be set. The following table captures the certificate profile for the TLS entity that may behave as a client, server or both. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 14 Table 4-1: TLS Client and Server Certificate Profile ORAN TLS Client and Server Certificate Profile Version v3. Serial Number Shall be Unique Positive Integer in the context of the issuing CA Subject DN (C=<country>), O=<Organization Name>, CN=<Some distinguishing name>. Organization and CN can be in UTF8 format. Note that C is optional element. cn=<hostname>, (ou=<servers>), dc=<domain>, dc=<domain>. Note that ou is an optional element. Validity Period 1 year or less. Signature See clause 6.1.1 of ETSI TS 133 310 [26] for the list of supported signature algorithms. Subject Public Key Info See clause 6.1.1 of ETSI TS 133 310 [26] for the list of supported public key types. Extensions Mandatory Criticality Value keyUsage TRUE TRUE digitalSignature for TLS clients and servers. extendedKeyUsage FALSE FALSE id-kp-clientAuth TLS clients. id-kp-serverAuth for TLS servers. Entities that may be both client and server will have both OIDs set. certificatePolicies FALSE FALSE If added, then, it should be populated with a CP Object Identifier(OID). These OIDs correspond to specific policies established by the certificate issuer. authorityKeyIdentifier FALSE FALSE This is same as the subjectKeyIdentifier of the Issuer's certificate. CA utilizes the method as defined in clause 2 of IETF RFC 7093 [60]. subjectKeyIdentifier FALSE FALSE This is calculated by the issuing CA utilizing method as defined in clause 2 of IETF RFC 7093 [60]. cRLDistributionPoint TRUE FALSE distributionPoint According to IETF RFC 5280 [64] this indicates if the CRL is available for retrieval using access protocol and location with HTTP URI or LDAP. subjectAltName TRUE TRUE Multiple subjectAltNames are permitted and has been defined above authorityInfoAccess FALSE FALSE id-ad-caIssuers According to IETF RFC 5280 [64] id-ad-caIssuers describes the referenced description server and the access protocol and location, for example, using one or multiple HTTP and/or LDAP URIs. The referenced CA issuers description is intended to aid certificate users in the selection of a certification path that terminates at a point trusted by the certificate user id-ad-ocsp According to IETF RFC 5280 [64] id-ad-ocsp defines the location of the OCSP responder using HTTP URI. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 15 ORAN TLS Client and Server Certificate Profile Extensions Mandatory Criticality Value TLS feature extension FALSE FALSE id-pe-tlsfeature: status_request(5)" This can be used according to IETF RFC 7633 [55] to prevent downgrade attacks that are not otherwise prevented by the TLS protocol; IETF RFC 7633 [55] also defines a certificate extension that indicates that clients expects stapled OCSP responses with a value of "status_request(5)" for the certificate and aborts the handshake ("hard-fail") if such a response is not available. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.3 Support NETCONF over secure Transport | TLS requirements for use with NETCONF on the O1 interface [10] are specified in clause 4.2. TLS requirements for use with NETCONF on the Open Fronthaul M-plane are specified in clause 5.4 of the O-RAN WG4 Management Plane Specification [6]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.4 DTLS | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.4.1 General requirements | O-RAN and 3GPP interfaces that implement mutual authentication, integrity protection, replay protection and confidentiality protection using DatagramTransport Layer Security (DTLS) shall: • Support DTLS 1.2 [23]; • Support DTLS for Stream Control Transmission Protocol [24]; and should support DTLS 1.3 [23]. DTLS 1.0 shall not be supported. NOTE: IETF RFC 6083 [24] specifies a user message limit of approximately 16 kbytes, which does not align with the unlimited user message size that exists when SCTP is used without DTLS. There could be applications messages exceeding the limit, preventing the use of DTLS over SCTP, so enforcing unsecured SCTP. Entities providing O-RAN components that support DTLS for authentication, confidentiality or integrity shall: • Stay current with the latest release of the DTLS software used to implement the protocol, such as OpenSSL. • Provide an upgrade path for new software releases. Pre-shared keys in the authentication mechanisms should not be used. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.4.2 DTLS 1.2 profiling | DTLS 1.2 is based on TLS 1.2, so all requirements defined in clause 4.2 in the present document for TLS 1.2 shall apply to DTLS as well. 3GPP control plane interfaces in O-RAN system implementing DTLS 1.2 shall support all requirements defined in the profiling for TLS 1.2 specified in ETSI TS 133 210 [20], clause 6.2.3. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.4.3 Certificate profiling | Certificate requirements defined in clause 4.2 in the present document for TLS 1.2 shall apply to DTLS as well. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 16 3GPP control plane interfaces in O-RAN system implementing DTLS 1.2 shall support certificate profiling as given in clause 6.1.3a of ETSI TS 133 310 [26]. DTLS client certificates shall be directly signed by the DTLS client CA in the operator domain that the DTLS client belongs to. DTLS server certificates shall be directly signed by the DTLS server CA in the operator domain that the DTLS server belongs to. