mneb/malware-text-db · Datasets at Hugging Face

input stringlengths 1 785	output dict	schema listlengths 0 0
We searched for the dis- tinctive SSL certificates in two Internet Census service probes, and SSL certificate scans from ZMap.	{ "entities": {} }	[]
We also contacted a team at TU Munich, who applied our fingerprints to their SSL scanning data.	{ "entities": {} }	[]
Across all of these sources, we obtained 31,345 indicator hits reflecting 555 IP addresses in 48 coun- tries.	{ "entities": {} }	[]
One SSL certificate returned by 175 of the servers was issued by'' /CN=RCS Certification Authority /O=HT srl,'' apparently referring to the name of the spyware and the company.	{ "entities": {} }	[]
Servers for 5 of our FSBSpy samples and 2 of our RCS samples re- sponded with this type of certificate.	{ "entities": {} }	[]
Some servers returned these certificates in chains that in- cluded another distinctive certificate.	{ "entities": {} }	[]
We found 175 distinct IP addresses (including the C & C's for 5 of our FSBSpy samples and 2 of our RCS samples) responded with this second type of certificate.	{ "entities": {} }	[]
We devised two more indicators: one that matched 125 IP addresses, including 7 of our FSBSpy samples' C & C's, and one that matched 2 IP addresses, in Italy and Kazakhstan.	{ "entities": {} }	[]
Server locations: On 11/4/13 we probed all of the IP ad- dresses that we collected, finding 166 active addresses match- ing one of our fingerprints in 29 different countries.	{ "entities": {} }	[]
We sum- marize the top providers and countries in Table 5.	{ "entities": {} }	[]
The prevalence of active servers either located in the USA or hosted by Linode is striking, and seems to indicate a pervasive use of out-of-country web hosting and VPS services.	{ "entities": {} }	[]
In addition, we found: (1) 3 IP addresses on a /28 named'' HT public subnet'' that is registered to the CFO of Hacking Team.	{ "entities": {} }	[]
The domain hackingteam.it resolves to an address in this range.	{ "entities": {} }	[]
(2) An address belonging to Omantel, a majority-state-owned telecom in Oman.	{ "entities": {} }	[]
This address was un- reachable when we probed it; a researcher pointed us to an FS- BSpy sample that contained an Arabic-language bait document about Omani poetry, which talked to a C & C in the UK.	{ "entities": { "entity": [ { "text": "an FS- BSpy sample", "start": 77, "end": 95 }, { "text": "an Arabic-language bait document about Omani poetry", "start": 111, "end": 162 }, { "text": "a C & C in the UK", "start": 1...	[]
(3) 7 IP addresses belonging to Maroc Telecom.	{ "entities": {} }	[]
Moroccan journal- ists at Mamfakinch.com were previously targeted by RCS in 2012.	{ "entities": {} }	[]
(4) Overall, servers in 8 countries with governments deemed'' authoritarian regimes'': Azerbaijan, Kazakhstan, Nigeria, Oman, Saudi Arabia, Sudan, UAE, Uzbekistan.	{ "entities": {} }	[]
Link to Hacking Team: All active servers match- ing one of our signatures also responded peculiarly when queried with particular ill-formed HTTP requests, respond- ing with'' HTTP1/1 400 Bad request'' (should be'' HTTP/1.1'') and a body of'' Detected error: HTTP code 400''.	{ "entities": {} }	[]
Googling for this response yielded a GitHub project em-http-server, a Ruby-based webserver.	{ "entities": {} }	[]
The project's author is listed as Alberto Ornaghi, a software architect at Hacking Team.	{ "entities": {} }	[]
We suspect that the Hacking Team C & C server code may incorporate code from this project.	{ "entities": {} }	[]
Links between servers: We identified many cases where several servers hosted by different providers, and in different countries, returned identical SSL certificates matching our fin- gerprints.	{ "entities": {} }	[]
We also observed 30 active servers used a global IPID.	{ "entities": {} }	[]
Only one active server had neither a global IPID nor an SSL certificate matching our fingerprints.	{ "entities": {} }	[]
We assessed whether servers returning SSL certificates were forwarding to the servers with global IPIDs by inducing bursts of traffic at the former and monitoring the IPID at the latter.	{ "entities": {} }	[]
For 11 servers, we found that the server's activity correlated to bursts sent to other servers.	{ "entities": {} }	[]
