Symbolic execution is a powerful systematic software analysis technique, but suffers from the high cost of constraint solving, which is the key supporting technology that affects the effectiveness of symbolic execution. Techniques like Green and GreenTrie reuse constraint solutions to speed up constraint solving for symbolic execution; however, these reuse techniques require syntactic/semantic equivalence or implication relationship between constraints. This paper introduces DeepSover, a novel approach to constraint solving with deep learning for symbolic execution. Our key insight is to utilize the collective knowledge of a set of constraint solutions to train a deep neural network, which is then used to classify path conditions for their satisfiability during symbolic execution. Experimental evaluation shows DeepSolver is highly accurate in classifying path conditions, is more efficient than state-of-the-art constraint solving and constraint solution reuse techniques, and can well support symbolic execution tasks.

中文翻译：

---

使用深度学习进行符号执行的约束求解

符号执行是一种强大的系统软件分析技术，但约束求解成本高，是影响符号执行有效性的关键支撑技术。Green 和 GreenTrie 等技术重用约束解决方案来加速符号执行的约束求解；然而，这些重用技术需要约束之间的句法/语义等价或蕴涵关系。本文介绍了 DeepSover，这是一种通过深度学习进行符号执行的约束求解的新方法。我们的关键见解是利用一组约束解决方案的集体知识来训练深度神经网络，然后使用该网络对路径条件进行分类，以确保它们在符号执行期间的可满足性。