Enhancing Wireless Communication Security using eBPF-Based Packet Filtering and GRU Models in Software-Defined Antenna Networks

Abstract

Modern wireless communication systems, particularly those involving smart antennas and software-defined radio networks, require efficient and secure data transmission under increasing cybersecurity threats. Traditional intrusion detection mechanisms often fail to scale with the dynamic demands of high-speed, low-latency wireless environments. This paper proposes an integrated framework that combines Extended Berkeley Packet Filter (eBPF)-based kernel-level packet telemetry with deep learning techniques, specifically Gated Recurrent Unit (GRU) models, for real-time anomaly detection in wireless communication systems. The approach leverages the CSE-CIC-IDS2018 dataset, processed through random sampling, Z-score normalization, and ANOVA-F feature selection, to train a GRU-based detection model capable of analyzing system-level metrics such as TCP session anomalies and malformed packet flows within wireless infrastructures. Packet capture and filtering are performed directly within the Linux kernel using eBPF, ensuring minimal latency. This mechanism is especially suited for software-defined antenna systems, where security must be maintained without compromising throughput. Experimental results show that the proposed system achieves an accuracy of 97.78%, demonstrating its viability for secure and adaptive intrusion detection in next-generation wireless networks