labtrace reconstructs a forensic timeline from raw Linux audit telemetry and correlates events across processes. Rules find the signals, taint analysis connects them, and you get ranked incident reports with the full process chain — with no agent installed on the host.
SIEMs show you events. Auditors need incidents. The distance between “500 suspicious log lines” and “this SSH session stole cloud credentials, shipped them to a raw IP, installed persistence and created a backdoor account” is manual, slow, and exactly what an incident responder has no time for. labtrace closes that gap in two passes.
LT001–LT010 mark single events on their own: reverse-shell primitives, key material access, cloud credential stores, persistence writes, log truncation, database dumps and sensitive archives.
Sensitive data (SSH keys, cloud tokens, cookies, DB dumps) is marked at the source and propagated through file writes, exec arguments and child processes. A finding is promoted to an incident when taint reaches a sink: an outbound connection or an archive.
Findings that share a process chain collapse into one incident. A chain confirmed by a taint flow gets its severity raised. Lone low-severity noise is dropped instead of reported.
It is a static Rust binary that consumes an event stream; there is nothing to install on the machines you analyse.
cargo install --path .docker build -t labtrace .Usage and the full rule catalogue live in the repository README.
The rules are public and versioned, so an incident report can be checked against the exact logic that produced it.