input stringlengths 1 785 | output dict | schema listlengths 0 0 |
|---|---|---|
The major identifier indicates which major routine group likely handles or creates the file. | {
"entities": {}
} | [] |
Not all data files have been recovered, so the information remains incomplete. | {
"entities": {}
} | [] |
Data files associated with Stage 4 kernel modules have not been recovered Table 8 lists recovered data files used by Stage 5 modules. | {
"entities": {}
} | [] |
The associated modules that supposedly manipulate those data files were not recovered. | {
"entities": {}
} | [] |
The following details can be used to help determine whether you have been impacted by this threat. | {
"entities": {}
} | [] |
Blue Coat researchers have uncovered a previously-undocumented, highly automated, and extremely sophisticated framework for performing targeted attacks. | {
"entities": {}
} | [] |
The framework is notable for a number of reasons, including (but not limited to) its use of a cloud-based infrastructure for command-and-control and its use of the WebDAV protocol to send instructions and receive exfiltrated information from compromised systems. | {
"entities": {
"entity": [
{
"text": "The framework",
"start": 0,
"end": 13
},
{
"text": "a cloud-based infrastructure",
"start": 92,
"end": 120
},
{
"text": "command-and-control",
"start": 125,
"end": 144
... | [] |
Initial malware components were embedded in Rich Text Format (RTF) files. | {
"entities": {}
} | [] |
Exploitation of vulnerabilities in this file format is leveraged to gain remote access to victim's computers. | {
"entities": {
"entity": [
{
"text": "Exploitation of vulnerabilities in this file format",
"start": 0,
"end": 51
},
{
"text": "gain remote access to victim's computers",
"start": 68,
"end": 108
}
],
"action": [
{
"text... | [] |
The framework, thus far, has been using the services of a cloud service provider based in Sweden, CloudMe.com, for its main command-and-control infrastructure. | {
"entities": {
"entity": [
{
"text": "The framework",
"start": 0,
"end": 13
},
{
"text": "the services of a cloud service provider based in Sweden, CloudMe.com",
"start": 40,
"end": 109
},
{
"text": "its main command-and-contro... | [] |
Malware payloads designed for a wide array of potential devices, including home routers and mobile devices running iOS, BlackBerryOS or Android, were also recovered during the course of our research. | {
"entities": {}
} | [] |
The framework is designed in such a way that all post-infection communication (i.e. target surveying, configuration updates, malware updates, and data exfiltration) can be performed via the cloud service. | {
"entities": {
"entity": [
{
"text": "all post-infection communication (i.e. target surveying, configuration updates, malware updates, and data exfiltration)",
"start": 45,
"end": 164
},
{
"text": "the cloud service",
"start": 186,
"end": 203
... | [] |
The malware components of this framework follow a plugin model, where new malware rely on other, previously delivered malware components to interact with the framework. | {
"entities": {}
} | [] |
Initial attacks were largely focused on Russia and a few other Eastern European countries. | {
"entities": {}
} | [] |
However, we have later seen that attackers are interested in targets all over the globe. | {
"entities": {}
} | [] |
The framework is itself target-agnostic, and seems highly automated. | {
"entities": {}
} | [] |
The operational security exhibited by the attackers is very good-among the best we have seen. | {
"entities": {}
} | [] |
Most interaction between attackers and their infrastructure is performed via a convoluted network of router proxies and rented hosts. | {
"entities": {
"entity": [
{
"text": "Most interaction between attackers and their infrastructure",
"start": 0,
"end": 59
},
{
"text": "a convoluted network of router proxies and rented hosts",
"start": 77,
"end": 132
}
],
"action": ... | [] |
Although the attackers have left a few clues, we have been unable to provide attribution with any degree of accuracy. | {
"entities": {}
} | [] |
The use of software vulnerabilities in order to execute malicious software on unsuspecting users' computers is an important parameter to monitor. | {
"entities": {}
} | [] |
