Abstract
Software testing is an important method to ensure the functional delivery of software
products and detect problems early, which has attracted widespread attention from academia and
industry, and software automatic testing is one of the hotspots. Automated software testing
technology adopts methods such as program analysis, which can continuously generate test cases
and execute programs. Its main purpose is to cover as many program paths as possible and
discover vulnerabilities that can cause program crashes. On this basis, we propose a method that
can adaptively adjust based on environmental feedback in the context of enhanced learning, and
apply it to the design of search strategies for dynamic symbol execution. This method can
effectively solve control transfer problems and can be applied to hybrid testing algorithms. On
this basis, a new hybrid detection algorithm, LCCT, was designed based on reinforcement
learning.
Keywords: dynamic symbol execution; Mixed testing; Layered reinforcement learning
