Public Exploits Released for Four Linux Kernel Flaws That Enable Local Root
A security researcher has released working exploit code for four Linux kernel vulnerabilities, all of which allow a local user to gain root access. While the vulnerabilities have been patched in recent kernel releases, the public availability of the exploit code raises concerns, especially given that many systems enable unprivileged user namespaces by default. The researcher discovered these flaws using AI-assisted bug hunting techniques, and this release marks the end of his automated search for similar vulnerabilities.
A security researcher, Asim Manizada, has released working exploit code for four Linux kernel vulnerabilities, all of which enable a local user to gain root access on a machine. Kernel maintainers have addressed these flaws in recent kernel releases, including versions 5.10.270, 5.15.221, 6.1.188, 6.6.157, 6.12.109, 6.18.50, and 7.2.4. However, the public availability of these exploits presents a risk, particularly because many Linux distributions enable unprivileged user namespaces by default, providing a pathway for attackers to escalate privileges.
Manizada discovered these flaws using an AI-assisted process that maps how the kernel handles memory and reasons about its layout. He disclosed a similar Open vSwitch flaw, OVSwrap, in July, and this release represents the conclusion of his automated bug hunting efforts. The vulnerabilities are DirtyAH6, TUNderflow, PPPoEject, and DiagSpill.
DirtyAH6 resides in the IPsec code handling IPv6 Authentication Header, where a crafted packet caused an internal pointer to move far out of bounds and write past the buffer. TUNderflow is found in the TUN and TAP virtual network devices, where an oversized value passed in through Open vSwitch caused the size calculation to wrap around, leading to packet data landing outside its buffer. PPPoEject is in the PPP over Ethernet code, where a pointer was kept in a network buffer while a device routine could free and move that buffer, resulting in later writes using freed memory (a use-after-free). DiagSpill is in the reporting code for SCTP, where a counter tracking connection endpoints wrapped back to zero after 65,536 connections, leading to 8 MiB of data being written past the buffer.
To mitigate the risk, Manizada recommends patching to the latest kernel releases. If patching is not immediately possible, disabling unprivileged user namespaces and disabling affected features (AH6, TUN/TAP, PPPoE, and SCTP) can reduce the attack surface. Manizada’s AI-assisted bug hunting process has been credited in the DirtyAH6 fix commit with an ‘Assisted-by’ line. This release follows a trend of Linux kernel privilege escalation flaws discovered through 2026, several of which were aided by large language models.