A Pyramidal Multistage Energy-Valley Attention Network for DDoS Attack Detection in Wireless Sensor Networks

Wireless Sensor Networks (WSNs) are seriously threatened by the growing frequency of Distributed Denial of Service (DDoS) attacks, which can result in major disruptions, monetary losses, and security breaches. These attacks abuse WSN protocol weakness and overload target systems with huge traffics and immediate and accurate detection is difficult. Conventional detection approach can barely deal with this kind of attacks due to the dynamism of the attacks where different attack types including DNS, SNMP, and NTP have different characteristics. This makes it mandatory to develop adaptative and high-performance detection systems to prevent and cope with DDoS attacks in WSNs, as well as providing reliable and real security in highly dynamised and restricted resources contexts. To overcome these issues, a novel DDoS Attack Detection using Pyramidal Scaled Multiscale Energy Valley Dot-Product Convolutional Attention Network in WSN is proposed. This research employs the cicaddos2019 dataset that contains over 50 million network records to solve the problem of accurately distinguishing the various DDoS attacks in the 12 classes. Among the pre-processing decisions, data cleaning, encoding, discretization and balancing are employed as part of preprocessing. Feature selection is performed using the Snow Geese Algorithm (SGA), which optimizes the choice of relevant features to enhance model efficiency. For attack detection, the Pyramidal Scaled Multiscale Energy Valley Dot-Product Convolutional Attention Network (PSMEV-SPCAN) is employed, leveraging advanced attention mechanisms to capture intricate patterns and dynamics in network flows. The model demonstrated an experimental accuracy of 99.9%, offering a high-performance solution for real-time DDoS detection in WSN. The first benefit found in this approach is that it can be implemented on the large and scalable networks with the ability to adapt depending on the nature of the network environment and with the capability to offer the broad and full spectrum of security against the constantly emerging new threats in the cyberspace.