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.5 IPSec | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.5.1 Overview | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.5.1.0 Supported IPsec capabilities | O-RAN and 3GPP interfaces that implement authentication, confidentiality and integrity using IPsec shall support IPsec according to [27]. The supported IPsec capabilities in this clause follow [20] for interworking purposes and further apply requirements given in [26], [28], [29] and [i.5]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.5.1.1 Supported IPsec capabilities | Services that SHALL be supported (see also [29], section 9.8.2): • Confidentiality, can be enabled/disabled (via ENCR_NULL) • Integrity protection, always enabled • Data origin authentication, always enabled • Anti-replay protection, can be enabled/disabled • Extended sequence numbers, can be enabled/disabled Protocol that SHALL be supported: ESP ([31], as profiled by [20]): • IPsec mode: Tunnel mode • Copying the DSCP value from the inner IP-header to the outer IP-header for encrypted packets • For encrypted packets the DSCP value of the inner packet will be copied to the outer packet • NAT traversal ESP encryption transforms that SHALL be supported (including authenticated encryption transforms, see also [20], section 5.3.3) as defined in [i.5] (the ones marked with 'MUST'). ESP authentication transforms that SHALL be supported according to [20], subsection 5.3.4: • IKE endpoint Identification that SHALL be supported, according to [26], subsection 6.2.1b: IP address • Fully Qualified domain name (FQDN) (if DNS is supported Authentication that SHALL be supported: • X.509v3 digital certificates provided by a Certificate Authority solution Authentication that MAY be supported: • Pre-shared Keys NOTE: Pre-shared keys should not be used. Key exchange that SHALL be supported: • IKEv2 [33] profile as described in [26] ETSI ETSI TS 104 107 V9.0.0 (2025-05) 17 • Certificate-based authentication according to [26] • Certificates according to the profile described by [26] • DH group 19 • (Optional) DH group 20 Security Association multiplicity that SHALL be supported: • Multiple IKE SAs (multiple IPsec tunnels) • Multiple IPsec SAs • Multiple IPsec SAs per IPsec tunnel IKE SA(s) and IPSEC SA(s) SHALL be regularly re-established: • When the full sequence number space of an IPSEC SA(s) is used, the transmission for that SA SHALL stop. • Dead-Peer-Detection (DPD) SHALL be supported as defined in [34]. • Each node SHALL support a traffic narrowing function for the traffic selector ([34]): If the traffic selector notified from the peer (e.g. Security GW or neighbouring node) is wider than the Local/Remote Address range in the SPD information set on the node side, the peers set the traffic selector values to the narrower range. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.5.2 Parallel usage of IPsec and other secure transport protocols | Implementations SHALL support IPSec configuration with one or more dedicated connections in parallel with other secure transport protocols. EXAMPLE: It should be possible to run SSH or TLS connections within an IPsec tunnel or in parallel to IPsec tunnel(s). |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.5.3 Responder mode and Initiator/Responder mode support | Implementations SHALL support a configurable option per IKE Security Association to use "Responder mode" instead of "Responder/Initiator mode". Responder mode is applied to IKE SA establishment only. The introduction of "Responder mode" does not change the IKE SA rekeying behaviour: Each node shall be able to operate as initiator and responder in IKE_SA rekeying. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.6 CMPv2 | Certificate Management Protocol version 2 (CMPv2) is based on Internet X.509 Public Key Infrastructure (PKI) Certificate Management Protocol (CMP) specified in IETF RFC 4210 [38] and IETF RFC 4211 [39]. ETSI TS 133 310 [26] specifies the use of CMPv2 used by base stations to obtain an operator-signed certificate using a secured communication based on the vendor-signed certificate in the base station and a vendor root certificate pre- installed in the CA/RA server. Certificates may be installed at initial system initialization or obtained through certificate enrolment with the operator's PKI. Certificate enrolment may be supported with the CMPv2 protocol between the Network Element (NE) and the operator's CA as defined in ETSI TS 133 310 [26]. A PNF that supports CMPv2 shall use the CMPv2 profile defined in ETSI TS 133 310 [26], clause 9.5. The CA/RA server may include the trust anchor for the operator issued certificate and the appropriate certificate chains in the initialization response message. