Real-Time Detection of Data Exfiltration Using Deep Learning in Edge Computing Systems

Abstract

Data exfiltration remains a critical cybersecurity threat, particularly in edge computing environments where vast amounts of sensitive information are processed and transmitted. Traditional security mechanisms often struggle to detect sophisticated data breaches due to their reliance on predefined rules and signatures. This study proposes a deep learning-based approach for real-time detection of data exfiltration, leveraging transformer, CNN, and RNN architectures to analyze network traffic patterns and identify malicious activities. The transformer-based model demonstrated superior performance, achieving a detection accuracy of 96.3%, with lower false positive and false negative rates compared to CNN and RNN models. The proposed solution effectively minimizes alert fatigue by reducing false positives while ensuring high recall rates to detect unauthorized data transfers with minimal oversight. Additionally, the model's computational efficiency makes it well-suited for deployment in resource-constrained edge computing environments. Experimental results highlight the robustness of the approach against adversarial evasion techniques, emphasizing its potential for real-world cybersecurity applications. The study also explores the integration of continuous learning mechanisms and explainable AI to enhance model adaptability and interpretability. These findings suggest that deep learning-based detection methods can significantly improve data security in edge computing, providing a scalable and effective solution to mitigate data exfiltration threats in dynamic and distributed environments.