New Lightweight Algorithm Enhances Vulnerability Detection in Python Code

• •AMFKD algorithm achieves high accuracy in detecting vulnerabilities in Python code.
• •The model reduces computational costs while maintaining performance compared to larger models.
• •Results were validated on two benchmark datasets, showing scores above 92%.

Researchers introduced a lightweight vulnerability detection algorithm named AMFKD, which utilizes adaptive multi-layered fusion knowledge distillation to enhance vulnerability detection in Python source code. This method addresses the challenges of manual vulnerability detection, which is time-consuming and requires extensive expertise, as well as the limitations of traditional deep learning models that are often too resource-intensive for deployment in constrained environments. The algorithm was evaluated on the VUDENC and CVEFixes datasets, achieving Macro-F1 scores of 92.25% and 93.76%, respectively, while consuming fewer computational resources than the larger teacher model. The study emphasizes the practical significance of deploying high-accuracy models in environments with limited memory and processing power. The findings suggest that AMFKD could significantly improve automated code inspection capabilities, although the results are based on specific datasets rather than broader production codebases.