Proj CDeepFuzz Paper Reading: Towards robustness of deep program processing models—detection, estimation, and enhancement
Abstract
背景: the robustness requires the model to produce consistent decisions given minorly perturbed code inputs
本文:CARROT
Github: https://github.com/SEKE-Adversary/CARROT
Task robustness detection, measurement, enhancement of DL models for source code processing
Method:
- CARROTA: optimization-based attack: generate code examples
- CARROTM: robustness metrics and measurement toolkit: 在允许的扰动范围内使用worst-case performance models
- CARROTT: improve the robustness of DL
实验:
subtasks: functionality classification, code clone detection, defect prediction
数据集:GRU, LSTM, ASTNN, LSCNN, TBCNN, CodeBERT, CDLH
效果:
- effective and efficient adversarial example detection
- tight robustness estimation
- effective robustness enhancement