A security researcher has publicly released working exploit code for four recently patched Linux kernel vulnerabilities, collectively enabling local users to achieve root privileges. While all four flaws have received fixes in recent kernel updates, the public availability of exploits underscores the urgency for system administrators and privacy-conscious users to update their systems promptly.  The four vulnerabilities, dubbed **DirtyAH6**, **TUNderflow**, **PPPoEject**, and **DiagSpill**, were discovered by researcher **Asim Manizada**. He reported them to the **Linux** kernel security team in mid-July and published a detailed technical write-up with proof-of-concept exploits on September 18, following a coordinated disclosure period. Currently, there are no reports of these flaws being exploited in real-world attacks. Manizada's exploits are tailored to specific kernel builds and can cause system crashes, making them primarily suitable for isolated testing environments. However, the release of public exploit code significantly increases the risk for systems running older, unpatched kernels, especially in multi-user or shared server environments where an attacker might already have low-privileged access. ### What's Affected Three of the four flaws (**DirtyAH6**, **TUNderflow**, and **PPPoEject**) are exploitable by an ordinary user only when unprivileged user namespaces are enabled. This **Linux** feature allows a standard user to operate as root within a private sandbox. Many distributions enable user namespaces by default, which inadvertently grants attackers the necessary network privileges to exploit these vulnerabilities. **DiagSpill** stands out as the exception, requiring no user namespaces or special privileges, provided the system has the **SCTP** networking module available. Manizada also identified that **DirtyAH6** and **DiagSpill** can be triggered remotely, though only in specific scenarios and primarily resulting in system crashes. **DirtyAH6** can crash an **IPv6** router or gateway that adds an **IPsec Authentication Header** in transport mode. **DiagSpill** can crash a host only when non-default **SCTP** options are enabled. While Manizada achieved remote root with **DirtyAH6** in his lab by carefully shaping memory on the target, he noted that doing so remotely without prior memory shaping "looks extremely difficult." For **DiagSpill**, he found no viable path to remote root. He also theorized that the flaws could potentially allow container escapes, though he did not develop such an exploit. ### Recommended Actions It is imperative to update your **Linux** kernel to a version that includes all four fixes. The first stable kernel releases with the complete set of patches are: * 5.10.270 * 5.15.221 * 6.1.188 * 6.6.157 * 6.12.109 * 6.18.50 * 7.2.4 Most users run kernels provided by their distribution (e.g., **Debian**, **Ubuntu**, **Red Hat**, **SUSE**), which maintain their own versioning and release schedules. It is crucial to consult your distribution's security advisories to confirm that an update including these four fixes has been shipped, rather than relying solely on the upstream kernel version numbers. If immediate patching is not feasible, two mitigation steps can reduce risk: * **Disable unprivileged user namespaces**: This action prevents ordinary users from exploiting **DirtyAH6**, **TUNderflow**, and **PPPoEject**. However, it does not mitigate **DiagSpill** or protect against processes already possessing network-admin privileges. * **Disable affected features**: If not in use, consider disabling **AH6**, **TUN/TAP**, **PPPoE**, and **SCTP**. Manizada strongly recommends patching as the primary solution, as other potential exploitation paths for these flaws may exist. ### How the Bugs Work All four vulnerabilities are memory-safety bugs located in different components of the kernel's networking code. Each flaw enables an attacker to corrupt kernel memory, which Manizada then leveraged to obtain a root shell. The underlying coding errors are long-standing, dating back between 10 and 21 years. * **DirtyAH6** resides in the **IPsec** code managing the **IPv6 Authentication Header**. The code trusted a routing-header field without proper verification against the number of addresses, leading a crafted packet to cause an out-of-bounds write. * **TUNderflow** affects the **TUN** and **TAP** virtual network devices. A single value was ambiguously used for both spare space and size. An oversized value passed via **Open vSwitch** led to a size calculation wrap-around, causing packet data to land outside its designated buffer. * **PPPoEject** is a use-after-free vulnerability in the **PPP over Ethernet** code. A pointer to a network buffer was retained while a device routine that could free and move that buffer was called, resulting in subsequent writes to freed memory. * **DiagSpill** is found in the reporting code for **SCTP**. A 16-bit counter tracking connection endpoints would wrap around to zero after 65,536 endpoints. The reporting code then failed to allocate sufficient space but copied the entire list, leading to an approximately 8 MiB out-of-bounds write. ### An AI-Assisted Discovery Manizada revealed that he discovered these four flaws using an AI-assisted process that maps kernel memory layout and reasons about it. The kernel's fix for **DirtyAH6** acknowledges this, crediting his custom AI tooling with an "Assisted-by" line in the commit. This marks the latest in a series of **Linux** kernel privilege escalation flaws disclosed in 2026, several of which were found with the aid of large language models. Manizada previously disclosed a similar **Open vSwitch** flaw, **OVSwrap**, in July. One of the new exploits also reuses a technique from **Dirty Frag**, another **Linux** kernel root vulnerability discovered in May by a different researcher. Manizada indicated in his write-up that this batch likely concludes the public phase of his AI-assisted bug hunting endeavors.