input stringlengths 1 785 | output dict | schema listlengths 0 0 |
|---|---|---|
The binary also checks a separate folder on the cloud service designated to contain new configuration information. | {
"entities": {
"entity": [
{
"text": "The binary",
"start": 0,
"end": 10
},
{
"text": "a separate folder on the cloud service designated to contain new configuration information",
"start": 23,
"end": 113
}
],
"action": [
{
... | [] |
If such a file is present on the server, the malware downloads the new configuration file then deletes it from the server. | {
"entities": {
"entity": [
{
"text": "the malware",
"start": 41,
"end": 52
},
{
"text": "the new configuration file",
"start": 63,
"end": 89
},
{
"text": "it",
"start": 103,
"end": 105
},
{
"... | [] |
The cloud storage provider in every case we have seen was the Swedish company CloudMe.com, which offers free and paid WebDAV cloud storage. | {
"entities": {}
} | [] |
The URI model used by the malware is http: //webdav.cloudme.com/% username% /CloudDrive/ which is a direct reference to file storage. | {
"entities": {}
} | [] |
It must be noted that the CloudMe service is not actively spreading the malicious content; the attackers are only using it for storing their files. | {
"entities": {}
} | [] |
We notified CloudMe.com about the abuse of their services. | {
"entities": {}
} | [] |
Their CEO, Mr. Daniel Arthursson, was none too happy about this, and was very helpful in our further research. | {
"entities": {}
} | [] |
CloudMe has shared a great deal of log information related to this attack. | {
"entities": {}
} | [] |
These indicate that there are many other accounts (over 100) likely related to this attack system. | {
"entities": {}
} | [] |
We have no way of verifying this with absolute certainty, but this is what we regard as a high confidence assumption. | {
"entities": {}
} | [] |
The cloud accounts are not used for one-way communication only. | {
"entities": {}
} | [] |
The malware also checks configured subfolders for updates; and if these are found they will be downloaded, decrypted and used as appropriate. | {
"entities": {
"entity": [
{
"text": "The malware",
"start": 0,
"end": 11
},
{
"text": "configured subfolders",
"start": 24,
"end": 45
},
{
"text": "updates",
"start": 50,
"end": 57
},
{
"tex... | [] |
One such case is the franko7046 account, used against the previously mentioned bank CEO. | {
"entities": {}
} | [] |
In this account there was hidden another encrypted configuration file which the malware downloaded and decrypted. | {
"entities": {
"entity": [
{
"text": "another encrypted configuration file",
"start": 33,
"end": 69
},
{
"text": "the malware",
"start": 76,
"end": 87
}
],
"action": [
{
"text": "downloaded",
"start": 88,
... | [] |
This is how we found the depp3353 account. | {
"entities": {}
} | [] |
In this new account there was another surprise waiting for us–a download folder with two new encrypted files, 921.bin and 922.bin. | {
"entities": {}
} | [] |
Once decrypted, these turned out to be PE executables. | {
"entities": {}
} | [] |
The two new executables are plugins-quite similar to each other and obviously compiled on the same setup. | {
"entities": {}
} | [] |
They are lightweight and intended to pull specific survey information from their target. | {
"entities": {
"entity": [
{
"text": "They",
"start": 0,
"end": 4
},
{
"text": "specific survey information",
"start": 42,
"end": 69
},
{
"text": "their target",
"start": 75,
"end": 87
}
],
"action... | [] |
Of interest, both of the DLLs originally had the same internal name (78wO13YrJ0cB.dll). | {
"entities": {}
} | [] |
Presumably the same PE sanitization script and parameters were used on both. | {
"entities": {}
} | [] |
None of these plugins contain any means of CnC communication. | {
"entities": {}
} | [] |
Instead, when they are executed they are passed a pointer to a function to use for sending data back home. | {
"entities": {
"entity": [
{
"text": "they",
"start": 32,
"end": 36
},
{
"text": "a pointer to a function",
"start": 48,
"end": 71
},
{
"text": "use for sending data back home",
"start": 75,
"end": 105
}
... | [] |
Neither are they ever written to disk. | {
"entities": {}
