Skip to main content
CybersecurityVulnerability Management

NoisyTCP Vulnerability Exposes Linux Servers to Crashing Attacks

Linux server room with several server racks in soft, ambient light.

CVE-2026-99288 — an 11-year-old Linux kernel vulnerability — can be triggered remotely to crash servers running Nginx, Apache, and SSH with default configurations.

CVE-2026-99288: what NoisyTCP does

Security researcher Hyunwoo Kim disclosed a pre-authenticated remote corruption flaw in the Linux kernel's TCP networking subsystem that the researcher calls NoisyTCP. The weakness stems from a NULL pointer dereference and allows unauthenticated actors to send maliciously crafted TCP packets that exploit timing between SYN cookie handling and socket hashing, forcing a kernel panic and taking the operating system offline.

Kim emphasized that while exploits for CVE-2026-99288 cause a kernel crash and Denial of Service (DoS), they "don't allow threat actors to alter data or gain arbitrary code execution." The impact, therefore, is a remote service outage rather than persistent system compromise.

Why the TCP timing and socket hashing matter

Red Hat's security team described the underlying fault as a race condition in lockless socket lookups: "Due to a race condition during lockless socket lookups, recycled socket memory can be accessed before it is properly initialized and hashed. A remote attacker could exploit this issue by generating network traffic that drives socket reuse, triggering a NULL pointer dereference and kernel crash that results in a Denial of Service (DoS)."

Kim provided further technical detail on the necessary sequence: "To trigger this vulnerability, a stale ehash lookup and the freeing and reallocation of an existing request socket must proceed concurrently in different RX contexts. Packet processing has to be spread out by using the difference in receive hash between IPv4 fragments and regular TCP packets, so it can be triggered against servers that have two or more RX queues, or where RPS distributes packets across multiple CPUs." In short, successful exploitation depends on a tight race across receive (RX) contexts and socket reuse.

Which systems are affected

The vulnerability affects almost all Linux distributions and any system that receives TCP packets directly and runs Nginx, Apache, or an SSH server. That scope includes default-configured servers presenting TCP services to the network.

Kim cautioned that "the race window is narrow, so triggering it against a quiet bare-metal server can take a long time." The requirement for concurrent events across multiple RX contexts — for example, servers with two or more RX queues or where Receive Packet Steering (RPS) spreads packets across CPUs — shapes which deployments are most practical targets.

Mitigation and patch status

A patch addressing CVE-2026-99288 has already been released and the fix was merged into the mainline Linux kernel two weeks ago. For organizations that cannot apply the patch or upgrade immediately, Kim advised two temporary mitigations: disable SYN cookies and serialize to a single RX queue.

Kim warned that disabling SYN cookies "will unfortunately lower protection against SYN flood attacks." Serializing packet reception to a single RX queue removes the concurrent RX contexts needed to win the race, but it is a configuration trade-off: it can blunt the exploit mechanism at the cost of performance or scalability in high-throughput environments.

What this means for system administrators, procurement leaders, and attackers

  • System administrators and security teams: prioritize deploying the upstream patch or kernel update where feasible. If immediate upgrade is impossible, consider the suggested temporary mitigations — disabling SYN cookies and moving to a single RX queue — while accounting for reduced SYN-flood protection and potential performance impacts.
  • Affected enterprises and procurement leaders: inventory externally-facing Linux hosts that run Nginx, Apache, or SSH with default configurations and plan patch windows. The vulnerability’s reliance on RX queue topology and RPS means some deployments may be at higher operational risk; those should be triaged first.
  • Adversaries and threat actors: the flaw produces a remote DoS through kernel panic but not data modification or code execution, and exploitation requires a narrow timing race. That combination makes opportunistic, high-volume disruption feasible in some environments, while practical exploitation against quiet, well-provisioned bare-metal servers may be slow.

Kim’s disclosure of NoisyTCP follows other recent kernel findings he reported earlier this year: in May he disclosed Dirty Frag, a local privilege escalation chain involving RxRPC (CVE-2026-43500) and xfrm-ESP (CVE-2026-43284), and in July he disclosed Januscape, a 16-year-old kernel vulnerability enabling virtual-machine escape and arbitrary code execution on the host. The sequence of disclosures underscores continuing attention to long-lived kernel code paths that remain exploitable under precise conditions.

The remedy for administrators is straightforward on paper — apply the patch merged into mainline Linux two weeks ago — but the trade-offs Kim describes for temporary mitigations make the operational path more complex for systems that must remain both available and resilient to network-level attacks. How swiftly organizations roll the mainline fix into production kernels will determine whether NoisyTCP becomes a recurring nuisance or a quickly contained incident.

Original story