AI Bug-Bounty Bot Misses SSRF Because It Checks HTTP, Not DNS

Summary

The creator says his AI bug-bounty bot missed an exploitable SSRF because it checked HTTP but not DNS callbacks. Manual monitoring and a second AI agent exposed internal data; he reports receiving a $5,000 bounty and shares testing-workflow improvements.

Key points

• The creator says his AI-driven bug-bounty bot flagged a webhook but treated repeated proxy errors as evidence it could not exploit the SSRF.
• The bot checked HTTP interactions but missed DNS callbacks; a DNS request to the creator’s monitoring server showed the target made an outbound lookup.
• Using a second AI agent, the creator says he reached internal JSON endpoints and retrieved usernames, emails, VPN profiles, and other personal data.
• He says reviewing the full JavaScript file exposed additional internal hosts, including access to Kubernetes and an actuator endpoint with environment variables and a heap dump.
• He reports receiving a $5,000 bounty for the finding; the target is not named in the transcript.
• His revised skill checks DNS and HTTP, maps filters and allowlists before bypass testing, records the finding’s source asset, and polls for delayed callbacks for up to 24 hours.