Welcome to this week’s edition of the Threat Source newsletter.  

Brothers and sisters, gather close for a moment. We are all security followers here gathered in fellowship and community, with one joyful spirit to fight the good fight and do good out there in the security world.   

It is with that spirit that I have to mention Clawdbot. Clawdbot (aka Moltbot or OpenClaw) is a locally run open-source agentic application that acts on your behalf. Want to check into a flight? Reply to an email? Vibe code Skynet? Clawdbot’s got you. As of writing this, it has 157k stars on Github. To make it work, the only teeny tiny thing you have to do is feed Clawdbot all of your private information (like logins, passwords, and API keys) and you’re off to the races. No big deal, right? It completely acts on your behalf, with little input if that’s what you desire. If that just made the hair on the back of your neck stand up a little, yeah, me too.

By now, the security hot mess that is Clawdbot has made its way from obscurity into the mainstream news, and it’s all bad. Shocker.   

This is important. I cannot stress this enough. Everyone in the room who ran as fast as possible and installed Clawd/Moltbot, I need you to rethink things. To make this agentic platform act on your request and/or autonomously, you mustsurrender private information to an unvetted, unsecured agentic engine. Now, as a result, your logins, passwords, and more are sitting in a plaintext file, ripe for easy stealing.   

And then there’s the Skills. You can teach your wildly productive agent to do new things! Edit a spreadsheet! Write GPOs! Play a game of global thermonuclear war! The sky is the limit. All it requires is you to give over complete system admin/root access to your Clawd agent. Just understand that Skills are unvetted and unsecured, and already are being actively exploited.

As disciples of security, we understand installing first and asking questions later is practically asking to get pwnt. It has never panned out well for the end user, but usually quite well for attackers who very much understand the threat landscape. Clawdbot is no exception.   

I need you to be highly skeptical of any AI tool rush. Do not be consumed by The Hype. Much like OpenAI’s Atlas, AI tools are being aggressively released to the market and installed, often with security vulnerabilities everywhere. Resist the urge to throw yourself upon tools or platforms that have rushed to address a market need — they usually had no forethought about security, or just push an unreasonable assumption of risk on the end user.  

Security is being sacrificed on the altar of convenience, as AI outpaces our ability to secure it. Brothers and sisters, I’m not asking you to reject the future. AI is going to neat places. I’m asking you to guard yourself as you walk into it. 

The one big thing 

In Talos’ latest blog, we share the discovery of “DKnife,” a modular Linux-based attack framework that compromises routers and edge devices to intercept network traffic, steal credentials, and deliver malware. Active since at least 2019, DKnife can hijack legitimate software updates and bypass endpoint security, posing a significant risk to both users and organizations. 

Why do I care? 

DKnife can take over routers and edge devices, letting attackers spy on users, steal passwords, and install malware without being easily noticed. Because it can break through traditional antivirus defenses and target many types of devices, even networks with good security could be at risk if these gateway devices are not protected. 

So now what? 

Review and harden the security of routers, gateways, and other Linux-based edge devices. Audit for unauthorized firmware or binaries, make sure you’re enforcing strong authentication and certificate validation, and monitor for unusual traffic patterns or update behaviors. Implement network segmentation and make sure your devices are getting updates directly from trusted vendors. 

Top security headlines of the week 

You mean, other than the mess that is Clawdbot? Sorry, the first headline shows we’re not escaping that any time soon: 

Weaponized VS Code add-on ClawdBot sneaks in ScreenConnect RAT 
Security researchers flagged a malicious VS Code extension named “ClawdBot Agent” on the Visual Studio Marketplace. Microsoft swiftly removed it after a report, but not before it tricked developers into installing a fully functional trojan. (Cyber Press) 

Windows malware uses Pulsar RAT for live chats while stealing data 
A newly discovered Windows malware campaign combines the Pulsar RAT with Stealerv37, using Donut loader shellcode injection into explorer.exe to operate entirely in memory while evading traditional antivirus detection. (HackRead) 

eScan confirms update server breached to push malicious update 
MicroWorld Technologies confirmed unauthorized access to a regional eScan antivirus update server resulted in malicious updates distributed to customers during a two-hour window on January 20. (Bleeping Computer) 

County pays $600,000 to pentesters it arrested for assessing courthouse security 
Two security professionals who were arrested in 2019 after performing an authorized security assessment of a county courthouse in Iowa will receive $600,000 to settle a lawsuit they brought alleging wrongful arrest and defamation. (Ars Technica)

Can’t get enough Talos? 

The TTP: Less ransomware, same problems 
Every quarter, Talos IR reviews the incidents we’ve responded to and looks for meaningful shifts in attacker behavior. Hazel is joined by Joe Marshall and Craig Jackson to break down what trends stood out in Q4. 

IR Tales from the Frontlines 
Go beyond the blog with Cisco Talos IR on February 11. This live session features candid stories, behind-the-scenes insights, and strategic lessons learned from the most critical real-world incidents we faced last quarter. 

UAT-8099: New persistence mechanisms and regional focus 
Talos uncovered a new wave of attacks by UAT-8099 targeting IIS servers across Asia, with a special focus on Thailand and Vietnam. Analysis confirms significant operational overlaps between this activity and the WEBJACK campaign. 

Talos Takes: What encryption can (and can’t) do for you 
Step into the fascinating world of cryptography. Amy, Yuri Kramarz, and Tim Wadhwa-Brown sit down to chat about what encryption really accomplishes, where it leaves gaps, and when defenders need to take proactive measures. 