This method of attack is not only known to have a considerable success rate, it is also often deployed by resourceful attackers and, as such, marks a threat worth paying attention to. | {
"entities": {}
} | [] |
The use of exploits in document formats like PDF, DOC and RTF is in some ways especially noteworthy. | {
"entities": {}
} | [] |
Documents are commonly exchanged via mail, which make them perfect for email-borne targeted attacks; what is otherwise known as spear phishing. | {
"entities": {}
} | [] |
In March, 2014, Microsoft published information about a new vulnerability in Rich Text Format (RTF). | {
"entities": {}
} | [] |
This vulnerability, named CVE-2014-1761 (Microsoft Word RTF Object Confusion), had already been used effectively by attackers at the time of the announcement. | {
"entities": {}
} | [] |
Two previous vulnerabilities in the RTF file format, known as CVE-2010-3333 and CVE-2012-0158, had become, by that time, mainstays of targeted attacks, so we tracked how attackers implemented this new exploit with keen interest. | {
"entities": {}
} | [] |
By late August, we identified a malware espionage operation that used both the CVE-2014-1761 and CVE-2012-0158 vulnerabilities to trigger execution of the malicious payload, and which leveraged a single cloud service as the backbone of its entire visible infrastructure. | {
"entities": {}
} | [] |
When we examined the suspicious documents, it was discovered that they were somewhat anomalous compared to the run-of-the-mill material. | {
"entities": {}
} | [] |
They turned out to belong to a highly advanced and professional targeted attack framework, which utilized a complex series of techniques to survey potential targets. | {
"entities": {
"entity": [
{
"text": "a highly advanced and professional targeted attack framework",
"start": 29,
"end": 89
},
{
"text": "a complex series of techniques",
"start": 106,
"end": 136
},
{
"text": "survey potential ... | [] |
Due to the many levels of obfuscation and indirection, we named this the Inception framework; but there ends all similarity with the movie by the same name. | {
"entities": {}
} | [] |
Leonardo DiCaprio is not associated with this investigation. | {
"entities": {}
} | [] |
We initially knew little about who the actual targets were; apart from one. | {
"entities": {}
} | [] |
In that particular case we had the actual phishing email, so we knew the apparent recipient–the CEO of a large Russian bank. | {
"entities": {
"entity": [
{
"text": "we",
"start": 24,
"end": 26
},
{
"text": "the actual phishing email",
"start": 31,
"end": 56
}
],
"action": [
{
"text": "had",
"start": 27,
"end": 30
}
]
... | [] |
The email was apparently sent from'' Mrs. World''; note the Mrs., and not Miss-World. | {
"entities": {}
} | [] |
We soon discovered that our malware repository contained several other, similar documents, but these had come from other sources which did not include the email message, or any identifiable information about the targets. | {
"entities": {}
} | [] |
However, the text of the documents covered a variety of topics mostly revolving around Russian issues relating to a variety of business sectors. | {
"entities": {}
} | [] |
The following pages highlight a representative selection of these documents. | {
"entities": {}
} | [] |
All documents that we have found so far have been rather standard Word documents, of the old 97-2003 compatible format based on OLE2. | {
"entities": {}
} | [] |
Such documents can, and typically do, contain quite a bit of metadata: The name of the document creator; the user who edited it most recently; the name of the company whose copy of Word was used to create the document, et al. | {
"entities": {}
} | [] |
Users can optionally configure Word to remove this metadata when a document is saved, and that's exactly what the creator of these documents did, stripping out this potential source of attribution data. | {
"entities": {}
} | [] |
However, Word documents in this format contain additional information, if you know where to look. | {
"entities": {}
} | [] |
All Word documents of this format contain what's known as a File Information Block (FIB). | {
"entities": {}
} | [] |
The FIB contains information about the file's internal structure, and also–to some extent–data on the program used to create the file. | {
"entities": {}
} | [] |
In the case of the samples we analyzed, all of the documents were saved using the same build of Microsoft Word from Office14 (better known as Office 2010). | {
