NatJack NAT Attack Hijacks TCP Sessions and Spoofs DNS

Security researcher Malcolm Stagg has disclosed a new attack class called NatJack that manipulates network address translation connection state to hijack active TCP sessions, spoof DNS responses, and disclose victim IP addresses and mapped ports across Windows, Linux, and macOS systems. Two implementation-specific flaws have been assigned CVEs: CVE-2026-56181 (CVSS score: 8.3) in Windows NAT used by Hyper-V, and CVE-2026-63913 (CVSS score: 8.2) in Linux Netfilter conntrack.

What Happened: NatJack NAT Attack Bypasses Windows and Linux Network Segmentation

The NatJack research was presented at Black Hat USA 2026 and targets a foundational assumption built into many NAT implementations. Most network administrators assume that hosts behind the same NAT are generally trusted and do not manipulate one another’s connection state. However, Stagg demonstrated that an attacker controlling a system behind the same NAT can manipulate connection-tracking entries belonging to another system.

The attack model places an attacker-controlled system behind the same NAT infrastructure as the victim. Furthermore, the exact capabilities and preconditions vary by technique and platform. All tested products were found vulnerable across network infrastructure devices, virtualization, containerization, and cloud infrastructure.

Technical Details of the NatJack Attack Class

NatJack is not a single exploit but a class of five distinct attacks against NAT implementations. The techniques leverage the time between when NAT entries are neutralized and when they are used by the victim.

Five Attack Vectors

1. UDP DNS Hijacking: The attacker fills the NAT table with dummy entries to evict the victim’s pending DNS entry. Then they create replacement entries pointing to themselves. Consequently, the legitimate DNS response is sent to the attacker instead of the victim, after which the attacker sends a spoofed replacement response back to the victim.
2. TCP/IP Hijacking (Downstream Variant): The attacker removes and replaces a victim’s NAT entry to hijack the connection using downstream spoofing techniques.
3. TCP/IP Hijacking (Upstream Variant): The attacker and victim can be in different subnets and broadcast domains. Moreover, this variant enables cross-segment hijacking.
4. Victim IP/Port Disclosure: The attacker can determine the victim’s IP address and mapped ports through NAT table manipulation.
5. Denial of Service: The attacker exhausts NAT table entries to disrupt legitimate traffic.

Implementation-Specific CVEs

Two implementation flaws have been assigned CVE identifiers:

• CVE-2026-56181 (CVSS 8.3): An origin validation error in Windows Network Address Translation (NAT) used by Hyper-V. Affected releases include Windows 11 24H2 before 26100.8875, 25H2 before 26200.8875, 26H1 before 28000.2525, and Windows Server 2025 before 26100.33158. Microsoft describes this as an origin-validation error that enables spoofing from an adjacent network.
• CVE-2026-63913 (CVSS 8.2): In the Linux kernel, the netfilter conntrack subsystem contains a flaw where a crafted SYN followed by a reset packet with an invalid sequence number can prematurely force an active NAT entry into a closed state. Therefore, the conntrack logic failed to validate direction before forcing the state change. Fixed stable releases include 5.10.259, 5.15.210, 6.1.176, 6.6.143, 6.12.93, 6.18.35, 7.0.12, and 7.1.

Impact Across Platforms

Stagg tested 32 products and configurations across 13 vendors. Consequently, all tested products were found vulnerable. The research demonstrated proof-of-concept exploitation in controlled environments. In addition, Stagg demonstrated a macOS virtualization attack. Apple considers the behavior a known transport-layer limitation rather than a vulnerability, but said it was considering additional hardening as a defense-in-depth measure.

Business and Operational Impact

The NatJack attack class poses significant risks for organizations that rely on NAT-based network segmentation as a security control. The impact extends across several domains:

• Cloud and Virtualization Infrastructure: Organizations using Hyper-V, containerization, and cloud NAT services face session hijacking risks between tenants or workloads sharing NAT infrastructure.
• Network Infrastructure Devices: Enterprise routers, firewalls, and edge devices may be vulnerable to DNS hijacking and connection manipulation between internal hosts.
• Remote Work Environments: Users behind corporate NAT gateways could have their sessions intercepted by compromised devices on the same internal network.
• Cross-Tenant Risk: In shared hosting and cloud environments, untrusted workloads sharing NAT infrastructure could attack other tenants’ connections.

There is no single patch for the broader NatJack attack class. Therefore, organizations must adopt defense-in-depth strategies rather than relying on a single fix.

Mitigation and Recommendations

Immediate Actions for Defenders

1. Apply Available Patches: Install Windows and Linux kernel updates addressing CVE-2026-56181 and CVE-2026-63913 as soon as possible.
2. Separate Untrusted Workloads: Isolate untrusted systems from trusted systems that share NAT infrastructure. Use VLANs, isolated subnets, or separate NAT gateways.
3. Encrypt Internal Traffic: Encrypt traffic even within internal networks to prevent session hijacking from yielding plaintext data.
4. Enable IP Source Guard: Where applicable, implement IP Source Guard to prevent spoofed traffic from reaching other hosts behind the same NAT.
5. Monitor for Anomalous NAT Behavior: Monitor NAT table exhaustion, unusual DNS redirections, and unexpected RST packets on internal networks.

Long-Term Architecture Changes

Organizations should reevaluate the assumption that NAT alone provides sufficient isolation between internal hosts. Moreover, network architects should consider microsegmentation approaches that do not rely on NAT state as a security boundary.

Bottom line: NatJack reveals that NAT-based network segmentation is weaker than previously assumed. Organizations should apply available patches for CVE-2026-56181 and CVE-2026-63913, separate untrusted workloads from trusted systems behind shared NAT infrastructure, and encrypt internal traffic. Furthermore, network architects must redesign internal segmentation to not rely on NAT state as a security boundary.

Incident Summary

CVE ID / Incident:	CVE-2026-56181 (Windows NAT, CVSS 8.3) and CVE-2026-63913 (Linux Netfilter, CVSS 8.2)
Affected Systems:	Windows 11 24H2/25H2/26H1, Windows Server 2025, Linux kernel 5.10 through 7.1, network infrastructure devices, virtualization and container platforms, macOS virtualization
Disclosure Date:	August 6, 2026 (Black Hat USA 2026)
Patch Status:	Partial — Linux kernel fixes released; Windows patches available; broader NatJack class requires architectural changes

References

1. Stagg, Malcolm. “NatJack: Manipulating NAT Connection State.” Black Hat USA 2026. SODIUM-24. https://natjack.io/, accessed August 7, 2026.
2. National Vulnerability Database. “CVE-2026-56181 Detail.” https://nvd.nist.gov/vuln/detail/CVE-2026-56181, accessed August 7, 2026.
3. National Vulnerability Database. “CVE-2026-63913 Detail.” https://nvd.nist.gov/vuln/detail/CVE-2026-63913, accessed August 7, 2026.
4. The Hacker News. “Malware Can Abuse Windows Hello for Business Keys for Persistent Entra ID Access.” August 7, 2026. https://thehackernews.com/2026/08/malware-can-abuse-windows-hello-for.html, accessed August 7, 2026.
5. BleepingComputer. “New TONTOU CPU attack bypasses Spectre v2 fixes, leaks Linux password hashes.” August 6, 2026. https://www.bleepingcomputer.com/news/security/new-tontou-cpu-attack-bypasses-spectre-v2-fixes-leaks-linux-password-hashes/, accessed August 7, 2026.