SignDrop exploit published this week reaches SYSTEM via CLFS
SignDrop turns a CLFS defense into SYSTEM, default passwords unlock cars, bee-js gets backdoored, and AI-driven bug hunting pushes Project Zero to tighten disclosure.
Microsoft built HMAC-based authentication codes into the Common Log File System to shut down a class of bug that had produced 19 of the last 24 CLFS CVEs over five years. A new privilege-escalation exploit called SignDrop uses that signing mechanism to reach SYSTEM. Researcher Abdelhamid Naceri published it this week, and per Cyderes it forces clfs.sys to create a .cnpf signature file as SYSTEM at an attacker-controlled path, which gives a privileged file-creation primitive. Because the resulting file keeps attacker-accessible security attributes, the exploit reopens it and replaces the signed content with a payload DLL. Microsoft says it is investigating.
Two other items this week have the same shape, a protection becoming the way in. Researchers found the admin console on two major automakers’ login servers still sitting behind its default password. Both brands authenticated customers through ForgeRock, the identity platform now part of Ping Identity. From that admin account the researchers took over customer accounts and used them to remotely unlock and start a test car, flash its headlights, and sound its horn. The carmakers fixed it about two hours after disclosure.
The supply-chain case hit @ethersphere/bee-js, the JavaScript client for the Swarm storage network. Version 13.3.0 shipped a malicious postinstall hook that, per SafeDep, pulls a second stage to steal the Swarm node’s wallet key and install a backdoor: a Go remote shell on Linux and macOS, and on Windows a hidden ScreenConnect client running as SYSTEM. The same code went out through Swarm Desktop 0.56.2 on GitHub, and the campaign is tied to SleepyDuck. GitHub logs attribute 13 pushes on November 6 to long-time maintainer nugaon; that day the attacker added GitHub Actions workflows to steal CI credentials and failed, then two days later published the malicious releases through the bee-worker release bot. How they got into the accounts is unknown.
Active exploitation showed up in the usual places. A set of N-able flaws tracked as 4EverAble2Shell is being used to drop cryptocurrency miners, with 21 compromised N-central servers counted across the US, South Africa, Canada, the UK, Germany, France, and Sweden; an N-able patch was due the same day. Google shipped a Chrome fix for CVE-2026-125350, a use-after-free in the XML component rated 8.8 and exploited in the wild, now in 157.0.8103.50/.51.
The economics under all of this are moving. Forescout ran AI models against 10 TCP/IP stacks and found 99 vulnerabilities in 90 hours, 66 of them new and just over half rated critical or high. Static analysis cost $935, about $2.58 per finding, and the AI rediscovered 40% of the bugs Forescout had previously disclosed on those stacks by hand. The firm put the time saved at nine months down to 90 hours. That is the reasoning behind Google Project Zero tightening its disclosure clock: the old policy gave vendors 90 days plus a 30-day soak for rollout, and Project Zero will now disclose 90 days after reporting whether or not a fix exists, dropping to 60 days on February 2, 2027 and 30 days on January 3, 2028.
For anyone running refrigeration controllers, two items are worth the attention. Midnight Blue found six pre-authentication RCE 0-days in Danfoss AK-SM 800A controllers within an hour; details are withheld, but the firm notes the devices lack secure boot, so code execution lets an attacker flash firmware that bricks the controller with no recovery short of hardware intervention. Separately, VulnCheck reports threat actors probing the related Copeland XWEB EVO supervisor, including an unauthenticated root command execution flaw, CVE-2026-124010. Five of that product’s 14 bugs affect the current 5.8.8 firmware, which is end-of-support and will not be patched; Copeland’s guidance is to move to XWEB Pro.
Keep Reading
workosOn October 30, an AI tool breached WorkOS
WorkOS says an attacker used the ARTEX AI offensive framework to breach an internal sales app and access CRM contacts and customer call summaries.
pwn2ownPwn2Own Ireland 2026 paid $1,262,000 for 98 zero-days
Pwn2Own Ireland 2026: 29 teams demonstrated 98 zero-days for $1,262,000, with Ikotas Labs taking Master of Pwn and a $300,000 Pixel 10 hack.
voice agentsGemini 3.8 Live broke two security assumptions
Gemini 3.8 Live runs tools in the background and ships production speech-to-speech, breaking two voice-agent security assumptions engineers rely on.
Latest on the Wire
Full wire →- 150,000 Uber Drivers Exposed in Fireworks BreachBleepingComputer
- 3dfx 1 Cards' Washed-Out Color Output ExplainedHacker News
- Aristotle Brings Formal Verification to RustHacker News
- Ethereum Heist Brothers Acquitted on Wire Fraud ChargesBleepingComputer
New signal daily · RSS
Stay in the loop
New writing delivered when it's ready. No schedule, no spam.