AI-Augmented Adversary Lab

Bring your favorite assistant. Challenges simulate real investigations: logs, binaries and artifacts. Some are noisy, some incomplete, some misleading. Flags sit behind analysis, not single tricks. Large language models are encouraged and expected throughout.

Realistic artifacts: Production-style logs, binaries and evidence. Not textbook puzzles.
Ambiguous data: Noisy, incomplete, sometimes misleading. Like the real thing.
AI encouraged: Bring your assistant. The skill is knowing when to trust it.
Reasoning wins: Ask good questions, validate tool output, challenge automation.

Open to every summit attendee, any pass, any role.

ctf — adversary-lab — 80×24

you> triage auth.log. did anyone actually get in?

agent● reading auth.log … 48,213 lines

agent● 3 hosts show bursts of failures followed by a success

agent● 10.4.7.22 → 312 failures, then success as svc_backup at 02:14

agent● svc_backup is a service account. checking for interactive logins…

agent● none in 90 days. this one matters.

you> show me the raw lines. i don't trust you yet.

Feb 25 02:14:07 bastion sshd[8812]: Accepted password for svc_backup from 10.4.7.22

you> good. now: what did svc_backup do next?

APRES{████████████████} — flags unlock Feb 25, 10:00 AM MST

you>

Illustrative. The real challenges are harder, and the logs lie. When it finishes, the prompt is live: try hint.

Warm-up round · open now

There is a flag hidden on this website.

Four fragments, in the places a curious human or a well-behaved agent would look first. Join them and submit in the terminal. No prizes, just bragging rights.

or press ~

AI-Augmented Adversary Lab

There is a flag hidden on this website.

Sponsor the CTF