HYBRID DEEP LEARNING MODELS FOR SECURITY DEFECT DETECTION IN ENTERPRISE SOFTWARE SYSTEMS

Abstract

Enterprise software systems are increasingly exposed to security defects due to complex architectures, large-scale codebases, third-party dependencies, and rapidly evolving cyber threats. Traditional security defect detection methods based on static analysis and manually defined rules often experience limitations in detecting sophisticated vulnerabilities and adapting to new attack patterns. This paper proposes a Hybrid Deep Learning Framework for Security Defect Detection in Enterprise Software Systems by integrating deep learning, machine learning, static code analysis, software metrics, vulnerability intelligence, and automated security testing. The proposed framework combines multiple deep learning models to analyze source code structures, execution patterns, dependency relationships, and historical vulnerability information for accurate defect identification. Hybrid learning mechanisms improve feature extraction, classification accuracy, and detection reliability. The framework continuously evaluates enterprise software components and prioritizes security defects based on severity and potential impact. Experimental evaluation demonstrates improvements in defect detection accuracy, vulnerability classification, falsepositive reduction, and continuous security monitoring. The proposed approach provides an intelligent solution for secure enterprise software development.