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 18 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7 OAuth 2.0 | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.1 Overview | The authorization framework described in clause 4.7 of the present document uses the OAuth 2.0 framework as specified in IETF RFC 6749 [35]. It allows the service producers to authorize the requests from service consumers, and the service consumers to obtain authorization credentials ("token endpoint"). Interfaces requiring the use of OAuth 2.0 for authorization purposes shall support the functionality according to clause 4.7 of the present document. Following options have been introduced to guarantee interoperability and align with the existing OAuth 2.0 authorization framework in ETSI TS 133 501 [29] (i.e. Service Based Architecture). |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.2 Basic Parameterization | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.2.1 General | Client access token authorization grants shall be supported with type Client Credentials Grant, as described in clause 4.4 of IETF RFC 6749 [35]. Mutual TLS authentication, as specified in the present document (clause 4.2.1), is the mechanism selected by O-RAN for this security procedure. In addition, still aligned with Client Credential Grant as described in section 4.4. of IETF RFC 6749 [35], Client Id and Client Secret may be supported for client authentication. In this case, one-way TLS may be used (certificate on server side). Access tokens shall be JSON Web Tokens (JWT) as described in IETF RFC 7519 [36] and shall be secured with digital signatures based on JSON Web Signature (JWS) as described in IETF RFC 7515 [37]. The 'scope' attribute as described in clause 3.3 of IETF RFC 6749 [35] shall be used to specify the allowed services per type of service producer. A more granular level of authorization may be defined by adding additional scope information with the token (e.g. to authorize specific operations, or access to particular resources or datasets), which requires to be verified by the service producer. OAuth 2.0 roles, as defined in clause 1.1 of IETF RFC 6749 [35], are as follows: • OAuth 2.0 Authorization server: new security function in O-RAN architecture; or provided by existing OAuth 2.0 infrastructure in the network. • OAuth 2.0 client: every API service consumer in O-RAN system (e.g. rApp, xApps). • OAuth 2.0 resource owner/server: every API service producer in O-RAN system (e.g. Near-RT RIC platform). |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.2.2 Registration process | OAuth 2.0 resource servers (API service producers) shall be registered with the OAuth 2.0 Authorization server. Service producers may include additional scope information related to specific allowed service operations and resources per client type. Before initiating the authorization protocol, OAuth 2.0 clients shall be registered with the OAuth 2.0 Authorization server. To achieve that, information about the service consumer instance and its type shall be made available in the OAuth 2.0 Authorization server. The registration process is subject to implementation procedures of the operator, with the following consideration on authentication procedure: OAuth 2.0 clients shall be capable to authenticate securely with the authorization server and client type shall be "confidential" as specified in clause 2.1 of IETF RFC 6749 [35]. Strong authentication mechanisms based on digital certificates shall be supported. In addition, client authentication mechanism based on client Id and Client Secret may be supported. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 19 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.2.3 Access Token request process | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.2.3.0 Server requirements | After TLS mutual authentication procedure between the OAuth 2.0 client and OAuth 2.0 Authorization server, or one- way TLS authentication with server certificate and client authentication based on Client Id and Client Secret has been executed, the Authorization server exposes a 'Token Endpoint' where the Access Token request service can be requested by OAuth 2.0 client. The following procedure depicts the procedure of the OAuth 2.0 client to obtain an access token from the OAuth 2.0 authorization server. OAuth 2.0 Client OAuth 2.0 Authorization Server 1.AccessToken_Get Request (Expected service name(s) and producer type, client type, client id, ...) 3. AccessToken_Get Response (expiration time, access_token) 2. Check whether the client (API service consumerr) is authorized. If client is authorized, generate an access token. Figure 4-1: Access Token request The OAuth 2.0 Client shall request an access token from the OAuth 2.0 Authorization server. For this operation the client shall send a HTTP POST request to the authorization server ('Token endpoint'), as described in IETF RFC 6749 [35], clause 3.2. The message, i.e. the body of the HTTP POST request, shall include the identifier of the client instance making the request, the 'scope' parameter indicating the expected services, and optionally additional scope information with more granular information about resources and operations on the resources, and optionally the type of client ('confidential') and the expected OAuth 2.0 resource owner/server. The message may include as well information referring to instance(s) of specific resource owner(s)/server(s) if required. