In short: Security researcher Malcolm Stagg presented NatJack at Black Hat USA 2026 — a new attack class that lets anyone sharing a NAT boundary (hotel, café, office, co-working space) hijack active TCP sessions, poison DNS responses, expose mapped ports, and crash the entire network. All 32 tested products across Windows, Linux, and macOS were vulnerable to at least some techniques. The underlying flaw is architectural; encryption is the primary defense.
What Is NatJack
NAT (Network Address Translation) has been internet infrastructure since the early 1990s — it lets many devices share a single public IP address. Home routers, café Wi-Fi, hotel networks, and corporate gateways all rely on it. The implicit assumption baked into every NAT implementation is that devices sharing the same NAT table are "trusted neighbors" — reasonable when that meant a household or a single office floor.
Malcolm Stagg demonstrated at Black Hat USA 2026 that this assumption no longer holds. NatJack is a class of attacks exploiting weaknesses in NAT connection-state tracking tables. Anyone sharing a NAT boundary — a stranger at the next café table, a fellow guest in a hotel, another tenant in a co-working space — can attack your active connections without elevated privileges or physical network access. Our security blog covers emerging threats like these as they are disclosed.
Four Attack Techniques
Stagg identified four distinct vectors, each exploiting the NAT state table differently:
- TCP session hijacking. The attacker forces a victim's NAT mapping into a closed state and redirects the connection to themselves — effectively taking over an established session with a bank, email provider, or any other service.
- DNS response poisoning. The attacker intercepts outbound DNS queries and injects forged responses, silently redirecting the victim to attacker-controlled servers.
- Port identification. The external port that NAT assigned to a victim's active connection can be determined — useful input for further targeted attacks.
- NAT table exhaustion (DoS). Flooding the NAT table with spoofed flows prevents any new legitimate connections from being established, knocking all users off the network.
Critically, none of these techniques require IP spoofing or Layer 2 broadcast domain access — the barriers that made older attacks like ARP poisoning far harder. NatJack operates at a higher level, making it more practical in real-world shared environments.
Scale: 32 Products Tested, All Vulnerable
Stagg notified 13 vendors and tested 32 products across 95 report cases. Every single implementation was vulnerable to at least some NatJack techniques — including systems built on entirely independent codebases: Windows (Hyper-V), Linux (Netfilter conntrack), macOS, and hardware from major networking vendors.
Two CVEs have been assigned:
- CVE-2026-56181 — Windows NAT in Hyper-V (CVSS 8.3)
- CVE-2026-63913 — Linux Netfilter conntrack (CVSS 8.2)
As of August 7, 2026, no exploitation in the wild has been detected — but Stagg notes these attacks are difficult to detect, so absence of reports does not mean absence of attacks.
Who Is at Risk
Anyone using shared networks is in the risk zone: Wi-Fi at cafés, restaurants, hotels, airports, libraries, co-working spaces, and shopping malls. In these environments, dozens of unknown devices share a single NAT infrastructure — exactly the condition NatJack requires.
Corporate users are also exposed if trusted and untrusted segments are not isolated. Cloud environments where multiple tenants share NAT infrastructure carry risk too — one of the CVEs targets Hyper-V, the virtualization platform widely used in Azure-compatible environments.
Why a Complete Patch Is Not Possible
NatJack is an architectural problem, not a bug in a specific codebase. NAT was designed as an IP address conservation mechanism — not as a security boundary. The "trusted neighbor" assumption made sense in the 1990s when a shared network meant one household or one company department. It does not hold in 2026 public infrastructure.
Microsoft and the Linux community have issued or are preparing patches for the specific CVEs, but fully eliminating the flaw without redesigning NAT itself is not possible. That is why Stagg's core recommendation was not "update your software" but "encrypt your traffic."
How to Stay Protected
Stagg stated plainly: "Encryption is a great help here — an attacker can still hijack a connection, but if they do, they can't send or receive any data over that connection unencrypted."
Practical steps for everyday users:
- Apply security updates. Patches for CVE-2026-56181 and CVE-2026-63913 reduce risk for specific scenarios — install them as soon as they become available for your platform.
- Use HTTPS everywhere. Most modern websites encrypt by default. Ensure your browser always connects over HTTPS, not plain HTTP.
- Encrypt all traffic with a VPN. LiMP VPN creates an encrypted tunnel for every byte your device sends. Even if an attacker hijacks a connection via NatJack, they receive only ciphertext — the content is unreadable, and DNS queries are protected inside the encrypted tunnel.
- Network admins: segment untrusted traffic. Place untrusted workloads on isolated NAT gateways, enable IP Source Guard where supported, and monitor NAT table saturation for anomalies.