Enhancing Code Security with CodeSheriff: A Multi-Language Vulnerability Scanner

Abstract Securing code has become a significant challenge in software development due to cyber threats. Identifying vulnerabilitiesduring development is crucial to reduce security risks. The solution, Codesheriff, is a Telegram bot that scans for vulnerabilitiesautomatically. Codesheriff uses a fastText model trained on the Rosetta Code dataset to detect programming languagesprecisely and employs customized models for detecting vulnerabilities in C, C++, Python, Java, PHP, and JavaScript. Itsperformance is measured against tools like Veracode and Fortify using various metrics. Experiments show Codesheriffachieved precision rates of 93.6% for Python, 93.5% for C/C++, 94.5% for Java, 91.8% for PHP, and 92.4% for JavaScript,demonstrating its robustness and practicality in secure code development.