Detection and Prediction of Inter-Slice Handover DDoS Attacks in 5G and Beyond Networks Using P4 and Deep Learning

This study addresses the challenge of securing 5G and beyond (6G) networks against Distributed Denial of Service (DDoS) attacks during inter-slice handovers. A hybrid model based on P4 programmable switches and the Gated Recurrent Unit (GRU) algorithm is proposed to detect and predict such attacks with high accuracy and low latency. P4 enables real-time extraction of key Quality of Service (QoS) parameters, including packet loss rate, latency, and priority, which are used for efficient traffic analysis and attack detection. The proposed model achieves a DDoS detection accuracy of 98.63%, sensitivity of 98.53%, and an F1 score of 98.58%, while predicting legitimate slices with an accuracy of 98.7%. The false positive rate (FPR) is reduced to less than 2.1%, and the total system delay for detection and decision-making is kept below 350 milliseconds, making it suitable for latency-sensitive applications such as URLLC. Scalability tests demonstrate that the system maintains over 90% detection accuracy and a delay of less than 500 milliseconds with up to 15 switches and 4 slices, even under high traffic loads. This research highlights the effectiveness of combining deep learning with P4 for enhancing security and scalability in advanced networks, providing a robust framework for next-generation network security.