Fragnesia and ssh-keysign-pwn: Two New Linux Kernel Vulnerabilities Explained

ssh‑keysign‑pwn (CVE‑2026‑46333): Reading Root Secrets

The second vulnerability, ssh‑keysign‑pwn, targets the Linux kernel’s ptrace access‑control logic. Instead of directly granting root privileges, it enables information disclosure, allowing an unprivileged user to read files normally restricted to root .

The flaw stems from incorrect handling of a process property called “dumpability” inside ptrace_may_access()—a function used to determine whether one process can inspect another’s memory . When the logic is applied to certain tasks without associated memory‑management structures, access checks can fail to enforce required privileges.

In practice, attackers can exploit this to steal highly sensitive data such as:

• SSH host private keys from /etc/ssh/
• Password hashes from /etc/shadow

Public proof‑of‑concept exploits target the setuid‑root helper ssh-keysign, which opens host key files before dropping privileges; attackers attach at just the right moment and extract the open file descriptor or process memory .

The bug was disclosed by security researchers and fixed in multiple stable Linux kernel releases, including 7.0.8, 6.18.31, 6.12.89, 6.6.139, 6.1.173, 5.15.207, and 5.10.256 .

Why These Bugs Are Part of a Broader Wave

Fragnesia and ssh‑keysign‑pwn appeared during an unusual cluster of Linux kernel vulnerability disclosures. Security researchers and vendors noted that several flaws—including Dirty Frag and Copyfail‑related bugs—were reported within roughly two weeks, all affecting areas that enable local privilege escalation after initial compromise .

Many of these issues share characteristics:

• They target core kernel subsystems such as networking or process tracing.
• They assume the attacker already has local unprivileged code execution.
• They allow attackers to escape containment and gain root access or sensitive data.

This pattern matters because once attackers obtain a low‑privilege foothold—through a web application compromise, container breakout, or stolen credentials—these kernel bugs can provide a straightforward path to full system control.

Patches and Immediate Actions for Administrators

Because these are kernel vulnerabilities, mitigation requires installing patched kernels and rebooting into the updated version.

Key remediation steps include:

• Update the kernel immediately. Most distributions have released patched packages; for example, AlmaLinux deployed updated kernels addressing both CVEs via normal system updates .
• Reboot after updating. Installing the patch without rebooting leaves the vulnerable kernel running.
• Verify the running kernel version using

  uname -r

  and ensure it matches a fixed release.

For environments where patching cannot happen immediately, temporary mitigations may reduce exposure.

Temporary mitigations for Fragnesia

If the vulnerable networking modules are not required, administrators can disable them until patched kernels are installed. Suggested mitigations include unloading or blocking modules such as:

• esp4
• esp6
• rxrpc

These mitigations reduce the vulnerable attack surface but may disable IPsec‑related functionality .

Temporary mitigations for ssh‑keysign‑pwn

Some administrators have used stricter ptrace restrictions to block known exploits, such as increasing the kernel setting:

kernel.yama.ptrace_scope=2 or 3

However, this can break debugging tools like strace or gdb, and it should be considered only a temporary workaround until patched kernels are deployed .

The Bottom Line

Fragnesia and ssh‑keysign‑pwn highlight how dangerous post‑compromise kernel vulnerabilities can be. Even if an attacker initially gains only low‑privilege access, these bugs can turn that foothold into full root control or expose critical secrets.

For most organizations, the correct response is straightforward:

• Apply vendor kernel updates that fix CVE‑2026‑46300 and CVE‑2026‑46333.
• Reboot into the patched kernel.
• Rotate sensitive secrets such as SSH host keys if compromise is suspected.

With proof‑of‑concept exploits already public, delaying patch deployment significantly increases the risk of real‑world exploitation.