A method and system for automatically constructing an adaptive penetration testing attack path based on a large language model
The penetration testing system driven by a large language model collects and infers test status in real time and dynamically decides attack paths, solving the problems of rigid execution logic and invalid operations in traditional penetration testing, and realizing efficient and interpretable adaptive penetration testing.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Filing Date
- 2026-06-11
- Publication Date
- 2026-07-17
AI Technical Summary
Existing automated penetration testing technologies suffer from rigid execution logic, numerous invalid operations, lack of contextual semantic reasoning capabilities, and no failure path backtracking mechanism. They are unable to dynamically and adaptively construct attack paths, resulting in low testing efficiency, severe resource waste, and uninterpretable test results.
It adopts a large language model as the penetration testing decision engine, collects test status in real time, and dynamically decides the optimal attack action through semantic understanding and logical reasoning of the large language model, constructs an adaptive attack path, and is equipped with an intelligent backtracking mechanism for failed paths to generate an interpretable attack path chain.
Significantly improves the intelligence and efficiency of penetration testing, reduces invalid operations, adapts to diverse target assets, enables intelligent decision-making and path optimization throughout the entire process, and generates interpretable test reports.
Smart Images
Figure CN122419980A_ABST