JS engine difference fuzz testing method based on deep learning

• Main Authors: FAN ZHENYE, YE GUIXIN, TANG ZHANYONG, ZHAO JINGYING, CHE XIAOKANG, YI WEN
• Format: Patent
• Language: Chinese; English
• Published: 14.07.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention belongs to the field of software automation testing, and particularly relates to a JS engine difference fuzzy testing method based on deep learning, and the whole operation process sequentially comprises a data collection stage, a model construction stage, a use case generation stage and a difference testing stage. Firstly, model training data and a high-quality test suite are collected; carrying out data preprocessing and submitting the data to a model for fine tuning training; using the fine tuning model to continuously write the test suite; finally, the generated test cases are handed over to JS engines in the differential test system to be executed, suspicious cases obtained after execution are subjected to further result judgment through a corresponding execution platform of the standard version supported by the engines, and whether the results are JS engine implementation errors or not is analyzed. According to the method, the deep learning text generation model is used for generating the