Skip to main content
ACTIVELY MONITORED//Verified: Nov 16, 2026
Vulnerability WatchCloud DefenseThreat Intelligence
5 min read

CVE-2026-99288: How to Mitigate NoisyTCP Remote Linux Crash

•By CyberNewsAI Threat Research Team•VERIFIED INTEL
CVE-2026-99288: How to Mitigate NoisyTCP Remote Linux Crash

SOC Briefing Summary :: Executive Key Takeaways

  • [01]NoisyTCP (CVE-2026-99288) is an 11-year-old flaw in the Linux kernel TCP networking stack allowing pre-authenticated remote denial of service.
  • [02]Race condition during lockless established hash (ehash) socket lookups triggers NULL pointer dereference panics on multi-queue NICs and RPS setups.
  • [03]Affects all major Linux distributions running default Nginx, Apache, or OpenSSH; mitigation requires kernel updates or disabling Receive Packet Steering.

Executive Incident Briefing (2-Minute CISO Brief)

Quick Answer (At-A-Glance): Security researcher Hyunwoo Kim disclosed NoisyTCP (CVE-2026-99288), an 11-year-old pre-authenticated remote vulnerability in the Linux kernel's TCP networking subsystem affecting nearly all major enterprise distributions. By abusing race conditions during lockless socket lookups (ehash) across multiple CPU cores, remote attackers can trigger a kernel panic that completely crashes bare-metal servers and cloud virtual machines running standard web and SSH services (Nginx, Apache, OpenSSH). While the flaw does not facilitate remote code execution or data leakage, it represents a severe availability risk. System administrators must immediately apply distribution kernel patches or disable Receive Packet Steering (RPS) on multi-queue network interfaces.

The Linux kernel's networking stack is engineered for ultra-high throughput, utilizing lockless data structures to process millions of packets per second. However, subtle concurrency edge cases can expose critical stability risks.

  • Vulnerability: CVE-2026-99288 (NoisyTCP)
  • Vulnerability Type: Remote Pre-Authenticated Denial of Service (Kernel Panic via NULL Pointer Dereference)
  • Affected Systems: Major Linux distributions (RHEL, Ubuntu, Debian, SUSE, CentOS Stream) running Linux kernel versions spanning the past 11 years.
  • Impacted Services: Any network service directly exposed to TCP traffic (Nginx, Apache HTTP Server, HAProxy, OpenSSH).
  • Exploit-DB & Research Context: Discovered by Hyunwoo Kim, who previously disclosed the Dirty Frag exploit chain (CVE-2026-43500 & CVE-2026-43284, documented in Exploit-DB ID 52591) and the Januscape VM escape vulnerability.

While quiet, single-core bare-metal servers present a narrow race window, busy cloud instances and multi-core enterprise servers with multi-queue network cards are highly susceptible to targeted exploitation.

Technical Vulnerability Analysis & Attack Chain

Attack Chain Flow
// Attack Chain Flow

NoisyTCP resides in the lockless established socket lookup table (ehash) implementation within the Linux kernel TCP stack.

1. Concurrency Architecture & Socket Memory Reuse

In high-performance Linux network stacks, when incoming TCP packets arrive, the kernel performs lockless lookups in the ehash table using a 4-tuple hash (source IP, source port, destination IP, destination port). To maximize packet processing speeds, the kernel rapidly recycles memory structures allocated for request sockets (struct request_sock).

Under normal operations, socket allocation, initialization, hashing, and deallocation follow strict atomic sequences. However, Hyunwoo Kim discovered that when packet receive contexts are distributed across distinct CPU cores, recycled socket memory can be read before it has been properly reinitialized and hashed.

2. Multi-Queue RX & Receive Packet Steering (RPS) Trigger

To trigger the race condition remotely, an attacker must cause two specific events to proceed concurrently across different CPU receive (RX) contexts:

  1. A stale ehash lookup on Core A.
  2. The concurrent freeing and reallocation of an existing request socket on Core B.

The attacker achieves this by exploiting differences in receive hash calculations between raw IPv4 fragments and regular TCP packets. On systems utilizing multi-queue Network Interface Cards (NICs) or Receive Packet Steering (RPS), the kernel spreads packet processing across multiple CPU cores based on hash signatures. By interleaving fragmented packets with standard TCP segments, the attacker orchestrates simultaneous socket reuse across different cores.

3. NULL Pointer Dereference & Kernel Panic

When the concurrent thread accesses the uninitialized socket structure during the stale lookup, the kernel attempts to dereference an invalid or NULL pointer in tcp_v4_rcv() / __inet_lookup_established().