We grouped servers by the SSL certificates they returned, and found that each group forwarded to only a sin- gle server, except for one case where a group forwarded to two different IPs (both in Morocco).	{ "entities": { "entity": [ { "text": "each group", "start": 73, "end": 83 }, { "text": "only a sin- gle server", "start": 97, "end": 119 }, { "text": "a group", "start": 147, "end": 154 }, { ...	[]
We also found two groups that forwarded to the same address.	{ "entities": { "entity": [ { "text": "two groups", "start": 14, "end": 24 }, { "text": "the same address", "start": 43, "end": 59 } ], "action": [ { "text": "forwarded", "start": 30, "end": 39 } ...	[]
There was a 1:1 mapping between the remaining 8 addresses and groups.	{ "entities": {} }	[]
We refer to a group along with the server (s) it forwards to as a server group.	{ "entities": {} }	[]
We identified several server groups that may be associated with victims or operators in a certain country.	{ "entities": {} }	[]
Some of these suggest possible further investigation: Turkey: We identified a group containing 20 servers in 9 countries.	{ "entities": {} }	[]
Two RCS and 5 FSBSpy samples from VirusTo- tal communicated with various servers in the group.	{ "entities": { "entity": [ { "text": "Two RCS and 5 FSBSpy samples from VirusTo- tal", "start": 0, "end": 46 }, { "text": "various servers in the group", "start": 65, "end": 93 } ], "action": [ { "text": "communicated",...	[]
The RCS samples also communicated with domains with lapsed registra- tions, so we registered them to observe incoming traffic.	{ "entities": { "entity": [ { "text": "The RCS samples", "start": 0, "end": 15 }, { "text": "domains with lapsed registra- tions", "start": 39, "end": 74 } ], "action": [ { "text": "communicated", "start": 21, ...	[]
We ex- clusively received RCS traffic from Turkish IP addresses.	{ "entities": {} }	[]
(RCS traffic is identifiable based on a distinctive user agent and URL in POST requests.	{ "entities": {} }	[]
) A sample of FSBSpy apparently installed from an exploit on a Turkish server talked to one of the servers in this group.	{ "entities": { "entity": [ { "text": "A sample of FSBSpy", "start": 2, "end": 20 }, { "text": "one of the servers in this group", "start": 88, "end": 120 } ], "action": [ { "text": "talked", "start": 78, ...	[]
We also found server groups containing servers in Uzbek- istan and Kazakhstan; we found FSBSpy samples on Virus- Total uploaded from these countries that communicated with servers in these groups.	{ "entities": { "entity": [ { "text": "FSBSpy samples on Virus- Total uploaded from these countries", "start": 88, "end": 148 }, { "text": "servers in these groups", "start": 172, "end": 195 } ], "action": [ { "text": "c...	[]
In the above cases, save Turkey, the country we have identi- fied is classified as an'' authoritarian regime,'' and may be using Hacking Team products against the types of targets we profile in this paper.	{ "entities": {} }	[]
In the case of Turkey, there are hints that the tool may be employed against dissidents.	{ "entities": {} }	[]
Targeted surveillance of individuals conducted by nation-states poses an exceptionally challenging security problem, given the great imbalance of resources and expertise between the victims and the attackers.	{ "entities": {} }	[]
We have sketched the nature of this problem space as reported to us by targeted individuals in three Middle Eastern countries.	{ "entities": {} }	[]
The attacks include spyware for ongoing monitoring and the use of'' IP spy'' links to deanonymize those who voice dissent.	{ "entities": { "entity": [ { "text": "The attacks", "start": 0, "end": 11 }, { "text": "spyware for ongoing monitoring", "start": 20, "end": 50 }, { "text": "the use of'' IP spy'' links to deanonymize those who voice dissent", ...	[]
The attacks, while sometimes incorporating effective so- cial engineering, in general lack novel technical elements.	{ "entities": { "entity": [ { "text": "The attacks", "start": 0, "end": 11 }, { "text": "effective so- cial engineering", "start": 43, "end": 73 } ], "action": [ { "text": "incorporating", "start": 29, "e...	[]
In- stead, they employ prepackaged tools developed by vendors or acquired from the cybercrime underground.	{ "entities": {} }	[]
This technol- ogy sometimes suffers from what strike us as amateurish mis- takes (multiple serious errors implementing cryptography, bro- ken protocol messages), as does the attackers' employment of it (identifying-information embedded in binaries, C & C servers discoverable via scanning or'' Google hacking'', clusters...	{ "entities": {} }	[]