Upcoming events where you can find Talos 

• S4x26 (Feb. 23 – 26) Miami, FL  

Most prevalent malware files from Talos telemetry over the past week 

SHA256: 9f1f11a708d393e0a4109ae189bc64f1f3e312653dcf317a2bd406f18ffcc507 
MD5: 2915b3f8b703eb744fc54c81f4a9c67f 
Talos Rep: https://talosintelligence.com/talos_file_reputation?s=9f1f11a708d393e0a4109ae189bc64f1f3e312653dcf317a2bd406f18ffcc507 
Example Filename: VID001.exe 
Detection Name: Win.Worm.Coinminer::1201  

SHA256: 90b1456cdbe6bc2779ea0b4736ed9a998a71ae37390331b6ba87e389a49d3d59 
MD5: c2efb2dcacba6d3ccc175b6ce1b7ed0a 
Talos Rep: https://talosintelligence.com/talos_file_reputation?s=90b1456cdbe6bc2779ea0b4736ed9a998a71ae37390331b6ba87e389a49d3d59 
Example Filename: d4aa3e7010220ad1b458fac17039c274_64_Dll.dll 
Detection Name: Auto.90B145.282358.in02  

SHA256: 96fa6a7714670823c83099ea01d24d6d3ae8fef027f01a4ddac14f123b1c9974 
MD5: aac3165ece2959f39ff98334618d10d9 
Talos Rep: https://talosintelligence.com/talos_file_reputation?s=96fa6a7714670823c83099ea01d24d6d3ae8fef027f01a4ddac14f123b1c9974 
Example Filename: d4aa3e7010220ad1b458fac17039c274_63_Exe.exe 
Detection Name: W32.Injector:Gen.21ie.1201  

SHA256: a31f222fc283227f5e7988d1ad9c0aecd66d58bb7b4d8518ae23e110308dbf91 
MD5: 7bdbd180c081fa63ca94f9c22c457376 
Talos Rep: https://talosintelligence.com/talos_file_reputation?s=a31f222fc283227f5e7988d1ad9c0aecd66d58bb7b4d8518ae23e110308dbf91 
Example Filename: d4aa3e7010220ad1b458fac17039c274_62_Exe.exe 
Detection Name: Win.Dropper.Miner::95.sbx.tg  

SHA256: 41f14d86bcaf8e949160ee2731802523e0c76fea87adf00ee7fe9567c3cec610 
MD5: 85bbddc502f7b10871621fd460243fbc 
Talos Rep: https://talosintelligence.com/talos_file_reputation?s=41f14d86bcaf8e949160ee2731802523e0c76fea87adf00ee7fe9567c3cec610 
Example Filename: 85bbddc502f7b10871621fd460243fbc.exe 
Detection Name: W32.41F14D86BC-100.SBX.TG 

SHA256: 38d053135ddceaef0abb8296f3b0bf6114b25e10e6fa1bb8050aeecec4ba8f55 
MD5: 41444d7018601b599beac0c60ed1bf83 
Talos Rep: https://talosintelligence.com/talos_file_reputation?s=38d053135ddceaef0abb8296f3b0bf6114b25e10e6fa1bb8050aeecec4ba8f55 
Example Filename: content.js 
Detection Name: W32.38D053135D-95.SBX.TG

Cisco Talos Blog – ​Read More

Over the past two months researchers have reported three vulnerabilities that can be exploited to bypass authentication in Fortinet products using the FortiCloud SSO mechanism. The first two – CVE-2025-59718 and CVE-2025-59719 – were found by the company’s experts during a code audit (although CVE-2025-59718 has already made it into CISA’s Known Exploited Vulnerabilities Catalog), while the third – CVE-2026-24858 – was identified directly during an investigation of unauthorized activity on devices. These vulnerabilities allow attackers with a FortiCloud account to log into various companies’ FortiOS, FortiManager, FortiAnalyzer, FortiProxy, and FortiWeb accounts if the SSO feature is enabled on the given device.

To protect companies that use both our Kaspersky Unified Monitoring and Analysis Platform and Fortinet devices, we’ve created a set of correlation rules that help detect this malicious activity. The rules are already available for customers to download from Kaspersky SIEM repository; the package name is: [OOTB] FortiCloud SSO abuse package – ENG.

Contents of the FortiCloud SSO abuse package

The package includes three groups of rules. They’re used to monitor the following:

• Indicators of compromise: source IP addresses, usernames, creation of a new account with specific names;
• critical administrator actions, such as logging in from a new IP address, creating a new account, logging in via SSO, logging in from a public IP address, exporting device configuration;
• suspicious activity: configuration export or account creation immediately after a suspicious login.

Rules marked “(info)” may potentially generate false positives, as events critical for monitoring authentication bypass attempts may be entirely legitimate. To reduce false positives, add IP addresses or accounts associated with legitimate administrative activity to the exceptions.

As new attack reports emerge, we plan to supplement the rules marked with “IOC” with new information.

Additional recommendations

We also recommend using rules from the FortiCloud SSO abuse package for retrospective analysis or threat hunting. Recommended analysis period: starting from December 2025.

For the detection rules to work correctly, you need to ensure that events from Fortinet devices are received in full and normalized correctly. We also recommend configuring data in the “Extra” field when normalizing events, as this field contains additional information that may need investigating.

Learn more about our Kaspersky Unified Monitoring and Analysis Platform at on the official solution page.

Kaspersky official blog – ​Read More