NAT Shortcomings Exposed: Understanding NatJack and Its Implications for Network Security

For years, Network Address Translation (NAT) has served as the backbone for IP address management in extensive networks, primarily designed to address the ongoing IPv4 address shortage. However, recent findings by researcher Malcolm Stagg at Black Hat USA 2026 challenge the presumed security of NAT, creating a pivotal moment for network security professionals.

Stagg unveiled NatJack, a series of vulnerabilities leveraging manipulation of the NAT connection tracking table. Should an attacker reside behind the same NAT boundary as a target, they can hijack active connections, poison DNS information, and execute denial of service attacks without needing prior network activity or complex spoofing methods typical of older Layer 2 threats. Testing for NatJack spanned 32 products and configurations, with all tested systems proving vulnerable.

Understanding NAT Vulnerabilities

NAT was originally introduced in the early 1990s as a stopgap, allowing multiple devices to utilize a single public IP by enforcing a trust model among connected devices. The assumption that all actors within a subnet can be trusted is fundamentally flawed. Stagg noted, “The underlying assumption in many NAT implementations is that the peers are trusted.” This inherent trust creates vulnerabilities that skilled adversaries can exploit.

Historically, NAT security has faced challenges. Techniques like NAT Pinning and NAT Slipstreaming were disclosed in prior years, demonstrating how attackers could manipulate NAT behavior for malicious purposes. NatJack, however, is distinct; it can manipulate NAT directly without needing the victim to interact with a malicious site, thus raising the stakes significantly for net defense.

The NAT Weakness: Four Attack Techniques

NatJack encompasses four primary techniques that exploit vulnerabilities in how NAT tables manage connections:

• TCP connection hijacking: Attackers can close an active connection using forged packets, replacing the corresponding NAT entry with their own. This is facilitated by the RFC 1337 TIME-WAIT Assassination mechanism, allowing rapid connection takeovers.
• DNS response poisoning: This technique intercepts and modifies DNS responses, effectively redirecting queries made by the target user without their knowledge.
• Denial of Service: By exhausting the NAT table, an attacker can render all devices sharing that NAT unable to communicate effectively.
• Connection port identification: Attackers can discover the NAT-assigned ports for active connections, supporting the execution of other attacks.

Vendor Response to Disclosure

Stagg approached the disclosure of NatJack responsibly, yet reactions from vendors have been mixed. Some recognized the vulnerabilities and issued patches, while others dismissed them outright. The Linux kernel security team initially labeled the vulnerabilities as insignificant, a reaction that surprised Stagg. “I was pretty taken aback by that response,” he said.

Eventually, Microsoft intervened, associated with its Azure Kubernetes Service, prompting necessary changes to the Linux kernel that resulted in CVE-2026-63913. Microsoft’s own vulnerability concerning Windows NAT was tagged as CVE-2026-56181.

Other organizations, including Cisco and Apple, rejected the vulnerabilities as inherent design limitations rather than security flaws. They argue that appropriate mitigation strategies exist to counter these issues, thus avoiding the need for a broader recognition of the security risks posed by NAT.

Mitigation Strategies for Network Professionals

While patches are rolling out, network professionals can actively reduce risks associated with NatJack through various strategies, as suggested by Stagg:

• Monitor for abnormal activities: Keeping an eye on NAT table sizes, unusual packet floods, or duplicate IP address appearances can hint at potential breaches.
• Implement source IP protection: Using IP Source Guard settings can mitigate risks by curtailing spoofed packets at routers and firewalls.
• Segregate untrusted traffic: Isolate untrusted users on separate subnets or VLANs to cap individual connections, ideally under 10,000.
• Disable lenient connection modes: Configuration changes to restrict loose connection tracking can fortify defenses.
• Restrict container network access: For Kubernetes workloads, limit network access for untrusted environments, and refrain from operating in privileged modes.
• Isolate cloud workloads: Ensuring untrusted and trusted workloads do not share NAT gateways can prevent cross-contamination.

Stagg cautioned that even with these mitigations, certain attack variations might persist, especially in scenarios involving differing subnets.

The crux of the NatJack revelation underscores the necessity for network professionals to reassess longstanding assumptions about NAT security. “A vast number of networks remain exposed to these vulnerabilities,” he pointed out. “Traditional Layer 2 protections may no longer suffice, and evaluating design choices in the context of modern threat models is essential.”