For those playing along at home, it's very much confirmed that the two PHP configuration changes don't mitigate the issues:
https://t.co/J3Ok3MzPxc
Having seen the patches, I can't stress the importance of patching enough. There be dragons.
PolinRider updating malware in an infected git repo with a "remove malware" message seems like a rare example of levity from a DPRK actor. You see it all the time from ecrime and some other actors but DPRK are usually all business in my experience.
@timboloman The same compromised account that commits the action file runs it once, shortly after dropping it. Same account that commits the PolinRider files
There's a strange overlap between PolinRider and a malicious Action spreading across GitHub. The Action collects runner secrets, vars and cloud credential files into a tarball, base64-encodes it and POSTs it to hxxps[:]//webhook[.]site/b61ec04f-205c-44b2-bb07-f9ae2d9741a0. The Action on runs when triggered manually.
The Action has been active since 27 September and has been planted in at least 450 repos across 56 namespaces. The webhook has received enough data to exceed its free limit. Around 1 in 10 infected repos also carry PolinRider, and in every case I checked, PolinRider was there first, for days to months before the Action arrived. It's hard to say at present whether this is victimisation by two independent threat actors or a new addition to the PolinRider toolkit.
PolinRider gets less attention than worms but is by far the most persistent GitHub-spreading threat, and stolen CI/CD credentials make a foothold in production infra or software supply chains more likely. One to watch.
The threat actor's ethereum wallet, used to trigger the blockchain dead drop C2 resolver: 0xb614155Fd88114d40549b259457Bcf921Df091B9
The transaction listing the C2 happened back on 21 September, so this has likely been cooking for a while
Drops hackbrowserdata binary to hosts (e.g. for linux, 16b8a1e335b7ad9856c65e3745950a5f76a197560749af1f0bb097de4ac57ad1) to obtain more browser credentials. Expanded collection from browser extensions as well, more oriented towards endpoint than CICD targeting
Math_Symbol.js b50a00900399ba99fb6ce1fc151519cb99d44320ef2a631f2237e1aea0ad6fec
Same structure as the keyv Shai-Hulud payloads, but a unique "globalThis.WORMTAG='tensrlake';" marker, so not reinfection.
The supplychain.local/MemOS worm from a couple of weeks ago has finally appeared in an infected GitHub repo: https://t.co/u6MMzbwhmy
This is the first apparent replication I've seen to date
Another new worm attempt - new NPM packages react-nodejs and express-javascript contain a postinstall script that sources simple payloads from codeberg that attempt to self-proliferate through NPM and AUR packages. Simple and unlikely to spread based on current foothold but another indication that some threat actors are still chasing Shai-Hulud's tail
https://t.co/ntu57xVRT6
Appears to drop the open source CHAOS golang RMM tool to Linux hosts and make it persistent using a systemd service. Also installs TOR and sets an environment variable to route C2 traffic over TOR
There may be a new wave of Shai-Hulud payloads kicking off.
vbenjs/vben-admin-thin-next and ant-design/pro-components were both compromised on GitHub two hours ago with obfuscated .claude and VSCode task files. The payload is the same as the May wave, so this may just be ongoing reinfection.