Apex – AI-Powered Pentester Attacks Apps in Black-Box Mode to Find Vulnerabilities
Apex is an AI-driven penetration testing tool that operates autonomously in black-box mode on live applications. It functions without the need for source code, hints, or predefined attack paths, enabling the identification of vulnerabilities at a pace…
Apex is an AI-driven penetration testing tool that operates autonomously in black-box mode on live applications. It functions without the need for source code, hints, or predefined attack paths, enabling the identification of vulnerabilities at a pace suited to modern software development.
The development of Apex addresses the limitations of current software security practices. AI coding agents are increasingly producing and integrating code at a rapid rate, which traditional scanners and manual assessments cannot match. Apex serves as an adversarial verification mechanism, simulating real attacker behavior to identify vulnerabilities before they lead to breaches.
Continuous Integration (CI) Pipelines: Apex validates deployments against a sandboxed application replica, assessing the attack surface prior to code integration. Production Environments: It provides ongoing detection of exploitable vulnerabilities in real time. On-Demand Testing: This replaces periodic assessments with a continuous feedback loop adapted to the pace of modern threats.
PensarAI introduced Argus as a benchmark to assess the capabilities of offensive security agents, including Apex. Argus consists of 60 Dockerized web applications featuring various vulnerabilities.
Apex is an AI-driven penetration testing tool that operates autonomously in black-box mode on live applications.
Existing benchmarks like XBOW were found inadequate due to their limited scope. Argus covers a broad range of frameworks and introduces categories such as multi-step exploit chains and modern authentication bypasses.
Apex, evaluated using the Argus benchmark, achieved a 35% pass rate, surpassing other tools like PentestGPT and Raptor. In more challenging scenarios, Apex demonstrated higher effectiveness, solving 80% of the hardest challenges.
Overall, Apex identified 271 unique vulnerabilities, including SQL injection, SSRF, and more, with an average cost of $8 per challenge. Notable successes included complex multi-step exploits executed in under 15 minutes.
Apex and the Argus benchmark are available as open source on GitHub .
Based on reporting by Cyber Security News.
