A trusting-trust backdoor built from GNU strip
A single tampered copy of GNU strip, placed in the NixOS binary seed, was enough to backdoor almost every binary in a complete graphical installer. The installer built from a real nixpkgs revision with no failures, and the payload rode from one generation of strip to the next until it reached the final standard environment, after the seed itself had already dropped out of the dependency closure.…
Julien Malka's research, "Trusting-Trust Attack against an Entire Linux Distribution through Binary Manipulation," revealed how a single compromised copy of GNU strip could infect an entire Linux distribution. Published in July 2026 and revised in September, Malka's paper shows that GNU strip, despite lacking the ability to read source code or generate binaries, can still carry a malicious payload through the distribution building process.
NixOS, which builds its distribution from a small binary seed and discards it before finalizing packages, is an example of a secure target. The tampered strip sits in this seed, processing binaries during the bootstrap and carrying the backdoor forward at each stage until it reaches the final standard environment. By the time the seed leaves the closure, the backdoor is already embedded in the standard environment, making it present in shipped packages.
The attack demonstrates that reproducible builds, which prove that published source and build inputs yield the published binary bit for bit, have limitations. If one of the build inputs is a poisoned strip in the seed, every independent rebuilder will reproduce the same backdoored binary, confirming the hash. However, reproducibility does not guarantee that the inputs were clean.
The trusted computing base of a build extends beyond the compiler to include every executable in the bootstrap seed. Therefore, auditing effort should focus on the entire seed, not just the compiler, as any tool handling binaries in the seed becomes part of the attack chain.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
