XBOW, an autonomous penetration tester, has reached the top spot on HackerOne
10 months ago
- #AI
- #BugBounty
- #Cybersecurity
- XBOW, an autonomous AI-driven penetration tester, reached the top spot on the US HackerOne leaderboard.
- Initial testing involved CTF challenges and custom benchmarks before moving to real-world zero-day vulnerability discovery in open-source projects.
- XBOW operates without human input, scaling rapidly to complete penetration tests in hours.
- The transition to real-world environments was challenging due to the diversity of systems, from modern to legacy.
- XBOW participated in public and private bug bounty programs on HackerOne, treating it like any external researcher.
- Scaling involved strategic target prioritization using a scoring system based on various signals like WAF presence and HTTP status codes.
- Domain deduplication was essential to avoid redundant testing of similar environments.
- XBOW uses validators to ensure the accuracy of vulnerabilities, reducing false positives.
- Over time, XBOW reported thousands of validated vulnerabilities, many affecting high-profile targets.
- XBOW climbed to the top position on the HackerOne US leaderboard, submitting over 1,060 vulnerabilities.
- Vulnerabilities included Remote Code Execution, SQL Injection, XXE, SSRF, and more, with many still awaiting resolution.
- A notable discovery was a vulnerability in Palo Alto’s GlobalProtect VPN, affecting over 2,000 hosts.
- XBOW’s findings were confirmed by program owners, demonstrating its real-world impact and adaptability.