ML-CSS@ICL

MITRE reports an increase in the number of software CVEs submitted yearly since 2016, reflecting the increased threat to the overall security of the software ecosystem. Accordingly, there has been a steady growth of research in software vulnerability detection across a spectrum of approaches such as static analysis, dynamic analysis and machine learning-based detection models. On the other hand, the advancement of deep learning in natural language processing, computer vision and image processing has snowballed over the years. With the recent rise of neural representation architectures such as Transformers and BERT-based architectures, they achieved outstanding task-specific performance while topping multiple benchmarks. However, can these architectures be applied to the software security domain, specifically vulnerability detection? In this talk, we briefly explore the application of neural representation learning for programming languages and how to use it to detect software vulnerabilities.