"entities": {}
} | [] |
In addition, documents can contain slack space in which old data remains. | {
"entities": {}
} | [] |
For example, the decoy that came with the attack named'' Organigrama Gobierno Rusia.doc'' contains Visual Basic leftovers indicating that it originally was created on a computer that was configured to be used by a native Spanish speaker, apparently by an advisor at the Spanish Embassy in Moscow. | {
"entities": {}
} | [] |
This document was presumably obtained by the attackers and repurposed for the attack. | {
"entities": {
"entity": [
{
"text": "This document",
"start": 0,
"end": 13
},
{
"text": "the attackers",
"start": 41,
"end": 54
},
{
"text": "the attack",
"start": 74,
"end": 84
}
],
"action": [
... | [] |
Despite the limited information at our disposal about the targets of these attacks, their content reveals some context about who the possible targets may have been. | {
"entities": {}
} | [] |
First of all, we have the decoy documents which indicate an interest in: We also have a set of phishing mails, which were targeted at: The shellcode used is a pretty standard variant previously used by a number of campaigns typically operating out of China, but with some minor changes. | {
"entities": {
"entity": [
{
"text": "we",
"start": 14,
"end": 16
},
{
"text": "the decoy documents",
"start": 22,
"end": 41
},
{
"text": "We",
"start": 73,
"end": 75
},
{
"text": "a set of p... | [] |
The malicious content is stored inside the document in encoded form, and the shellcode decodes and writes this to disk. | {
"entities": {
"entity": [
{
"text": "The malicious content",
"start": 0,
"end": 21
},
{
"text": "the document",
"start": 39,
"end": 51
},
{
"text": "encoded form",
"start": 55,
"end": 67
},
{
... | [] |
Upon successful execution this code drops a Word document and a Visual Basic script. | {
"entities": {
"entity": [
{
"text": "this code",
"start": 26,
"end": 35
},
{
"text": "a Word document and a Visual Basic script",
"start": 42,
"end": 83
}
],
"action": [
{
"text": "drops",
"start": 36,
... | [] |
The Word document is displayed to the user to avoid arousing any suspicion while the script is executed in the background. | {
"entities": {
"entity": [
{
"text": "the script",
"start": 81,
"end": 91
},
{
"text": "the background",
"start": 107,
"end": 121
}
],
"action": [
{
"text": "is executed",
"start": 92,
"end": 103
}... | [] |
Unusual for many exploit campaigns, the names of the dropped files vary; for example HyHa9AJ.vbs, ewRg.vbs, 0QHdN.vbs, etc. | {
"entities": {
"entity": [
{
"text": "the names of the dropped files",
"start": 36,
"end": 66
}
]
}
} | [] |
–clearly randomized in order to avoid detection by name. | {
"entities": {
"action": [
{
"text": "randomized",
"start": 9,
"end": 19
}
],
"modifier": [
{
"text": "in order to",
"start": 20,
"end": 31
}
],
"entity": [
{
"text": "avoid detection by name",
"start": ... | [] |
The VBScript dropper code is also a little unusual. | {
"entities": {}
} | [] |
It declares a Windows Management Instrumentation (WMI) object in order to reach components like the registry and file system. | {
"entities": {}
} | [] |
This seems adapted from Microsoft example code, like the one found at http: //msdn.microsoft.com/en-us/library/aa387236 (v=vs.85) .aspx | {
"entities": {}
} | [] |
When the VBSript is run it drops two files to disk. | {
"entities": {
"entity": [
{
"text": "it",
"start": 24,
"end": 26
},
{
"text": "two files",
"start": 33,
"end": 42
},
{
"text": "disk",
"start": 46,
"end": 50
}
],
"action": [
{
"text... | [] |
One is a polymorphed dll file and the other a binary data file with no obvious internal structure. | {
"entities": {}
} | [] |
This data file turns out to be encrypted using AES-256. | {
"entities": {
"entity": [
{
"text": "This data file",
"start": 0,
"end": 14
},
{
"text": "AES-256",
"start": 47,
"end": 54
}
],
"action": [
{
"text": "be encrypted",
"start": 28,
"end": 40
}
]... | [] |
The files will be installed in several locations: These locations will vary some between operating system versions. | {
"entities": {
"entity": [
{
"text": "The files",
"start": 0,
"end": 9
},
{
"text": "several locations",
"start": 31,
"end": 48
}
],
"action": [
{
"text": "be installed",
"start": 15,
"end": 27
}
... | [] |
