SecureCoder
Security-Aware Code Generation and Evaluation
Large Language Models are reshaping software development, but their ability to generate secure code remains an open challenge. SecureCoder is a collaborative research initiative at TU Darmstadt that tackles this problem end-to-end: from curating security-focused training data and building robust evaluation benchmarks, to training security-aware models and deploying constraint-based decoding techniques that prevent vulnerabilities at generation time.
Our work brings together expertise in programming languages, natural language processing, and machine learning to produce developer-friendly tools, including an IDE plugin and a unified benchmarking CLI, that raise the bar for safe AI-assisted development.
