Bottom line: The NatJack attack, presented at Black Hat USA 2026, shows that Network Address Translation is vulnerable to connection hijacking, DNS poisoning and denial-of-service across Windows, Linux and macOS, with Layer 2 protections such as VLANs offering no defense.
At Black Hat USA 2026, independent researcher Malcolm Stagg (Synack Red Team) presented NatJack, an attack class that manipulates the connection-tracking table of Network Address Translation (NAT). All 32 tested products and configurations from 13 notified vendors were affected.
NAT has been considered the standard mechanism since the early 1990s for working around IPv4 address scarcity by letting multiple devices share a single public IP address. According to Stagg, the underlying assumption that private addresses remain private and that peers within the same NAT boundary are trustworthy no longer holds up to closer scrutiny. Across 95 test reports covering 32 products and configurations, every implementation examined was vulnerable to at least one of the NatJack techniques, and in some cases to all four — confirmed on Windows, Linux and macOS, even though these systems do not share a common NAT codebase. According to Stagg, this points to a shared design assumption rather than an isolated bug, presumably caused by underspecification in the relevant RFCs.
An attacker who shares the NAT boundary with a victim can hijack active connections, poison DNS responses and force denial-of-service conditions — without IP spoofing or access to the broadcast domain, as older Layer 2 attacks required. The technique attacks at Layer 3 and Layer 4, which means VLAN segmentation and switch port isolation provide no protection. In detail, NatJack comprises four techniques: TCP Connection Hijacking (taking over a connection after forcing it into a closed state using spoofed packets, accelerated by an RFC 1337 TIME-WAIT assassination mechanism), DNS Response Poisoning (manipulating UDP DNS responses that pass through the NAT), Denial of Service (exhausting the NAT table, causing an outage for all devices behind the same NAT), and Connection Port Identification (determining the port number assigned to an active connection process as the basis for the other three techniques).
Earlier NAT attacks such as Samy Kamkar’s NAT Pinning (DEF CON 18 / Black Hat 2010) and NAT Slipstreaming (2020, extended in 2021 with Armis) relied on manipulating Application Level Gateways and required victim action, such as visiting a crafted website. Those vulnerabilities have since been patched. NatJack, by contrast, requires neither ALG abuse nor any victim action beyond an existing active connection through the same NAT.
Coordinated disclosure proceeded unevenly: while some vendors delivered formal patches, the Linux kernel security team initially rejected the report, reportedly calling it “totally bogus” according to Stagg. For CISOs, this means that network segmentation at Layer 2 alone is not sufficient to secure shared NAT infrastructure, and that connection-tracking security at Layer 3/4 as well as DNS integrity behind NAT gateways should be reviewed, particularly in environments with shared NAT boundaries between peers that are not fully trusted.
Source: www.csoonline.com · Published August 6, 2026
Lumi AI News — AI-assisted curation in accordance with Art. 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.8.3.