} | [] |
They are executed in memory only, and once they have completed the memory is freed. | {
"entities": {
"entity": [
{
"text": "They",
"start": 0,
"end": 4
},
{
"text": "memory",
"start": 21,
"end": 27
},
{
"text": "they have completed",
"start": 43,
"end": 62
},
{
"text": "the me... | [] |
This makes these modules extremely stealthy, flexible and compatible with multiple toolsets independent of what CnC method is being used. | {
"entities": {}
} | [] |
921.bin retrieves several datapoints about the infected machine: Domain info; a list of running processes with all loaded modules in each; the list of installed software; and a complete hardware profile of the target machine. | {
"entities": {
"entity": [
{
"text": "921.bin",
"start": 0,
"end": 7
},
{
"text": "several datapoints",
"start": 18,
"end": 36
},
{
"text": "the infected machine",
"start": 43,
"end": 63
},
{
... | [] |
922.bin compiles a dirwalk–a complete listing of every file path–of each fixed drive. | {
"entities": {
"entity": [
{
"text": "922.bin",
"start": 0,
"end": 7
},
{
"text": "a dirwalk–a complete listing of every file path–of each fixed drive",
"start": 17,
"end": 84
}
],
"action": [
{
"text": "compiles",
... | [] |
All of this information is exfiltrated back via the same WebDAV connection. | {
"entities": {
"entity": [
{
"text": "All of this information",
"start": 0,
"end": 23
},
{
"text": "the same WebDAV connection",
"start": 48,
"end": 74
}
],
"action": [
{
"text": "is exfiltrated",
"start": 24,
... | [] |
This model makes it possible to do the intrusion in steps, with verification stages in between; and the files will not be easily found on affected computers. | {
"entities": {}
} | [] |
Based on the information gathered from these modules, the attackers appear to move to the next stage of their attack by placing more new components on the WebDAV shares. | {
"entities": {}
} | [] |
Information about these uploads is limited by the fact that we do not have the AES keys to decrypt much of the uploaded data, but we have been able to see some upload patterns. | {
"entities": {}
} | [] |
What we assume to be third-stage plugins appear on the shares as *.bin files of roughly 72kb. | {
"entities": {}
} | [] |
As with other plugins, these are downloaded and deleted from the share in one go. | {
"entities": {
"entity": [
{
"text": "these",
"start": 23,
"end": 28
},
{
"text": "the share",
"start": 61,
"end": 70
},
{
"text": "one go",
"start": 74,
"end": 80
}
],
"action": [
{
... | [] |
However, the next day, another *.bin file of the same size will be uploaded to the share. | {
"entities": {}
} | [] |
This is a pattern that repeats itself over all live accounts. | {
"entities": {}
} | [] |
It seems that because the plugins exist in memory only, they are injected daily to ensure persistence on victim computers. | {
"entities": {
"entity": [
{
"text": "they",
"start": 56,
"end": 60
},
{
"text": "daily",
"start": 74,
"end": 79
},
{
"text": "ensure persistence on victim computers",
"start": 83,
"end": 121
}
],
... | [] |
Our theory is that this malware is a more typical datastealer, and we have observed that after this type of file is planted on the account, encrypted data uploads from compromised users increase. | {
"entities": {
"entity": [
{
"text": "this malware",
"start": 19,
"end": 31
},
{
"text": "a more typical datastealer",
"start": 35,
"end": 61
},
{
"text": "encrypted data uploads from compromised users",
"start": 140,
... | [] |
Victims of this attack will connect using the Windows WebDAV redirector, and the HTTP request user-agent string will reflect this. | {
"entities": {
"entity": [
{
"text": "Victims of this attack",
"start": 0,
"end": 22
},
{
"text": "the Windows WebDAV redirector",
"start": 42,
"end": 71
}
],
"action": [
{
"text": "connect",
"start": 28,
... | [] |
For Windows XP this will typically be'' Microsoft-WebDAV-MiniRedir/5.1.2600'', and for Windows 7 a common user-agent is'' Microsoft-WebDAV-MiniRedir/6.1.7601''. | {
"entities": {}
} | [] |
Security researchers–and there are a few of them-connect in a variety of ways; first of all, we see a number of connections that are indistinguishable from the way victims connect. | {
