Adaptive malware identification via integrated SimCLR and GRU networks.

Area of Science:

• Computer Science
• Cybersecurity
• Machine Learning

Background:

• Malware increasingly evades traditional signature-based detection through obfuscation and dynamic behaviors.
• Existing methods exhibit limitations in identifying novel threats, creating vulnerabilities in digital infrastructures.
• The need for adaptive, real-time malware detection systems is critical for enterprise and institutional security.

Purpose of the Study:

• To develop an adaptive and efficient malware detection framework capable of real-time analysis.
• To address the limitations of conventional methods in detecting sophisticated malware.
• To enhance the accuracy and resilience of malware detection systems.

Main Methods:

• Introduced SimCLR-GRU, a novel ensemble architecture combining SimCLR for feature extraction and Gated Recurrent Unit (GRU) for sequential pattern analysis.
• Incorporated Graph Neural Network (GNN)-based feature selection to minimize redundancy.
• Utilized Fish School Search (FSS) for hyperparameter optimization to improve model performance.

Main Results:

• Achieved a classification accuracy of 99% on a Portable Executable (PE) malware dataset, surpassing baseline models by 15%.
• Demonstrated high generalizability and accuracy with an Area Under the Curve (AUC) of 98.2% and an F1-score of 96.8%.
• Reported a low false positive rate of 0.02% and low inference latency, suitable for real-time applications.

Conclusions:

• SimCLR-GRU offers a scalable and effective solution for modern, evolving malware detection challenges.
• The framework's performance highlights its potential for real-time and resource-constrained environments.
• The integration of contrastive learning, GRU, GNN, and FSS provides a robust approach to cybersecurity threats.