Some of these errors assisted our efforts to assemble strong circumstantial evidence of governmental origins.	{ "entities": {} }	[]
In addition, we mapped out the global use of two'' governmental'' hacking suites, including identify- ing 11 cases in which they appeared to be used in countries governed by'' authoritarian regimes.	{ "entities": {} }	[]
'' We aim with this work to inspire additional research efforts addressing the difficult problem of how to adequately protect individuals with very limited resources facing very powerful adversaries.	{ "entities": {} }	[]
Open questions include robust, practical detection of targeted attacks designed to exfiltrate data from a victim's computer, as well as detection of and defense against novel at- tack vectors, like tampering with Internet connections to insert malware.	{ "entities": {} }	[]
The task is highly challenging, but the potential stakes are likewise very high.	{ "entities": {} }	[]
An opposition member, reflecting on gov- ernment hacking in Libya, speculated as to why some users would execute files even while recognizing them as potentially malicious:'' If we were vulnerable we couldn't care less.	{ "entities": {} }	[]
.	{ "entities": {} }	[]
.	{ "entities": {} }	[]
we were desperate to get our voices out.	{ "entities": {} }	[]
.	{ "entities": {} }	[]
.	{ "entities": {} }	[]
it was a matter of life or death.	{ "entities": {} }	[]
.	{ "entities": {} }	[]
.	{ "entities": {} }	[]
it was just vital to get this information out.	{ "entities": {} }	[]
'' This work was supported by the National Science Foundation under grants 1223717 and 1237265, and by a Citizen Lab Fel- lowship.	{ "entities": {} }	[]
Any opinions, findings, and conclusions or recom- mendations expressed in this material are those of the authors and do not necessarily reflect the views of the sponsors.	{ "entities": {} }	[]
The authors would like to thank the following individuals for their help in various aspects of our analysis: Bernhard Am- man, Collin D. Anderson, Brandon Dixon, Zakir Durumeric, Eva Galperin, Claudio Guarnieri, Drew Hintz, Ralph Holz, Shane Huntley, Andrew Lyons, Mark Schloesser, and Nicholas Weaver.	{ "entities": {} }	[]
Previous work by Guarnieri on scanning for FinSpy servers found that in response to a request such as GET/, the Bahraini FinSpy C & C server returns a response with the string'' Hallo Steffi''.	{ "entities": {} }	[]
Guarnieri searched a database of such responses compiled by the Critical.IO Internet scanning project, locating 11 additional servers in 10 countries.	{ "entities": {} }	[]
We refer to this fingerprint as α1.	{ "entities": {} }	[]
Concurrent with this ef- fort, we devised our own fingerprint β1 that tested three as- pects of the handshake between a FinSpy infectee and a Fin- Spy C & C server, which follows a custom TLV-based protocol running on ports such as 22, 53, 80, and 443.	{ "entities": {} }	[]
We conducted targeted scanning of several countries using β1, and also con- firmed Guarnieri's findings for those servers still reachable af- ter he published his findings.	{ "entities": {} }	[]
We observed a trend: changes in HTTP response behavior by FinFisher after publication of findings about the software.	{ "entities": {} }	[]
In July 2012, for example, after a post about Bahraini FinSpy samples, servers closed the TCP connection in response to a GET/or HEAD/request (although servers continued to behave consistently with β1.	{ "entities": {} }	[]
Other changes followed later in 2012, including a new response to GET/requests that in- cluded an imperfect copy of an Apache server's HTTP response (the Date header used UTC rather than GMT).	{ "entities": {} }	[]
We fingerprinted this error as α2, and later in 2012 fingerprinted other distinctive behavior in response to GET/requests as α3.	{ "entities": {} }	[]
Subsequent scans of /0 for α2 and α3, and five service probes of the Internet Census for α1 through α3, located several additional servers.	{ "entities": {} }	[]
In Feburary 2013 we identified and finger- printed new HTTP response behavior with α4 and modified β1 to produce β2, which tests only two of the three aspects of the FinSpy handshake (the third test of β1 was broken when Fin- Spy servers were updated to accept types of invalid data they had previously rejected).	{ "entities": {} }	[]