"entities": {}
} | [] |
This happens when researchers use lab machines with live internet access to run the malware. | {
"entities": {}
} | [] |
The only way we can tell these are researchers is because they connect from IP address ranges that are unlikely to be victims; and they also tend to consist of short-lived sessions. | {
"entities": {}
} | [] |
Some researchers set up scheduled tasks to scan the shares for new updates and malware. | {
"entities": {}
} | [] |
We see a few variations of these–one typical configuration is where the requests contain a Python-related user-agent string. | {
"entities": {}
} | [] |
Attackers, on the other hand, don't appear to use Windows. | {
"entities": {}
} | [] |
Common across multiple accounts, multiple IP's, and over time, is that the probable attackers have used a HTTP user-agent of'' davfs2/1.4.6 neon/0.29.6''. | {
"entities": {
"entity": [
{
"text": "the probable attackers",
"start": 71,
"end": 93
},
{
"text": "a HTTP user-agent of'' davfs2/1.4.6 neon/0.29.6''",
"start": 104,
"end": 153
}
],
"action": [
{
"text": "used",
... | [] |
We know these are not researchers, because we can see malware files being uploaded by them: We have a log fragment in which the attackers uploaded a sequential series of updates (from 1746.bin to 1774.bin) within 1.5 hours on Sept 17th, spread over 27 different accounts and using 27 different IP addresses in the proces... | {
"entities": {}
} | [] |
The user-agent string shows that attackers likely have used a client based on the open source davfs2 file system for Linux to mount the WebDAV shares. | {
"entities": {}
} | [] |
This client is used when uploading new malware, but also when the attackers scan their shares for new victim updates, in which case the shares are enumerated by requests in a scheduled manner. | {
"entities": {}
} | [] |
The attackers have used a large number of IP addresses to access the shares. | {
"entities": {}
} | [] |
As mentioned above, there is a rotation scheme in place in which a new IP address will be used after a few minutes of access against CloudMe accounts. | {
"entities": {}
} | [] |
These IP's are distributed widely over geographical locations and service providers, with a heavy bias towards South Korean ranges. | {
"entities": {}
} | [] |
At first we thought these IP's belonged to some commercial proxy service, particularly since several such proxy services also offer IP rotation. | {
"entities": {}
} | [] |
However, this turned out to be a wrong assumption. | {
"entities": {}
} | [] |
A superficial examination of the proxy IP addresses that connected to CloudMe showed them to be internet-connected devices of various kinds. | {
"entities": {}
} | [] |
Many were Korean Tera-EP home routers; but there were several other products represented. | {
"entities": {}
} | [] |
It is believed that the attackers were able to compromise these devices based on poor configurations or default credentials. | {
"entities": {
"entity": [
{
"text": "the attackers",
"start": 20,
"end": 33
},
{
"text": "these devices",
"start": 58,
"end": 71
},
{
"text": "poor configurations or default credentials",
"start": 81,
"end": 12... | [] |
We were able to do some forensic work on a compromised Tera-EP TE-800 device and discovered another dimension of the attacker's infrastructure. | {
"entities": {}
} | [] |
Under the ramfs mounted partition we found a stripped and statically linked MIPS-el binary named tail-. | {
"entities": {}
} | [] |
Instances of this were also found under the running process list. | {
"entities": {}
} | [] |
tail- serves as a SOCKS proxy for the attackers. | {
"entities": {
"entity": [
{
"text": "tail-",
"start": 0,
"end": 5
},
{
"text": "a SOCKS proxy",
"start": 16,
"end": 29
},
{
"text": "the attackers",
"start": 34,
"end": 47
}
],
"action": [
{... | [] |
Each sample of the binary we were able to acquire was configured with a unique 32byte blowfish key and a small, encrypted section appended to the end of the binary. | {
"entities": {}
} | [] |
Upon execution the binary uses its hardcoded key to decrypt the configuration section and retrieve the listening port to use for incoming connections. | {
"entities": {
