Discover Artificial Intelligence· 2026Q1
Application of dynamic symbolic execution and test coverage enhancement combined with artificial intelligence in large-scale web software systems
- 0citations
- Q1SCImago
- 2026year
Short summary
An AI-driven approach using Actor-Critic networks and curriculum learning significantly boosts path exploration (216-225 paths) and test coverage (73.4%-74.9%) in large-scale web systems, achieving 79.8% solution success and 88% vulnerability detection.
AI-generated from the title and abstract; the full text is not read.
Key points
- An Actor-Critic dual network model based on PPO optimizes path selection probability and state value evaluation in DSE.
- Curriculum learning and a generalized advantage estimation/pruning mechanism dynamically adjust path priority and depth.
- Key branch priority solving enhances constraint solving efficiency.
- Online model updates using coverage increment as reward dynamically refine path exploration strategies.
- Experiments show improved path exploration (216-225 paths), test coverage (73.4%-74.9%), 79.8% solution success, and 88% vulnerability detection.
AI-generated from the title and abstract; the full text is not read.
Abstract
Aiming at the bottleneck problems of path explosion and low efficiency of constraint solving faced by dynamic symbolic execution (DSE) in large-scale Web software systems, this paper applies a DSE and test coverage enhancement method combined with artificial intelligence (AI). Firstly, an Actor-Critic dual network model is constructed based on the PPO (Proximal Policy Optimization) algorithm. The Actor network outputs the probability distribution of path selection, and the Critic network evaluates the value of path state. Secondly, the path priority sorting strategy is updated through the generalized advantage estimation and pruning mechanism, and the path depth threshold is gradually increased by combining curriculum learning. The key branch priority solving strategy is applied to optimize the efficiency of constraint solving. Finally, after each round of symbolic execution, the coverage increment is used as a reward signal to update the model parameters online and dynamically optimize the path exploration strategy. Experimental results show that this method significantly improves the efficiency of path exploration (the number of paths reaches 225 and 216, respectively) and test coverage (74.9% and 73.4%) on large-scale Web systems, with an average solution success rate of 79.8% and an average vulnerability detection rate of 88%. It has good vulnerability detection capabilities and resource adaptability, and provides an efficient and scalable new paradigm for automated testing of complex systems.
The authors' abstract, as published at the source. Discover Artificial Intelligence, 2026 · DOI ↗
Continue with a free account
Ask the paper: 3 free questions a day about this paper; save it, get its citation, new summaries every day for your field. Takeaways are Premium.
Continue free on the webSign in with Google or Apple; no card needed. You come back to this paper.
On your phone:
Field: Software
SoftwareComputer Science