As of 3/13/13, all servers that matched any α fingerprint matched β2.	{ "entities": {} }	[]
Our clients often ask us to assess the threat Russia poses in cyberspace.	{ "entities": {} }	[]
Russia has long been a whispered frontrunner among capable nations for performing sophisticated network operations.	{ "entities": {} }	[]
This perception is due in part to the Russian government's alleged involvement in the cyber attacks accompanying its invasion of Georgia in 2008, as well as the rampant speculation that Moscow was behind a major U.S. Department of Defense network compromise, also in 2008.	{ "entities": {} }	[]
These rumored activities, combined with a dearth of hard evidence, have made Russia into something of a phantom in cyberspace.	{ "entities": {} }	[]
In this paper we discuss a threat group whose malware is already fairly well-known in the cybersecurity community.	{ "entities": {} }	[]
This group, unlike the China-based threat actors we track, does not appear to conduct widespread intellectual property theft for economic gain.	{ "entities": {} }	[]
Nor have we observed the group steal and profit from financial account information.	{ "entities": {} }	[]
The activity that we profile in this paper appears to be the work of a skilled team of developers and operators collecting intelligence on defense and geopolitical issues–intelligence that would only be useful to a government.	{ "entities": {} }	[]
We believe that this is an advanced persistent threat (APT) group engaged in espionage against political and military targets including the country of Georgia, Eastern European governments and militaries, and European security organizations since at least 2007.	{ "entities": {} }	[]
They compile malware samples with Russian language settings during working hours consistent with the time zone of Russia's major cities, including Moscow and St. Petersburg.	{ "entities": {} }	[]
While we don't have pictures of a building, personas to reveal, or a government agency to name, what we do have is evidence of long- standing, focused operations that indicate a government sponsor–specifically, a government based in Moscow.	{ "entities": {} }	[]
We are tracking this group as APT28.	{ "entities": {} }	[]
APT28 targets insider information related to governments, militaries, and security organizations that would likely benefit the Russian government.	{ "entities": {} }	[]
Malware compile times suggest that APT28 developers have consistently updated their tools over the last seven years.	{ "entities": {} }	[]
Since 2007, APT28 has systematically evolved its malware, using flexible and lasting platforms indicative of plans for long-term use.	{ "entities": {} }	[]
The coding practices evident in the group's malware suggest both a high level of skill and an interest in complicating reverse engineering efforts.	{ "entities": {} }	[]
• Malware compile times suggest that APT28 developers have consistently updated their tools over the last seven years.	{ "entities": {} }	[]
• APT28 malware, in particular the family of modular backdoors that we call CHOPSTICK, indicates a formal code development environment.	{ "entities": {} }	[]
Such an environment would almost certainly be required to track and define the various modules that can be included in the backdoor at compile time.	{ "entities": {} }	[]
• APT28 tailors implants for specific victim environments.	{ "entities": {} }	[]
They steal data by configuring their implants to send data out of the network using a victim network's mail server.	{ "entities": { "entity": [ { "text": "They", "start": 0, "end": 4 }, { "text": "data", "start": 11, "end": 15 }, { "text": "their implants", "start": 31, "end": 45 }, { "text": "data", ...	[]
• Several of APT28's malware samples contain counter- analysis capabilities including runtime checks to identify an analysis environment, obfuscated strings unpacked at runtime, and the inclusion of unused machine instructions to slow analysis.	{ "entities": { "entity": [ { "text": "Several of APT28's malware samples", "start": 2, "end": 36 }, { "text": "counter- analysis capabilities", "start": 45, "end": 75 }, { "text": "runtime checks", "start": 86, ...	[]
Indicators in APT28's malware suggest that the group consists of Russian speakers operating during business hours in Russia's major cities.	{ "entities": {} }	[]

