A recent Stanford University study reveals that an AI agent named ARTEMIS significantly outperformed human penetration testers in identifying network vulnerabilities, achieving an 82% valid submission rate during a 16-hour operation. Despite costing only $18 per hour, the AI still faced limitations, such as difficulty interacting with graphical user interfaces (GUIs), leading to missed vulnerabilities and increased false positives. Experts, including Diana Kelley from Noma Security, argue that while ARTEMIS showed promise in efficiency and cost reduction, human testers excel in creativity, contextual understanding, and nuanced risk assessment. The study concludes that AI, including ARTEMIS, will enhance human capabilities rather than replace them, emphasizing the importance of human oversight in penetration testing. As AI continues to evolve, researchers anticipate improvements that will address current shortcomings, reinforcing the critical role of human expertise in cybersecurity efforts while benefiting from automation’s efficiency.
Source link
Share
Read more