AI brokers are more and more seen as a approach to reinforce the capabilities of cybersecurity groups — however which might do one of the best job? Wiz has developed a benchmark suite of 257 real-world challenges spanning 5 offensive domains: zero-day discovery, CVE (code vulnerability) detection, API security, net security, and cloud security to search out out.
Wiz checks totally different mixtures of AI brokers and their underlying AI fashions in opposition to the check suite to see which rating the very best in every of the 5 classes. Scoring is deterministic and programmatic utilizing a number of elements: multi-dimensional rubrics for zero-day and CVE detection; endpoint-and-severity matching for API security and lag seize for net and cloud challenges.
The benchmark checks run inside remoted Docker containers with ample sources and no per-challenge timeouts, so scores mirror functionality moderately than throttling. Every agent makes use of its native instruments and execution mannequin out of the field, and will get three goes at each problem to see the way it performs on common.
