Automated Vulnerability Detection in Source Code Using Quantum Natural Language Processing

One of the most important challenges in field software code audit is presence vulnerabilities source code. Every year, more and flaws are found, either internally proprietary or revealed publicly. These highly likely exploited lead to system compromise, data leakage, denial service. C C++ open-source codes now available order create a large-scale, classical machine-learning quantum for function-level vulnerability identification. We assembled sizable dataset millions functions that point potential exploits. created an efficient scalable detection method based on deep neural network model– Long Short-Term Memory (LSTM), machine learning (QLSTM), can learn features extracted from codes. The first converted into minimal intermediate representation remove pointless components shorten dependency. Previous studies lack analyzing causes models recognize real-life examples. Therefore, keep semantic syntactic information using state-of-the-art word embedding algorithms such as Glove fastText. embedded vectors subsequently fed convolutional networks classify possible vulnerabilities. To measure performance, we used evaluation metrics F1 score, precision, recall, accuracy, total execution time. made comparison between results derived LSTM basic feature well representation. found QLSTM with detects significantly accurate runs faster than its counterpart.