A Combined Scheme Based on Artificial Immune System for Selective Forwarding Attack Detection in Event Driven Wireless Sensor Networks

Abstract

In general, the wireless sensor networks (WSNs) can be easily targeted by attacks due to their open nature. Among these attacks, the selective forwarding attack is particularly challenging to detect within WSNs. In this type of attack, malicious nodes deliberately discard received data packets, making it difficult to identify such attacks. Existing methods for selective forwarding attacks often suffer from low accuracy or high algorithm complexity, when dealing with Distributed Denial of Service attacks (DDoS). We propose an artificial immune system based on the danger model to detect network attacks. Our approach combines variational mode decomposition (VMD) and LSTM to identify selective forwarding attacks among other DoS attacks. We determine an optimal danger threshold and compare the outcomes to confirm the presence of attacks. This approach improves detection accuracy while minimizing computation requirements. The simulation results demonstrate that our proposed method achieves a low missing detection rate (MDR) of 0.6% and maintains a false detection rate (FDR) below 3.3%. Furthermore, in comparison with previous works, our system produces low algorithm complexity, making it more efficient in practical implementations.

Keywords

• clustering
• LSTM
• selective forwarding
• variational mode decomposition (VMD)
• wireless sensor networks