Skip to content

AI security

Attackers hit a Mythos-found HFS bug within a day

4 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

Image: theregister.com

Attackers started exploiting CVE-2026-61500 on October 1. It is a critical authentication bypass in Rejetto HTTP File Server (HFS) that Horizon3 researcher Zach Hanley found with Anthropic’s Mythos, and the attacks began a day after he published his write-up and a video of the exploit, The Register reports. If you run HFS, update to v3.2.1 or later. That release fixes this flaw, which can lead to full admin access and remote code execution, along with other security bugs.

HFS is an open source web file server. It had already been on the US Cybersecurity and Infrastructure Security Agency’s Known Exploited Vulnerabilities catalog in 2024. That catalog is CISA’s list of bugs confirmed to be used in real attacks. Hanley’s video shows the steps to exploit the server and run code on it remotely.

Where the attacks came from

“We started detecting exploitation of CVE-2026-61500 in Rejetto HFS this evening,” VulnCheck security researcher Patrick Garrity posted on LinkedIn on Thursday. He added that Hanley and his team had reported the bug to VulnCheck for CVE assignment. “Our canaries detected an actor in China targeting real vulnerable hosts in the US,” Garrity wrote.

Garrity told The Register that Thursday night’s activity came from a single IP address in China and targeted vulnerable servers in the US and Japan. On Friday he counted four hits from two US addresses, 173.239.211[.]248 and 173.239.211[.]249. Both sit in the same subnet and “appear to be coming from a proxy,” he said. China-linked intruders routinely route traffic through compromised devices to hide where they really are. In April, a security advisory from 10 countries warned that China-nexus operators use proxy networks “strategically, and at scale.”

How Mythos got to the key

The bug is in how HFS authenticates users. HFS generates a random value with Math.random() and passes it to Koa, the Node.js web framework HFS is built on. Koa then uses keygrip to sign every session cookie with that value. If an attacker can work out the signing key, they can forge valid session cookies. That should be impossible as long as the random number generator is cryptographically secure.

V8’s Math.random() is not. Mythos found that the output of the xorshift128+ algorithm behind it was fully reversible, and that HFS was leaking raw Math.random() outputs through a separate code path. The model then suggested using Z3 to recover the generator’s seed. Z3 is Microsoft’s publicly available SMT solver, a tool that searches for values satisfying a set of logical constraints. Hanley says Horizon3’s researchers could not remember seeing an SMT solver used this way to break cryptography in a real application and bypass authentication.

“What makes this impressive is that Mythos didn’t just flag the insecure PRNG in isolation – it simultaneously identified that the application leaked raw Math.random() outputs through a separate code path, recognized those two facts as a chain, and determined the leak produced exactly the observations needed to make state recovery feasible,” Hanley wrote. He credits the model’s grasp of mathematics and its skill at solving the constraints needed to reach remote code execution. Horizon3 joined Project Glasswing in July and has used Mythos for vulnerability research since then. Hanley says the company has found “many critical vulnerabilities” with it.

The second of 286

Glasswing gives selected partners access to Mythos for defensive work. Anthropic has said the model is too powerful to release to the general public. Garrity has tracked CVEs credited to Mythos and Glasswing since shortly after the program was announced in April. In September his tracker held 225 such CVEs, and one of them, a critical SQL injection in Ghost (CVE-2026-26980), had been exploited in the wild. At the time Garrity said, “There’s a big difference between finding vulnerabilities and whether they’re actually useful to and will be used by threat actors.” He put the share of vulnerabilities that historically get weaponized at “just under one percent to two percent.”

As of Friday the tracker lists 286 CVEs, 61 more than in September. HFS is the second of them known to be exploited in the wild, which puts the exploited share at about 0.7 percent.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.