Software Defect Detection Based on Feature Fusion and Alias Analysis
Xuejian Li, Zhengguang Zhu · 2023
As the scale of software systems continues to increase, predicting program defects in a quick and efficient manner has become a significant research area. Recent studies have introduced deep learning models that use neural networks to extract code features and build classifiers for defect prediction. However, most existing research focuses on extracting code features at a single granularity and single level, resulting in a lack of rich code features and low prediction accuracy. To address this issue, this paper proposes a defect prediction framework based on feature fusion and alias analysis for predicting the presence of non-inferable aliases and vulnerabilities in programs. The proposed approach parses the program into two different program representations, namely Abstract Syntax Tree (AST) and Program dependency Graph (PDG), and extracts code features for feature fusion using Long Short-Term Memory (LSTM) networks and Graph Convolutional Networks (GCN), respectively. To evaluate the effectiveness of the proposed approach, the Software Assurance Reference Dataset (SARD) from the National Institute of Standards and Technology (NIST) is chosen as the experimental dataset. The classifiers constructed using the proposed approach are used to predict the presence of non-inferable aliases and vulnerabilities in programs. The experimental results demonstrate the effectiveness of the proposed approach in predicting program defects.