Posts

Showing posts with the label TamperedChef

Malware Analysis: Fake Google Meet Application

Image
Malware Analysis of a fake Google Meet Application that drops SystemShock Loader. TL;DR “Fake Google Meet installer unpacks an Electron app that performs anti-analysis checks and downloads follow-on stealer/downloader payloads (not signed by Google). Check out more information on SystemShock Loader .”  While hunting suspicious Electron applications in VirusTotal, I came across  Google_Meet 1.2.1.exe , which isn't signed by Google. The application is instead signed by " Gucheng County Sili Technology Co., Ltd .", a now revoked certificate signer.  However, VirusTotal has, at the time of this writing, 0 detections for this. So how bad can it be? This is an NSIS installer executable, meaning I can extract the files with 7Zip. Inside the $PLUGINSDIR, is an app-64.7z file, in recent EvilAI and TamperedChef campaigns this usually means this is an Electron based application that is extracted. Quick Note on EvilAI...

Elevate.exe Understanding the Uses and Abuses

Image
TL;DR elevate.exe is an open-source UAC helper commonly bundled with Windows installers and Electron builds — usually legitimate, but sometimes recompiled/signed and abused by threat actors (observed in recent EvilAI, TamperedChef, and BaoLoader campaigns). I was reading through an article by TRUESEC , and one thing that struck myself and my peers was the elevate.exe file that exists in the samples that were analyzed. I mentioned to my peer that I see this file in many electron apps I've analyzed this year. We kicked around the idea of how to build detection for this, but we needed to understand what it is. Firstly, as pointed out by TRUESEC, elevate.exe is a tool by Johannes Passing, found on the github repo . The tool can be shipped/packaged with windows executables which, on their own do not elevate privileges, this helper will invoke UAC to elevate privileges of the designated application. Initially, after readi...

EvilAI: Fake Online Speedtest Application

Image
Several Windows applications that present themselves as legitimate utilities—Internet speed testers, “manual reader” and “finder” tools, certain PDF utilities, and even some AI search frontends such as   justaskjacky   have been observed to drop a portable Node runtime folder alongside a heavily obfuscated JavaScript payload. The visible executable performs as expected to the users, however the installer also extracts the Node runtime, a scheduled task, and  an obfuscated *.js file that don’t appear necessary for the application's primary function. That JavaScript is executed by the dropped Node instance via a scheduled task (observed to run on roughly a 12-hour cycle). Its capabilities include encoded/obfuscated network communications and the potential to execute arbitrary code delivered by the server. Because the JS runs independently from the main executable and is persistent via scheduled tasks, it significantly increases the attack surface...

Tauri Based Malware Analysis on Fake Opulous AI Application

Image
This is just a page of rough notes I'm taking as I go through Opulous analysis from https://x.com/sta5i/status/1962591346407154110 POST https://glitch.footballismy[.]life/api/4/envelope/ HTTP/1.1 x-sentry-auth: Sentry sentry_key=71878a140ea8482c86abc998e4ca02bb, sentry_version=7, sentry_timestamp=1757078735.0221975, sentry_client=sentry.rust/0.42.0 accept: */* host: glitch.footballismy[.]life content-length: 13132 sdk":{"name":"sentry.rust","version":"0.42.0","integrations":["attach-stacktrace","debug-images","contexts","panic","process-stacktrace"],"packages":[{"name":"cargo:sentry","version":"0.42.0"}]}} " POST http://ipc.localhost/loadData HTTP/1.1 Host: ipc.localhost Proxy-Connection: keep-alive Content-Length: 2 sec-ch-ua-platform: "Windows" sec-ch-ua: "Chromium";v="139...

EvilAI: PDF Editor Malware Analysis

Image
 Looking at sample: da3c6ec20a006ec4b289a90488f824f0f72098a2f5c2d3f37d7a2d4a83b344a0 AppSuites-PDF-1.0.28 Specifically looking at the malicious JS file loaded. pdfeditor.js (b3ef2e11c855f4812e64230632f125db5e7da1df3e9e34fdb2f088ebe5e16603). Once this is deobfuscated using  https://obf-io.deobfuscate.io/  I can launch this using node.exe --inspect-brk, then launch edge://inspect to step through the JS. This write-up is a work-in-progress of findings as I go. Here are some of the interesting screenshots so far:  _0x279589 is populated by iterating through and decoding the variables below it. After running through these, here is a list of strings extracted for this sample: A4FDDP7: "ew-key" A6C2XCU: "id" A6C7C7N: "add" A43AUWU: "0.0.0.0" A64CEBI: "usid" A6882RQ: "Url" B4CB2TX: "split" B5D13XX: "https://sdk.appsuites.ai" B5D95P7: "Item2" B5E9U50: "api/s3/event" B40DLY6: "process" ...

EvilAI: Fake Manual Software

Image
UPDATE: September 18, 2025  Another variant of what I've been looking at over the past week caught my eye today. Just wanted to give a real quick review of this, showing the POST encoding/decoding process. OpenMyManual.exe (9f948215b9ee7e7496ce3bc9e46fda56b50cc8905b88535225c7651007f660d5) Drops a copy of node.exe as well as a malicious JS file (2cd68ea7f02e8cfaded52d64c2cb71b64560b3799c948960db37e827618ff22d) Like in my last post, you can run this through a js deobfuscato, make a couple quick edits, and the run node.exe with the --insert-brk to set breakpoints and step through the program: What you eventually get is the C2 (api.evil.com/nss), the POST data, and how that POST data is being encoded and sent out to the C2. The 16 byte buffer is the XOR key, the 176 byte buffer is what becomes the Base64 (shown here as _0xda4df9).  To date, I have not got a response back from any of these C2's. I've looked at several of these, they all so far have a pattern like ^api.[a-zA-...

EvilAI: Another variant of TamperedChef?

Image
  **Update** After a bit of analysis, it looks like I can force the deobfuscated code to run by "patching" following bit: "return (await _0x324dc7(_0x26c49f.HKLM, "Software\\Microsoft\\Cryptography", "MachineGuid")).value;" Once that done, I can use --inspect-brk, use dev view and set a breakpoint at the end, on the eval. In Process Hacker I can see the POST. The Payload can be decoded in cyberchef, it's Base64 decode, send to HEX, use the first 16 bytes as an XOR key, then take the remaining HEX, (From HEX) and apply the XOR key. You will get something like: {"Event":"heartbeat","MachineId":"f4f34c43-9bc1-4a9a-b55f-1d4dd97e0e88","SessionId":"125cc3b3-bc56-474d-817b-badc96d07202","Version":"0.0.2","OSVersion":"10.0.26100","args":""} Since Args doesn't return anything, it's likely not weaponized yet. The deobfuscated...