"entity": [
{
"text": "the binary",
"start": 15,
"end": 25
},
{
"text": "the listening port",
"start": 99,
"end": 117
},
{
"text": "use for incoming connections",
"start": 121,
"end": 149
}
... | [] |
This acts as a management interface. | {
"entities": {}
} | [] |
From here the attackers can request a specific port to be opened as one of the following types: SOCKET, SOCKSS, SOCKAT, SOCKS5, or STATUSPORT. | {
"entities": {}
} | [] |
To prevent anyone else from accessing this service all communication on the management interface is encrypted using the same blowfish key. | {
"entities": {}
} | [] |
This means that the attackers must maintain a list of where each of these implants are installed, as well as what port and key each is configured to use. | {
"entities": {
"entity": [
{
"text": "the attackers",
"start": 16,
"end": 29
},
{
"text": "a list of where each of these implants are installed, as well as what port and key each is configured to use",
"start": 44,
"end": 152
}
],
"a... | [] |
This setup makes it difficult to identify embedded devices compromised with this malware by scanning open ports. | {
"entities": {}
} | [] |
In the wild we witnessed the attackers connect to the management port and request SOCKSS connections. | {
"entities": {}
} | [] |
This would open the specified port and wait for configuration data, which consists of a domain name (webdav.cloudme.com), the destination port, and a variable length RC4 key, all of which encrypted using the blowfish key. | {
"entities": {}
} | [] |
Once received the malware would attempt to connect to the domain name on the specified port and would start tunneling all traffic received from the SOCKSS port to the destination and vise-versa. | {
"entities": {
"entity": [
{
"text": "the malware",
"start": 14,
"end": 25
},
{
"text": "the domain name",
"start": 54,
"end": 69
},
{
"text": "the specified port",
"start": 73,
"end": 91
},
{
... | [] |
The communication between the attacker and the SOCKSS is encoded using the RC4 key. | {
"entities": {
"entity": [
{
"text": "The communication between the attacker and the SOCKSS",
"start": 0,
"end": 53
},
{
"text": "the RC4 key",
"start": 71,
"end": 82
}
],
"action": [
{
"text": "is encoded",
"st... | [] |
The graphic below illustrates a typical session. | {
"entities": {}
} | [] |
The router proxy network provides another layer of indirection masking the attackers' infrastructure. | {
"entities": {}
} | [] |
However, because we captured traffic through one of these embedded devices we could identify other parts of their operation. | {
"entities": {}
} | [] |
We identified four IP addresses that connected to the proxy malware: This host belongs to a Bulgarian VPS service and would use the router proxy to connect to webdav.cloudme.com. | {
"entities": {
"entity": [
{
"text": "four IP addresses",
"start": 14,
"end": 31
},
{
"text": "the proxy malware",
"start": 50,
"end": 67
},
{
"text": "This host",
"start": 69,
"end": 78
},
{
... | [] |
This host does all scanning of webdav shares for stolen user data, and also uploads new malware components. | {
"entities": {
"entity": [
{
"text": "This host",
"start": 0,
"end": 9
},
{
"text": "new malware components",
"start": 84,
"end": 106
}
],
"action": [
{
"text": "uploads",
"start": 76,
"end": 83
}
... | [] |
This IP would make connections hourly and poll the status of the router proxy malware. | {
"entities": {}
} | [] |
This machine is most likely used to track which compromised routers are currently available for use. | {
"entities": {}
} | [] |
Traffic from this IP had a very specific purpose: It unlocked routers for proxying in connection with the sending of phishing emails. | {
"entities": {
"entity": [
{
"text": "It",
"start": 50,
"end": 52
},
{
"text": "routers",
"start": 62,
"end": 69
},
{
"text": "proxying",
"start": 74,
"end": 82
},
{
"text": "the sending of p... | [] |
In the wild we observed this IP connect to our router on the malware management port and specify a SOCKSS proxy port to be opened. | {
"entities": {
"entity": [
{
"text": "this IP",
"start": 24,
"end": 31
},
{
"text": "our router",
"start": 43,
"end": 53
},
{
"text": "the malware management port",
"start": 57,
