A critical vulnerability in the Linux kernel, identified as CVE-2026-99288 and nicknamed NoisyTCP, has been disclosed following its discovery by security researcher Hyunwoo Kim. The flaw, which has existed in the Linux TCP networking stack since October 2015, allows an unauthenticated remote attacker to trigger a system panic, effectively crashing the entire host machine. Because the vulnerability resides within the kernel, it poses a significant risk to multi-tenant environments, as a crash initiated against a service running inside a container or virtual machine can bring down the underlying host and all other processes sharing that kernel.
The root cause of the issue is a race condition within the kernel’s TCP networking code. When the system processes incoming traffic across multiple receive queues or CPUs, two kernel components can conflict, leading to a use-after-free condition where the kernel attempts to access memory that has already been deallocated. This memory corruption forces the operating system to panic, resulting in a denial-of-service state.
Testing conducted by the researcher demonstrated that the exploit is highly effective against virtualized environments. Using QEMU/KVM virtual machines, Kim was able to crash common services such as Nginx, Apache, and SSH within a timeframe ranging from five seconds to two minutes. While bare-metal servers are also susceptible, they proved more resilient during testing; triggering a crash on a physical server running Nginx required approximately 14 hours of targeted exploitation.
Despite the severity of the denial-of-service capability, there is currently no evidence that the vulnerability can be leveraged to steal data or achieve remote code execution. While Kim identified a secondary, more complex use-after-free path, he has not demonstrated a general-purpose remote code execution attack. For now, the confirmed impact remains limited to remote system crashes.
The vulnerability affects nearly all Linux distributions, as the flawed code has been present in the kernel for 11 years. Red Hat has officially acknowledged the severity of the issue, classifying it as Important with a CVSS score of 7.5. The upstream Linux kernel maintainers committed a patch to address the flaw on November 3, and the details were made public on November 15 following the expiration of a disclosure embargo.
Administrators are encouraged to verify whether their specific Linux distribution has released an updated kernel containing the fix. For those unable to patch their systems immediately, the researcher has suggested two temporary workarounds, though both involve significant trade-offs. Disabling TCP SYN cookies can mitigate the risk but leaves the system more vulnerable to SYN flood attacks. Alternatively, restricting incoming packet processing to a single receive queue can prevent the race condition but may result in degraded network performance.
The discovery of NoisyTCP follows Kim’s previous work earlier this year on the Dirty Frag kernel vulnerability, which allowed for local privilege escalation. The longevity of the NoisyTCP bug highlights the challenges of maintaining complex networking code within the Linux kernel, where flaws can persist undetected for over a decade. Organizations running containerized workloads should prioritize patching, as the ability for an attacker to crash a host from an unprivileged container represents a substantial security risk to infrastructure stability.