The VBScript then sets a startup key in the'' HKCU\Software\Microsoft\Windows\CurrentVersion\Run'' registry path to execute the DLLs at boot time. | {
"entities": {
"entity": [
{
"text": "The VBScript",
"start": 0,
"end": 12
},
{
"text": "a startup key",
"start": 23,
"end": 36
},
{
"text": "the'' HKCU\\Software\\Microsoft\\Windows\\CurrentVersion\\Run'' registry path",
... | [] |
Regardless of whether the registry launches the DLL or when another malware executable starts the DLL directly, the DLL is launched using regsrv32.exe with the /s (silent) option. | {
"entities": {
"entity": [
{
"text": "the DLL",
"start": 112,
"end": 119
},
{
"text": "regsrv32.exe",
"start": 138,
"end": 150
},
{
"text": "the /s (silent) option",
"start": 156,
"end": 178
}
],
"... | [] |
The names of these dropped files change from attack to attack. | {
"entities": {}
} | [] |
The one above drops ctfmonrc.dll. | {
"entities": {
"entity": [
{
"text": "The one above",
"start": 0,
"end": 13
},
{
"text": "ctfmonrc.dll",
"start": 20,
"end": 32
}
],
"action": [
{
"text": "drops",
"start": 14,
"end": 19
}
]
... | [] |
Other names observed were: The encrypted data files are named using random words apparently taken from a dictionary–'' acholias'','' arzner'','' bicorporate'','' crockrell'','' damnatorily'' etc. | {
"entities": {}
} | [] |
Looking at one of the dropped dlls we can see the authors originally called it 95Num3P3gm.dll.polymorphed.dll. | {
"entities": {}
} | [] |
When executed it will rebuild the original dll (95Num3p3gm.dll, presumably), load it from memory and pass over execution. | {
"entities": {
"action": [
{
"text": "load",
"start": 77,
"end": 81
}
],
"entity": [
{
"text": "it",
"start": 82,
"end": 84
},
{
"text": "memory",
"start": 90,
"end": 96
}
],
"modifier": [
... | [] |
It is hard to describe the polymorphed dlls with any real depth, as there is little consistency between them. | {
"entities": {}
} | [] |
When two nearly identical dlls are encoded using the polymorphic scheme there is very little code in common. | {
"entities": {}
} | [] |
The call graphs are different and key functions have varying number of arguments. | {
"entities": {}
} | [] |
The polymorphing mechanism also generates, and inserts, unique functions all of which make calls to different floating-point operations–all done just to obfuscate the actual decoding process. | {
"entities": {
"entity": [
{
"text": "The polymorphing mechanism",
"start": 0,
"end": 26
},
{
"text": "the actual decoding process",
"start": 163,
"end": 190
}
],
"action": [
{
"text": "obfuscate",
"start": 153,... | [] |
The sizes of buffers allocated are also randomized to mask their intent. | {
"entities": {}
} | [] |
What is common is that somewhere along the execution cycle is one extremely large function (over 200 kb in length) where early in a large allocation is made where the un-obfuscated binary will be placed. | {
"entities": {}
} | [] |
The binary is then built from de- obfuscating segments of it that have been dispersed through the'.rdata' section. | {
"entities": {}
} | [] |
The order, size, and locations of these segments vary from build to build but somewhere near the end of the large function there will be a call to a subfunction that loads the PE image into memory, followed by a call to free the PE image allocation from memory. | {
"entities": {
"entity": [
{
"text": "a call to a subfunction",
"start": 137,
"end": 160
},
{
"text": "the PE image",
"start": 172,
"end": 184
},
{
"text": "memory",
"start": 190,
"end": 196
},
{
... | [] |
Simply halting execution before this function call permits a researcher to extract the reconstructed DLL from memory. | {
"entities": {}
} | [] |
Here, pausing execution before the call to'loadpefrommemory' reveals the extracted PE at the memory address pointed to by edx. | {
"entities": {}
} | [] |
This reconstructed DLL, once loaded, will decode a configuration structure from its'.data' section which contains three important details: the name of the encrypted data file dropped by the VBScript; the AES key used to decrypt the file; and the name of a unique global mutex to hold while running to prevent multiple in... | {
"entities": {}
} | [] |
This configuration information is used to load the encrypted file into memory and decrypt it. | {
"entities": {
"entity": [
{
"text": "This configuration information",