"end": 84
}
],
"act... | [] |
Immediately after, the newly opened port would be connected to by another IP and used to send phishing emails with malicious attachments. | {
"entities": {
"entity": [
{
"text": "the newly opened port",
"start": 19,
"end": 40
},
{
"text": "another IP",
"start": 66,
"end": 76
},
{
"text": "phishing emails with malicious attachments",
"start": 94,
"end... | [] |
However, later we observed that the Email sender IP at VOLIA vanished and the Unlocker server taking over its role as well. | {
"entities": {}
} | [] |
After a router SOCKSS port was opened by Unlocker, this IP would connect to the opened port and tunnel its email traffic through the router. | {
"entities": {
"entity": [
{
"text": "this IP",
"start": 51,
"end": 58
},
{
"text": "the opened port",
"start": 76,
"end": 91
},
{
"text": "its email traffic",
"start": 103,
"end": 120
},
{
"... | [] |
Each of these connections used the correct encryption key, so we know that these accesses came from the attackers and not some opportunistic third party. | {
"entities": {}
} | [] |
Mail proxies: Through our router monitoring we identified two mail proxies used by the attackers. | {
"entities": {
"entity": [
{
"text": "two mail proxies",
"start": 58,
"end": 74
},
{
"text": "the attackers",
"start": 83,
"end": 96
}
],
"action": [
{
"text": "used",
"start": 75,
"end": 79
}
... | [] |
We were later notified by Symantec (thanks, guys!) about a third. | {
"entities": {}
} | [] |
These servers were hosted on domains that were registered by the attackers, using domain names clearly meant to look legitimate. | {
"entities": {}
} | [] |
This is the only time we have seen attackers register domains in this investigation. | {
"entities": {}
} | [] |
The mail proxies were: Registrant WHOIS information seems forged: The connections made from the Ukrainian host to the router were interesting. | {
"entities": {}
} | [] |
After being proxied though the router, each of these would authenticate with one of the dedicated mail proxies and send out phishing attacks. | {
"entities": {
"entity": [
{
"text": "each of these",
"start": 39,
"end": 52
},
{
"text": "phishing attacks",
"start": 124,
"end": 140
}
],
"action": [
{
"text": "send",
"start": 115,
"end": 119
}
... | [] |
From captured traffic it appears that the mail proxies have SOCKSv5 services running on obscure high ports. | {
"entities": {}
} | [] |
We have documented that the attackers log in to these using apparently randomly generated usernames and passwords, a unique pair for each server. | {
"entities": {}
} | [] |
The mail proxy would then relay the spearphishing mail as seen below. | {
"entities": {
"entity": [
{
"text": "The mail proxy",
"start": 0,
"end": 14
},
{
"text": "the spearphishing mail",
"start": 32,
"end": 54
},
{
"text": "seen below",
"start": 58,
"end": 68
}
],
"ac... | [] |
This way the attack can be mistaken to come from legitimate businesses and trusted organizations. | {
"entities": {
"entity": [
{
"text": "the attack",
"start": 9,
"end": 19
},
{
"text": "come from legitimate businesses and trusted organizations",
"start": 39,
"end": 96
}
],
"action": [
{
"text": "be mistaken",
... | [] |
In some cases the organization from which the phishing email originates would appear to be a known associate to the target. | {
"entities": {
"entity": [
{
"text": "the organization from which the phishing email originates",
"start": 14,
"end": 71
},
{
"text": "be a known associate to the target",
"start": 88,
"end": 122
}
],
"action": [
{
"tex... | [] |
The email shown above was one of a number of messages sent to targets in the oil industry. | {
"entities": {
"entity": [
{
"text": "a number of messages",
"start": 33,
"end": 53
},
{
"text": "targets in the oil industry",
"start": 62,
"end": 89
}
],
"action": [
{
"text": "sent",
"start": 54,
"end... | [] |
Investigating the target email addresses, we saw several of these were found in this public document from the World Petroleum Council, including some addresses that are, at the present time, no longer valid. | {
"entities": {}
} | [] |
And then, the ground shifted again. | {
"entities": {}
} | [] |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.