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.2.3.1 Server requirements | • The OAuth 2.0 Authorization server shall verify that the input parameters in the access token (e.g. client type) match with the corresponding ones in the public certificate of the client, or those in the client profile previously registered. • If the client is authorized, the Authorization server shall generate an access token with appropriate scope as defined in clause 3.3 of IETF RFC 6749 [35] included. • The Authorization server shall digitally sign the generated access token based on a private key as described in IETF RFC 7515 [37]. If the client is not authorized, the Authorization server shall not issue an access token to that client. If the authorization is successful, the Authorization server shall send the access token to the client ('200 OK'), otherwise it shall reply based on OAuth 2.0 error response defined in IETF RFC 6749 [35]. • The success response should include in addition the expiration time for the token as indicated in IETF RFC 6749 [35]. • The response shall include information to identify the resource owner(s)/server(s) if they differ from the related information included in the token request. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 4.7.2.4 Service access request based on token verification | Once the client is in possession of a valid access token, it may proceed with the request of service access towards the service producer (OAuth 2.0 resource owner/server). ETSI ETSI TS 104 107 V9.0.0 (2025-05) 20 OAuth 2.0 Client OAuth 2.0 Resource Owner/Server 1.Service request (access token) 3. Service response 2. Verify integrity and claims in the access token. If successful, execute the requested service Figure 4-2: Service request 1. The OAuth 2.0 client requests service from the OAuth 2.0 resource owner/server. The OAuth 2.0 client shall include the access token in the request. The OAuth 2.0 client and OAuth 2.0 resource owner/server shall authenticate each other via mutual TLS, as defined in the present document (clause 4.2.1) In addition, one-way TLS authentication with certificate only on server side may be supported, as defined in the clause 4.2.1. 2. The OAuth 2.0 resource owner/server shall verify the token: i. It ensures the integrity of the token by verifying the signature using the OAuth 2.0 Authorization server's public key. ii. If the integrity check is successful, the OAuth 2.0 resource owner/server shall verify the claims. 3. If the verification is successful, the OAuth 2.0 resource owner/server shall execute the requested service and respond back to the OAuth 2.0 client. Otherwise, it shall send a proper error response. If the HTTP protocol is used, that error response shall be based on the OAuth 2.0 error response defined in IETF RFC 6749 [35]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 5 Cryptographic operations | Table 5-1 outlines the main cryptographic operations involved in the protection of (1) App/VNF/CNF packages for ensuring their integrity, authenticity and confidentiality (for sensitive artifacts) during delivery, onboarding, and instantiation phases, (2) the communication channel over O-RAN interfaces in terms of authenticity, confidentiality and integrity, and (3) the stored assets within O-RAN system. It contains the allowed list of algorithms, key sizes and standards according to BSI [i.6] and NIST [40], [48], [i.9] cryptographic guidelines. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 21 Table 5-1 Cryptographic operations Algorithms Key sizes Applicable standards Usage Note Signature generation and verification RSASSA-PSS RSA_PKCS1_V1 _5 ≥ 2048-bits FIPS 186 [42] For ensuring the integrity and authenticity of Apps/VNF/CNF packages during delivery, onboarding, and instantiation phases Existing code should use PKCS #1 v1.5 padding mode for compatibility only. Use of null padding is not recommended. Signature generation and verification ECDSA NIST- approved curves (P-256, P-384, or P-521) ≥ 256-bits FIPS 186 [42], SP 800-186 [i.8] Parameter key size for DSA shall not allow the following combination: Bit lengths of L and N parameters L = 2048, N = 224 (BSI TR-02102-1 [i.6]). Symmetric Encryption/Decryption AES_128, AES_192 and AES_256 128, 192 and 256 bits FIPS 197 [44], SP 800-38 [i.7] (operation modes) For ensuring the confidentiality of sensitive artifacts Asymmetric Encryption/Decryption RSAES-OAEP ≥ 2048-bits IETF RFC 8017 [50] Hashing SHA-2 family (SHA- 224, SHA-256, SHA-384, SHA-512, SHA-512/224 and SHA-512/256) SHA-3 family (SHA3-224, SHA3-256, SHA3-384, and SHA3-512) FIPS 180-4 [73] for SHA2 FIPS 202 [47] for SHA 3 For ensuring the integrity of Apps/VNF/CNF packages The signature operation shall involve X.509-based PKI certificates [i.10]. The following recommendations should be considered: • For the protection of cryptographic keys, Hardware Security Modules, or HSMs, should be used. • Along with HSMs, the principle of least privilege should be used with keys, to ensure only users who need the key have access to it. • In case a legacy system does not support ECDSA, RSA signing algorithms should be used instead. Otherwise, the use of the Elliptic curve signing algorithms is recommended. • If libraries or frameworks are in use that do not support PSS padding scheme, one of the RSA PKCS1 algorithms should be used instead. Otherwise, the use of one of the RSA PSS algorithms is recommended. • In general, the largest key size available for an algorithm family should be used: - For RSA, the largest supported key size is 4 096 bits. - For ECDSA, the largest supported key size is 512 bits. - For AES, the largest supported key size is 256 bits. