Deep learning assisted program analysis

Deep learning is a powerful tool to overcome the awkward situation of extracting endless rules from real world. We propose some novel approach in program analysis. In DEEPVSA, we employ a layer of instruction embedding along with a bi-directional sequence-tosequence neural network to learn the machine code pattern pertaining to memory region accesses to help value-set analysis. In RENN, we designed a new recurrent neural architecture that could capture the data dependency between machine code segments for alias analysis.

People

Ligeng Chen

2017-2023(Now at HONOR)

Yi Qian

2021-now

Yuyang Wang

2021-now