HARNESSING LLMS FOR AUTOMATING BOLA DETECTIONS

Summary from the TLDR InfoSec newsletter

Broken Object Level Authorization (BOLA) vulnerabilities are notoriously difficult to automatically detect as they do not follow set patterns and often require an understanding of the application logic and interdependencies. To address this problem, Unit 42 developed BOLABuster, which leverages LLMs for detecting BOLAs. BOLABuster takes an OpenAPI specification and then identifies all potentially vulnerable endpoints. It then establishes which endpoints are producers and which are consumers to map dependencies. The LLM determines the execution path to the potential BOLA and generates a test script. Unit 42 was able to find new vulnerabilities in Grafana, Harbor, and Easy!Appointments using BOLABuster.