Because the fault occurs in interrupt context within kernel space, the kernel cannot recover or terminate a single user process; instead, it immediately triggers an Oops followed by a fatal Kernel Panic, halting the machine and forcing a reboot or complete system hang.

4. Exploit-DB & PoC Comparison

Unlike Dirty Frag (Exploit-DB ID 52591)—where Hyunwoo Kim chained RxRPC and xfrm page-cache write flaws to achieve local root privilege escalation—NoisyTCP is strictly an availability exploit. It requires no authentication and no local access, making it a powerful weapon for distributed denial-of-service (DDoS) campaigns targeting critical cloud perimeters.

MITRE ATT&CK Tactics, Techniques & Procedures (TTPs)

MITRE ATT&CK • OPERATIONAL TTP MAPPING
TacticTechnique IDTechnique NameOperational Context
Initial AccessT1190Exploit Public-Facing ApplicationRemote transmission of malformed TCP traffic targeting Nginx, Apache, or SSH.
ImpactT1499.004Application or System ExploitationTriggering kernel panic and fatal host crash via TCP ehash race condition.
ImpactT1499.003Operating System ExhaustionForcing remote bare-metal and cloud VM restarts to disrupt enterprise availability.

Detection, Hunting & SOC Mitigation Playbook

1. Immediate Containment Checklist

  • [ ] Update Linux kernel packages to patched distribution releases (apt update && apt upgrade linux-image-generic or dnf update kernel).
  • [ ] Temporary Workaround: Disable Receive Packet Steering (RPS) across network interfaces by setting /sys/class/net/<interface>/queues/rx-*/rps_cpus to 0.
  • [ ] Deploy perimeter firewall rules to drop non-standard or unsolicited IPv4 fragments destined for public-facing TCP service ports.
  • [ ] Ensure kdump is configured to capture vmcore crash dumps for forensic post-incident verification.

2. Behavioral Detection Strategy

  • Monitor system logs (/var/log/messages, dmesg, journalctl) for kernel panic signatures referencing NULL pointer dereference, ehash, or tcp_v4_rcv.
  • Inspect network traffic for sudden spikes in fragmented IPv4 packets preceding server unresponsiveness.
  • Configure out-of-band monitoring to alert on sudden, uncoordinated cloud instance reboots.

3. Verification & Diagnostic Audit Checklist

Because CVE-2026-99288 (NoisyTCP) is a pre-authentication network subsystem race condition causing remote kernel panics without shellcode execution or persistence, defense focuses on network perimeter filters and kernel parameter hardening:

  • [ ] Kernel Version Audit: Identify all running Linux kernel instances below 6.12.18 / 6.6.80 across cloud VMs and bare-metal clusters via automated vulnerability scanners.
  • [ ] Temporary Sysctl Hardening: On unpatched edge hosts unable to reboot immediately, evaluate disabling selective TCP optimizations: sysctl -w net.ipv4.tcp_sack=0.
  • [ ] Receive Packet Steering (RPS) Isolation: Reconfigure RPS / RSS network interface queues to restrict network packet processing to dedicated CPU cores, preventing complete system lockup during hash bucket collisions.
  • [ ] Perimeter IDS / IPS Signatures: Deploy Suricata / Snort network edge rules to detect and drop anomalous TCP SYN bursts exhibiting out-of-order option flags and malformed window scales targeting exposed server ports.

Indicators of Compromise (IOCs)

Indicator TypeValueForensic Description
cveCVE-2026-99288Primary vulnerability tracking NoisyTCP Linux kernel panic flaw
cveCVE-2026-43500Related Dirty Frag RxRPC privilege escalation vulnerability (EDB-ID 52591)
cveCVE-2026-43284Related Dirty Frag xfrm-ESP privilege escalation vulnerability (EDB-ID 52591)
otherExploit-DB ID 52591 (Linux Kernel Local Privilege Escalation)Authoritative Exploit-DB PoC research by same discoverer (Hyunwoo Kim)
otherLinux kernel TCP networking stack (ehash lockless socket lookup)Impacted kernel component responsible for remote race condition

---

Related Topic Cluster Intelligence

Indicators of Compromise (IOCs)
5 Identified
CVECVE-2026-99288
CVECVE-2026-43500
CVECVE-2026-43284
EXPLOIT-DB REFERENCEExploit-DB ID 52591 (Linux Kernel Local Privilege Escalation)
VULNERABLE SUBSYSTEMLinux kernel TCP networking stack (ehash lockless socket lookup)
SHARE INTEL:Reddit