Research Article
A Novel Packet-Based Preprocessing Approach for Spiking Neural Networks Models to Enhance DDoS Detection Accuracy
@INPROCEEDINGS{10.1007/978-3-032-44094-5_15, author={Lan Le Duc and An Nguyen Vo Thien and Phu Nguyen Phan Hai and Trang Hoang}, title={A Novel Packet-Based Preprocessing Approach for Spiking Neural Networks Models to Enhance DDoS Detection Accuracy}, proceedings={Industrial Networks and Intelligent Systems. 12th EAI International Conference, INISCOM 2026, Da Nang, Vietnam, February 26--27, 2026, Proceedings}, proceedings_a={INISCOM}, year={2026}, month={11}, keywords={DDoS attack detection Spiking Neural Networks neuromorphic computing energy-efficient computing network intrusion detection IoT security temporal spike encoding}, doi={10.1007/978-3-032-44094-5_15} }- Lan Le Duc
An Nguyen Vo Thien
Phu Nguyen Phan Hai
Trang Hoang
Year: 2026
A Novel Packet-Based Preprocessing Approach for Spiking Neural Networks Models to Enhance DDoS Detection Accuracy
INISCOM
Springer
DOI: 10.1007/978-3-032-44094-5_15
Abstract
As Distributed Denial of Service (DDoS) attacks evolve in complexity, traditional Deep Learning approaches face limitations in computational efficiency and real-time processing. While Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks achieve high detection accuracy, they suffer from energy consumption and processing delays that hinder deployment in resource-constrained IoT and edge environments. This study proposes a novel Spiking Neural Network (SNN) architecture for DDoS detection that processes information through discrete spike trains, offering energy efficiency advantages while preserving temporal processing capabilities. We introduce a packet-based preprocessing methodology that transforms network traffic features into optimized spike representations, preserving critical temporal and flow-level information. Our experimental evaluation on CIC-IoT2023 datasets demonstrates that the SNN-based approach achieves superior detection accuracy compared to conventional methods while maintaining significantly lower computational overhead, making it suitable for real-time DDoS detection in distributed network environments.