符号执行是保障软件质量与可信性的重要方法之一。近年来，由于相关算法、约束求解技术以及计算能力的发展，符号执行方法取得了长足的发展。然而，路径爆炸与约束求解导致的可扩展性问题仍然是阻碍符号执行得到进一步大规模应用的主要因素。与此同时，人工智能技术的发展为可信软件领域带来了新的机遇。项目将约束求解、深度学习、强化学习等多种人工智能方法应用于符号执行的不同层次，旨在提高符号执行的可扩展性。.研究内容包括：（1）基础支持层：符号执行的约束求解增强方法，挖掘约束求解器的能力，以缓解约束求解开销与路径爆炸问题；（2）路径搜索策略层：面向程序热点的智能符号执行方法，使用深度学习方法制定高效的路径搜索策略，以提高符号执行发现程序缺陷的效率；（3）技术融合层：模糊测试与符号执行方法的结合策略研究，探明两种技术的最佳融合方法，以提高程序分析的效率。项目成果将对可信软件领域的发展起到重要的推动作用。

Symbolic execution is one of the important techniques in the area of trustworthy software. Recently, symbolic execution has experienced great development due to the impressive progress of constraint solving, related algorithms and computation power. However, the scalability caused by path explosion and constraint solving overhead is still impeding the application of symbolic execution in the industry. Meanwhile, artificial intelligence is bringing new opportunities for trustworthy software area. This project applies multiple artificial intelligence techniques including constraint solving, deep learning and reinforcement learning, etc. to different layers of symbolic execution, aiming to tackle the scalability problem. .The main content of this project includes three aspects as follows. (1) constraint solving augmented symbolic execution releases the power of constraint solver, aiming to reduce the constraint solving overhead. (2) Program hotspots oriented intelligent symbolic execution improves the efficiency of bug detection, aiming to tackle the path explosion problem. (3) The optimal strategy for fusing fuzzing and symbolic execution aims to accelerate the testing and analysis of programs. The project is expected to advance the development of trustworthy software area.