We searched for the dis- tinctive SSL certificates in two Internet Census service probes, and SSL certificate scans from ZMap.

{ "entities": {} }

[]

We also contacted a team at TU Munich, who applied our fingerprints to their SSL scanning data.

{ "entities": {} }

[]

Across all of these sources, we obtained 31,345 indicator hits reflecting 555 IP addresses in 48 coun- tries.

{ "entities": {} }

[]

One SSL certificate returned by 175 of the servers was issued by'' /CN=RCS Certification Authority /O=HT srl,'' apparently referring to the name of the spyware and the company.

{ "entities": {} }

[]

Servers for 5 of our FSBSpy samples and 2 of our RCS samples re- sponded with this type of certificate.

{ "entities": {} }

[]

Some servers returned these certificates in chains that in- cluded another distinctive certificate.

{ "entities": {} }

[]

We found 175 distinct IP addresses (including the C & C's for 5 of our FSBSpy samples and 2 of our RCS samples) responded with this second type of certificate.

{ "entities": {} }

[]

We devised two more indicators: one that matched 125 IP addresses, including 7 of our FSBSpy samples' C & C's, and one that matched 2 IP addresses, in Italy and Kazakhstan.

{ "entities": {} }

[]

Server locations: On 11/4/13 we probed all of the IP ad- dresses that we collected, finding 166 active addresses match- ing one of our fingerprints in 29 different countries.

{ "entities": {} }

[]

We sum- marize the top providers and countries in Table 5.

{ "entities": {} }

[]

The prevalence of active servers either located in the USA or hosted by Linode is striking, and seems to indicate a pervasive use of out-of-country web hosting and VPS services.

{ "entities": {} }

[]

In addition, we found: (1) 3 IP addresses on a /28 named'' HT public subnet'' that is registered to the CFO of Hacking Team.

{ "entities": {} }

[]

The domain hackingteam.it resolves to an address in this range.

{ "entities": {} }

[]

(2) An address belonging to Omantel, a majority-state-owned telecom in Oman.

{ "entities": {} }

[]

This address was un- reachable when we probed it; a researcher pointed us to an FS- BSpy sample that contained an Arabic-language bait document about Omani poetry, which talked to a C & C in the UK.

{ "entities": { "entity": [ { "text": "an FS- BSpy sample", "start": 77, "end": 95 }, { "text": "an Arabic-language bait document about Omani poetry", "start": 111, "end": 162 }, { "text": "a C & C in the UK", "start": 1...

[]

(3) 7 IP addresses belonging to Maroc Telecom.

{ "entities": {} }

[]

Moroccan journal- ists at Mamfakinch.com were previously targeted by RCS in 2012.

{ "entities": {} }

[]

(4) Overall, servers in 8 countries with governments deemed'' authoritarian regimes'': Azerbaijan, Kazakhstan, Nigeria, Oman, Saudi Arabia, Sudan, UAE, Uzbekistan.

{ "entities": {} }

[]

Link to Hacking Team: All active servers match- ing one of our signatures also responded peculiarly when queried with particular ill-formed HTTP requests, respond- ing with'' HTTP1/1 400 Bad request'' (should be'' HTTP/1.1'') and a body of'' Detected error: HTTP code 400''.

{ "entities": {} }

[]

Googling for this response yielded a GitHub project em-http-server, a Ruby-based webserver.

{ "entities": {} }

[]

The project's author is listed as Alberto Ornaghi, a software architect at Hacking Team.

{ "entities": {} }

[]

We suspect that the Hacking Team C & C server code may incorporate code from this project.

{ "entities": {} }

[]

Links between servers: We identified many cases where several servers hosted by different providers, and in different countries, returned identical SSL certificates matching our fin- gerprints.

{ "entities": {} }

[]

We also observed 30 active servers used a global IPID.

{ "entities": {} }

[]

Only one active server had neither a global IPID nor an SSL certificate matching our fingerprints.

{ "entities": {} }

[]

We assessed whether servers returning SSL certificates were forwarding to the servers with global IPIDs by inducing bursts of traffic at the former and monitoring the IPID at the latter.