"start": 0,
"end": 30
},
{
"text": "the encrypted file",
"start": 47,
"end": 65
},
{
"text": "memory",
"start": 71,
"end": 77
}
],... | [] |
This turns out to be yet another dll. | {
"entities": {}
} | [] |
The first ordinal exported by this dll is located and then called, passing in the configuration and the name of the encrypted file on disk as parameters. | {
"entities": {}
} | [] |
This last dll is the heart of the threat (originally called q5Byo.dll in this instance. | {
"entities": {}
} | [] |
This file contains the true intent of this campaign. | {
"entities": {}
} | [] |
It is designed as a survey tool. | {
"entities": {
"entity": [
{
"text": "It",
"start": 0,
"end": 2
},
{
"text": "a survey tool",
"start": 18,
"end": 31
}
],
"action": [
{
"text": "is designed",
"start": 3,
"end": 14
}
],
"mo... | [] |
The PE file gathers system information including OS version, computer name, user name, user group membership, the process it is running in, locale ID's, as well as system drive and volume information. | {
"entities": {
"entity": [
{
"text": "The PE file",
"start": 0,
"end": 11
},
{
"text": "system information",
"start": 20,
"end": 38
},
{
"text": "OS version",
"start": 49,
"end": 59
},
{
"tex... | [] |
All of this is encrypted and then sent to cloud storage via WebDAV. | {
"entities": {
"entity": [
{
"text": "All of this",
"start": 0,
"end": 11
},
{
"text": "cloud storage",
"start": 42,
"end": 55
},
{
"text": "WebDAV",
"start": 60,
"end": 66
}
],
"action": [
{... | [] |
WebDAV is a communication standard that allows file management over HTTP or HTTPS. | {
"entities": {}
} | [] |
Windows allows WebDAV sessions to be mapped as network resources. | {
"entities": {}
} | [] |
The use of WebDAV as the communication channel is atypical for most malware samples we see. | {
"entities": {}
} | [] |
By using a network resource, the actual web traffic originates from the system itself, and not from the process in which the malware resides. | {
"entities": {}
} | [] |
Additionally, once the resource is established, the malware can transfer files to and from the command and control servers using standard file IO commands. | {
"entities": {
"entity": [
{
"text": "the malware",
"start": 48,
"end": 59
},
{
"text": "files",
"start": 73,
"end": 78
},
{
"text": "the command and control servers",
"start": 91,
"end": 122
},
{
... | [] |
All the authentication information for the WebDAV session including the URL, folders, path, user name, and password is stored within this last DLL in another AES-encrypted configuration structure in the binary. | {
"entities": {}
} | [] |
A unique path, username, and password were used for each malware instance we've seen in the wild. | {
"entities": {}
} | [] |
This allows the attackers to uniquely identify every targeted attack and track how successful each phishing campaign is. | {
"entities": {
"entity": [
{
"text": "the attackers",
"start": 12,
"end": 25
},
{
"text": "every targeted attack",
"start": 47,
"end": 68
}
],
"action": [
{
"text": "identify",
"start": 38,
"end": 46
... | [] |
Also contained within the configuration structure is information on how to name the survey data on the remote file server. | {
"entities": {}
} | [] |
The binary reads from its configuration a string on how to generate the remote filename, and a list of extensions to use. | {
"entities": {}
} | [] |
An example would be'' 1-7d0-8s'', ['' TIF'','' TAR'','' SIT''] which instructs the binary to generate a filename with 1 to 7 numeric digit characters followed by 0 to 8 ASCII letters with one of the three listed extensions such as'' 664gher.TAR''. | {
"entities": {}
} | [] |
The survey is then uploaded to the server in a specified folder with the generated name. | {
"entities": {
"entity": [
{
"text": "The survey",
"start": 0,
"end": 10
},
{
"text": "the server",
"start": 31,
"end": 41
},
{
"text": "a specified folder",
"start": 45,
"end": 63
},
{
"text... | [] |
Files are compressed using a modified LZMA-compression and encrypted using AES cipher-block-chaining (CBC) before being uploaded to the cloud server. | {
"entities": {
"entity": [
{
"text": "Files",
"start": 0,
"end": 5
},
{
"text": "a modified LZMA-compression",
"start": 27,
"end": 54
},
{
"text": "AES cipher-block-chaining (CBC)",
"start": 75,
"end": 106
... | [] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.