# NoisyTCP: 11-year-old Linux kernel flaw can crash servers remotely

> **TL;DR:** NoisyTCP (CVE-2026-99288) is an 11-year-old Linux kernel flaw that affects nearly every distribution. Attackers need no authentication. A specially crafted TCP packet is enough to trigger a kernel panic and crash the server. The flaw does not allow data tampering or arbitrary code execution, but it can cause a denial of service (DoS). A patch has been merged into the mainline Linux kernel, and users should update soon. If you can't patch yet, you can disable SYN cookies and serialize traffic onto a single RX queue to block exploitation.

- Original: https://migovi.com/en/noisytcp-loi-hong-nhan-linux-sap-may-chu/
- Author: [The Strongest Man on Earth](https://migovi.com/en/author/king/)
- Published: 2026-11-16T17:53:24+07:00
- Categories: News, Technology
- Source: [BleepingComputer](https://www.bleepingcomputer.com/news/linux/new-noisytcp-linux-kernel-flaw-lets-attackers-crash-servers/)

---

The NoisyTCP flaw (CVE-2026-99288) has sat in the Linux kernel for 11 years, letting remote attackers crash servers running Nginx, Apache, or SSH with default settings.

## What is NoisyTCP

NoisyTCP is the name given to a flaw that has lived in the Linux kernel for 11 years. According to Hyunwoo Kim, the security researcher who discovered it, the bug affects nearly every Linux distribution. Servers running Nginx, Apache, or SSH with default configurations are all potentially vulnerable.

It is a pre-authentication remote memory corruption bug tracked as CVE-2026-99288. The root cause is a NULL pointer dereference in the kernel's TCP networking subsystem. Attackers don't need to log in. They only need to send specially crafted TCP packets. These packets exploit a timing gap between SYN cookie handling and socket hashing, crashing the kernel and the operating system of the target server.

## Exploit conditions and impact

When exploited, the server hits a kernel panic and stops working. However, the flaw does not let attackers modify data or run arbitrary code. The main consequence is a denial of service (DoS).

Kim says the race window is very narrow. On a low-traffic bare-metal server, triggering the bug can take a long time. The Red Hat security team explains that because of a race condition in lockless socket lookup, recycled socket memory can be accessed before it is fully initialized and hashed. An attacker who generates traffic that causes a socket to be reused will trigger a NULL pointer dereference and crash the kernel.

![Illustration of the NoisyTCP flaw in the Linux kernel that can crash servers](https://migovi.com/wp-content/uploads/2026/11/noisytcp-loi-hong-nhan-linux-sap-may-chu-1.jpg)

According to Kim, triggering the bug requires a stale ehash lookup and the freeing and reallocation of a request socket to happen concurrently in different RX contexts. The attacker must spread packet processing across contexts using the difference in receive hash values between IPv4 fragments and regular TCP packets. As a result, the bug can be exploited on servers with two or more RX queues, or where RPS distributes packets across multiple CPUs.

## Mitigation and Kim's earlier bugs

A patch for CVE-2026-99288 is available. The fix was merged into the mainline Linux kernel about two weeks ago. For systems that can't be patched or upgraded right away, Kim recommends disabling SYN cookies and serializing traffic onto a single RX queue to prevent the race the exploit depends on. Note that disabling SYN cookies weakens protection against SYN flood attacks.

This isn't the first time Kim has disclosed a serious Linux bug. In May, he published Dirty Frag, a local privilege escalation flaw. It chains two page-cache write bugs, RxRPC (CVE-2026-43500) and xfrm-ESP (CVE-2026-43284), to gain root on major distributions such as Ubuntu, Red Hat Enterprise Linux, CentOS Stream, and Fedora. In July, he followed up with Januscape, a 16-year-old Linux kernel flaw that lets attackers escape a virtual machine and run arbitrary code on the physical host.

**MIGOVI VIEWPOINT**

NoisyTCP doesn't allow takeover of a system, but a remotely triggered kernel crash is still a significant risk for web servers, SSH hosts, and public-facing services that need to stay online. The narrow race window lowers the odds of mass exploitation in the short term. Still, servers with multi-queue network cards or RPS enabled are common in practice. Administrators should prioritize kernel updates as soon as the patch lands in their distribution. Disabling SYN cookies should be treated only as a stopgap, since it trades away SYN flood protection.

Source: [BleepingComputer](https://www.bleepingcomputer.com/news/linux/new-noisytcp-linux-kernel-flaw-lets-attackers-crash-servers/)