{ "entities": {} }

[]

For 11 servers, we found that the server's activity correlated to bursts sent to other servers.

{ "entities": {} }

[]

We grouped servers by the SSL certificates they returned, and found that each group forwarded to only a sin- gle server, except for one case where a group forwarded to two different IPs (both in Morocco).

{ "entities": { "entity": [ { "text": "each group", "start": 73, "end": 83 }, { "text": "only a sin- gle server", "start": 97, "end": 119 }, { "text": "a group", "start": 147, "end": 154 }, { ...

[]

We also found two groups that forwarded to the same address.

{ "entities": { "entity": [ { "text": "two groups", "start": 14, "end": 24 }, { "text": "the same address", "start": 43, "end": 59 } ], "action": [ { "text": "forwarded", "start": 30, "end": 39 } ...

[]

There was a 1:1 mapping between the remaining 8 addresses and groups.

{ "entities": {} }

[]

We refer to a group along with the server (s) it forwards to as a server group.

{ "entities": {} }

[]

We identified several server groups that may be associated with victims or operators in a certain country.

{ "entities": {} }

[]

Some of these suggest possible further investigation: Turkey: We identified a group containing 20 servers in 9 countries.

{ "entities": {} }

[]

Two RCS and 5 FSBSpy samples from VirusTo- tal communicated with various servers in the group.

{ "entities": { "entity": [ { "text": "Two RCS and 5 FSBSpy samples from VirusTo- tal", "start": 0, "end": 46 }, { "text": "various servers in the group", "start": 65, "end": 93 } ], "action": [ { "text": "communicated",...

[]

The RCS samples also communicated with domains with lapsed registra- tions, so we registered them to observe incoming traffic.

{ "entities": { "entity": [ { "text": "The RCS samples", "start": 0, "end": 15 }, { "text": "domains with lapsed registra- tions", "start": 39, "end": 74 } ], "action": [ { "text": "communicated", "start": 21, ...

[]

We ex- clusively received RCS traffic from Turkish IP addresses.

{ "entities": {} }

[]

(RCS traffic is identifiable based on a distinctive user agent and URL in POST requests.

{ "entities": {} }

[]

) A sample of FSBSpy apparently installed from an exploit on a Turkish server talked to one of the servers in this group.

{ "entities": { "entity": [ { "text": "A sample of FSBSpy", "start": 2, "end": 20 }, { "text": "one of the servers in this group", "start": 88, "end": 120 } ], "action": [ { "text": "talked", "start": 78, ...

[]

We also found server groups containing servers in Uzbek- istan and Kazakhstan; we found FSBSpy samples on Virus- Total uploaded from these countries that communicated with servers in these groups.

{ "entities": { "entity": [ { "text": "FSBSpy samples on Virus- Total uploaded from these countries", "start": 88, "end": 148 }, { "text": "servers in these groups", "start": 172, "end": 195 } ], "action": [ { "text": "c...

[]

In the above cases, save Turkey, the country we have identi- fied is classified as an'' authoritarian regime,'' and may be using Hacking Team products against the types of targets we profile in this paper.

{ "entities": {} }

[]

In the case of Turkey, there are hints that the tool may be employed against dissidents.

{ "entities": {} }

[]

Targeted surveillance of individuals conducted by nation-states poses an exceptionally challenging security problem, given the great imbalance of resources and expertise between the victims and the attackers.

{ "entities": {} }

[]

We have sketched the nature of this problem space as reported to us by targeted individuals in three Middle Eastern countries.

{ "entities": {} }

[]

The attacks include spyware for ongoing monitoring and the use of'' IP spy'' links to deanonymize those who voice dissent.

{ "entities": { "entity": [ { "text": "The attacks", "start": 0, "end": 11 }, { "text": "spyware for ongoing monitoring", "start": 20, "end": 50 }, { "text": "the use of'' IP spy'' links to deanonymize those who voice dissent", ...

[]

The attacks, while sometimes incorporating effective so- cial engineering, in general lack novel technical elements.

{ "entities": { "entity": [ { "text": "The attacks", "start": 0, "end": 11 }, { "text": "effective so- cial engineering", "start": 43, "end": 73 } ], "action": [ { "text": "incorporating", "start": 29, "e...

