Optimized memory augmented graph neural network-based DoS attacks detection in wireless sensor network

Area of Science:

• Computer Science
• Network Security
• Artificial Intelligence

Background:

• Wireless Sensor Networks (WSNs) are crucial for remote data monitoring but vulnerable to sophisticated attacks.
• Traditional attack detection methods struggle with the increasing complexity and frequency of malicious activities in WSNs.

Purpose of the Study:

• To propose an Optimized Memory Augmented Graph Neural Network-based method for Denial of Service (DoS) attack detection in Wireless Sensor Networks (DoS-AD-MAGNN-WSN).
• To enhance the accuracy and effectiveness of detecting various DoS attack types, including blackhole, flooding, and grayhole attacks.

Main Methods:

• Utilized the WSN-DS dataset for data collection and pre-processing using a secure adaptive event-triggered filter.
• Implemented nested patch-based feature extraction for optimal feature identification.
• Employed a Memory Augmented Graph Neural Network (MAGNN) optimized by gradient-based optimizers for attack classification.

Main Results:

• The DoS-AD-MAGNN-WSN achieved superior accuracy in classifying DoS attacks compared to existing methods.
• Demonstrated significant accuracy improvements of 31.20%, 23.30%, and 26.43% over CNN-DoS-AD-WSN, TB-DoS-AD-WSN-RDT, and FBDR-DoS-AD-RM-WSN, respectively.

Conclusions:

• The proposed DoS-AD-MAGNN-WSN offers a more effective and accurate solution for detecting DoS attacks in Wireless Sensor Networks.
• The optimized MAGNN approach shows promise for improving the security and reliability of WSNs against advanced threats.