Li, Z.-J.; Li, T.; Chen, H.-D.; Yu, Q.; Qiao, M.-Q.; Li, L. Software Vulnerability Detection Method Based on Abstract Syntax Tree Feature Migration（AST-FMVD）. Preprints2023, 2023090374. https://doi.org/10.20944/preprints202309.0374.v1
APA Style
Li, Z. J., Li, T., Chen, H. D., Yu, Q., Qiao, M. Q., & Li, L. (2023). Software Vulnerability Detection Method Based on Abstract Syntax Tree Feature Migration（AST-FMVD）. Preprints. https://doi.org/10.20944/preprints202309.0374.v1
Chicago/Turabian Style
Li, Z., Meng-qing Qiao and Lin Li. 2023 "Software Vulnerability Detection Method Based on Abstract Syntax Tree Feature Migration（AST-FMVD）" Preprints. https://doi.org/10.20944/preprints202309.0374.v1
Abstract
In the broad context of vulnerability detection, deep learning has achieved considerable progress but faces generalization challenges in multilingual environments . We introduce a novel approach named AST-FMVD, which leverages transfer learning and abstract syntax trees. By employing semantic similarity clustering and context-aware technology, the method constructs node mapping relationships between different languages, enabling zero-shot learning in vulnerability detection. The method was validated by applying Java's vulnerability detection model in the Python domain, successfully demonstrating that AST-FMVD retains the original model's detection capabilities in the target domain. In conclusion, the proposed method offers a promising solution to the inherent problems in multi-language vulnerability detection, signifying a potential leap in the application of deep learning, transfer learning, and abstract syntax trees for improved cross-domain performance.
Keywords
deep learning; transfer learning; zero-shot; vulnerability detection; abstract syntax tree
Subject
Computer Science and Mathematics, Security Systems
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