[]

In- stead, they employ prepackaged tools developed by vendors or acquired from the cybercrime underground.

{ "entities": {} }

[]

This technol- ogy sometimes suffers from what strike us as amateurish mis- takes (multiple serious errors implementing cryptography, bro- ken protocol messages), as does the attackers' employment of it (identifying-information embedded in binaries, C & C servers discoverable via scanning or'' Google hacking'', clusters...

{ "entities": {} }

[]

Some of these errors assisted our efforts to assemble strong circumstantial evidence of governmental origins.

{ "entities": {} }

[]

In addition, we mapped out the global use of two'' governmental'' hacking suites, including identify- ing 11 cases in which they appeared to be used in countries governed by'' authoritarian regimes.

{ "entities": {} }

[]

'' We aim with this work to inspire additional research efforts addressing the difficult problem of how to adequately protect individuals with very limited resources facing very powerful adversaries.

{ "entities": {} }

[]

Open questions include robust, practical detection of targeted attacks designed to exfiltrate data from a victim's computer, as well as detection of and defense against novel at- tack vectors, like tampering with Internet connections to insert malware.

{ "entities": {} }

[]

The task is highly challenging, but the potential stakes are likewise very high.

{ "entities": {} }

[]

An opposition member, reflecting on gov- ernment hacking in Libya, speculated as to why some users would execute files even while recognizing them as potentially malicious:'' If we were vulnerable we couldn't care less.

{ "entities": {} }

[]

.

{ "entities": {} }

[]

.

{ "entities": {} }

[]

we were desperate to get our voices out.

{ "entities": {} }

[]

.

{ "entities": {} }

[]

.

{ "entities": {} }

[]

it was a matter of life or death.

{ "entities": {} }

[]

.

{ "entities": {} }

[]

.

{ "entities": {} }

[]

it was just vital to get this information out.

{ "entities": {} }

[]

'' This work was supported by the National Science Foundation under grants 1223717 and 1237265, and by a Citizen Lab Fel- lowship.

{ "entities": {} }

[]

Any opinions, findings, and conclusions or recom- mendations expressed in this material are those of the authors and do not necessarily reflect the views of the sponsors.

{ "entities": {} }

[]

The authors would like to thank the following individuals for their help in various aspects of our analysis: Bernhard Am- man, Collin D. Anderson, Brandon Dixon, Zakir Durumeric, Eva Galperin, Claudio Guarnieri, Drew Hintz, Ralph Holz, Shane Huntley, Andrew Lyons, Mark Schloesser, and Nicholas Weaver.

{ "entities": {} }

[]

Previous work by Guarnieri on scanning for FinSpy servers found that in response to a request such as GET/, the Bahraini FinSpy C & C server returns a response with the string'' Hallo Steffi''.

{ "entities": {} }

[]

Guarnieri searched a database of such responses compiled by the Critical.IO Internet scanning project, locating 11 additional servers in 10 countries.

{ "entities": {} }

[]

We refer to this fingerprint as α1.

{ "entities": {} }

[]

Concurrent with this ef- fort, we devised our own fingerprint β1 that tested three as- pects of the handshake between a FinSpy infectee and a Fin- Spy C & C server, which follows a custom TLV-based protocol running on ports such as 22, 53, 80, and 443.

{ "entities": {} }

[]

We conducted targeted scanning of several countries using β1, and also con- firmed Guarnieri's findings for those servers still reachable af- ter he published his findings.

{ "entities": {} }

[]

We observed a trend: changes in HTTP response behavior by FinFisher after publication of findings about the software.

{ "entities": {} }

[]

In July 2012, for example, after a post about Bahraini FinSpy samples, servers closed the TCP connection in response to a GET/or HEAD/request (although servers continued to behave consistently with β1.

{ "entities": {} }

[]

Other changes followed later in 2012, including a new response to GET/requests that in- cluded an imperfect copy of an Apache server's HTTP response (the Date header used UTC rather than GMT).

{ "entities": {} }

[]

We fingerprinted this error as α2, and later in 2012 fingerprinted other distinctive behavior in response to GET/requests as α3.