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 22 |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6 Secure File Transfer protocols | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6.1 General | File Transfer management is required in several O-RAN use cases, e.g. for export of log files. An O-RAN architectural element that implements file management shall support at least one of these secure file transfer protocols: SFTP, FTPES, HTTPS. This is aligned with, for example, how 3GPP specifies File Management in [67] and [68]. SFTP is authenticated with username/password, SSH keys or X.509 certificates. FTPES is authenticated with X.509 certificates. HTTPS is mutually authenticated with X.509 certificates. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6.2 SFTP | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6.2.1 General Requirements | O-RAN architectural elements that implement secure file transfer using SFTP shall: • Support secure connection and authentication using SSHv2 Authentication Protocol [2] with all O-RAN specific requirements from clause 4.1 in the present document. • Support SFTPv3 as the SFTP Industry Best Practice [65]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6.3 FTPES | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6.3.1 General Requirements | O-RAN architectural elements that implement secure file transfer using FTPES shall: • Support secure connection and authentication using TLS with all O-RAN specific requirements from clause 4.2 in the present document. • Support FTPES as defined in [66]. |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6.4 HTTPS | |
5f9c441dec96cf3060075b9f996af316 | 104 107 | 6.4.1 General Requirements | O-RAN architectural elements that implement secure file transfer using HTTPS shall: • Support secure connection and authentication using TLS with all O-RAN specific requirements from clause 4.2 in the present document. • Support HTTPS as defined in [69]. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 23 Annex A (informative): Change history Date Revision Description 2024.03 09.00 Certificate Profile for O-RAN TLS entity Reference Update 2023.11 08.00 Secure file transfer protocols added SSH Update on no root remote login ssh-ed448 changed to optional Reference on SSH added 2023.07 07.00 Editorial alignments TLS entity certificate profile CMPv2 profile update Introduction of one-way TLS authentication for OAuth2.0 2023.03 06.00 Cryptographic operations update 2022.11 05.00 TLS Cipher update 2022.07.20 04.00 Addition of CMPv2 Update of O-Cloud Image protocols Addition of mTLS Update of OAuth 2.0 2021.11.08 03.00 Update the O-RAN security protocols and specifications to include mandatory support for TLS 1.3 2021.07.05 02.00 Addition of DTLS and IPsec requirements. Alignment of TLS 1.2 and TLS 1.3 profiles with 3GPP TS 33.210. Typographical changes. 2021.04.01 01.00 Initial version of the document with requirements for TLS and SSH. ETSI ETSI TS 104 107 V9.0.0 (2025-05) 24 History Document history V9.0.0 May 2025 Publication |
1a382157d5a45af8b664bdd32c73890e | 104 090 | 1 Scope | The present document describes the requirements for consumer receivers designed to be used with an Emergency Warning System (EWS) based on DAB, and the necessary test methods that lead to compliance with the requirements. It may be used as the technical basis for an EWS Certification Mark scheme. An EWS Certification Mark is designed to be used on product packaging and provides an easily recognized mark to correspond to public information campaigns on the necessary features and benefits of an EWS based on DAB. Manufacturers are, of course, free to include additional features or increased performance compared to the requirements specified in the present document. A DAB based EWS may also be used to deliver information to public signage and specialized receivers. Such devices may have additional features, such as addressability, which are not applicable to consumer receivers and are not within the scope of the present document. In addition to receivers with a minimum set of features and performance, an EWS also requires appropriate transmission infrastructure and authorization mechanisms. These latter are not within the scope of the present document. |
1a382157d5a45af8b664bdd32c73890e | 104 090 | 2 References | |
1a382157d5a45af8b664bdd32c73890e | 104 090 | 2.1 Normative references | References are either specific (identified by date of publication and/or edition number or version number) or non-specific. For specific references, only the cited version applies. For non-specific references, the latest version of the referenced document (including any amendments) applies. Referenced documents which are not found to be publicly available in the expected location might be found in the ETSI docbox. NOTE: While any hyperlinks included in this clause were valid at the time of publication, ETSI cannot guarantee their long term validity. The following referenced documents are necessary for the application of the present document. [1] ETSI TS 103 461: "Digital Audio Broadcasting (DAB); Domestic and in-vehicle digital radio receivers; Minimum requirements and Test specifications for technologies and products". [2] ETSI TS 104 089: "Digital Audio Broadcasting (DAB); Emergency Warning System (EWS); Definition and rules of behaviour". |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.