{ "entities": {} }

[]

Subsequent scans of /0 for α2 and α3, and five service probes of the Internet Census for α1 through α3, located several additional servers.

{ "entities": {} }

[]

In Feburary 2013 we identified and finger- printed new HTTP response behavior with α4 and modified β1 to produce β2, which tests only two of the three aspects of the FinSpy handshake (the third test of β1 was broken when Fin- Spy servers were updated to accept types of invalid data they had previously rejected).

{ "entities": {} }

[]

As of 3/13/13, all servers that matched any α fingerprint matched β2.

{ "entities": {} }

[]

Our clients often ask us to assess the threat Russia poses in cyberspace.

{ "entities": {} }

[]

Russia has long been a whispered frontrunner among capable nations for performing sophisticated network operations.

{ "entities": {} }

[]

This perception is due in part to the Russian government's alleged involvement in the cyber attacks accompanying its invasion of Georgia in 2008, as well as the rampant speculation that Moscow was behind a major U.S. Department of Defense network compromise, also in 2008.

{ "entities": {} }

[]

These rumored activities, combined with a dearth of hard evidence, have made Russia into something of a phantom in cyberspace.

{ "entities": {} }

[]

In this paper we discuss a threat group whose malware is already fairly well-known in the cybersecurity community.

{ "entities": {} }

[]

This group, unlike the China-based threat actors we track, does not appear to conduct widespread intellectual property theft for economic gain.

{ "entities": {} }

[]

Nor have we observed the group steal and profit from financial account information.

{ "entities": {} }

[]

The activity that we profile in this paper appears to be the work of a skilled team of developers and operators collecting intelligence on defense and geopolitical issues–intelligence that would only be useful to a government.

{ "entities": {} }

[]

We believe that this is an advanced persistent threat (APT) group engaged in espionage against political and military targets including the country of Georgia, Eastern European governments and militaries, and European security organizations since at least 2007.

{ "entities": {} }

[]

They compile malware samples with Russian language settings during working hours consistent with the time zone of Russia's major cities, including Moscow and St. Petersburg.

{ "entities": {} }

[]

While we don't have pictures of a building, personas to reveal, or a government agency to name, what we do have is evidence of long- standing, focused operations that indicate a government sponsor–specifically, a government based in Moscow.

{ "entities": {} }

[]

We are tracking this group as APT28.

{ "entities": {} }

[]

APT28 targets insider information related to governments, militaries, and security organizations that would likely benefit the Russian government.

{ "entities": {} }

[]

Malware compile times suggest that APT28 developers have consistently updated their tools over the last seven years.

{ "entities": {} }

[]

Since 2007, APT28 has systematically evolved its malware, using flexible and lasting platforms indicative of plans for long-term use.

{ "entities": {} }

[]

The coding practices evident in the group's malware suggest both a high level of skill and an interest in complicating reverse engineering efforts.

{ "entities": {} }

[]

• Malware compile times suggest that APT28 developers have consistently updated their tools over the last seven years.

{ "entities": {} }

[]

• APT28 malware, in particular the family of modular backdoors that we call CHOPSTICK, indicates a formal code development environment.

{ "entities": {} }

[]

Such an environment would almost certainly be required to track and define the various modules that can be included in the backdoor at compile time.

{ "entities": {} }

[]

• APT28 tailors implants for specific victim environments.

{ "entities": {} }

[]

They steal data by configuring their implants to send data out of the network using a victim network's mail server.

{ "entities": { "entity": [ { "text": "They", "start": 0, "end": 4 }, { "text": "data", "start": 11, "end": 15 }, { "text": "their implants", "start": 31, "end": 45 }, { "text": "data", ...

[]

• Several of APT28's malware samples contain counter- analysis capabilities including runtime checks to identify an analysis environment, obfuscated strings unpacked at runtime, and the inclusion of unused machine instructions to slow analysis.

{ "entities": { "entity": [ { "text": "Several of APT28's malware samples", "start": 2, "end": 36 }, { "text": "counter- analysis capabilities", "start": 45, "end": 75 }, { "text": "runtime checks", "start": 86, ...

[]

Indicators in APT28's malware suggest that the group consists of Russian speakers operating during business hours in Russia's major cities.

{ "entities": {} }

[]