Hybrid Convolutional Neural Networks To Create An Attack Detection Framework For A Wireless Sensor Network Based Health Care Application

Healthcare Wireless Sensor Networks (H-WSNs) play a crucial role in modern healthcare applications, enabling remote patient monitoring and data collection. A unique Hybrid Deep Learning Model (HDML) for attack detection in H-WSNs is presented in this paper. Convolutional neural networks (CNNs) and long short-term memory (LSTM) networks are combined to create HDML, which enables the efficient detection of common, black, and grey hole attacks. The proposed approach comprises three phases: pre-processing, dimensionality reduction, and attack detection. For the purpose of facilitating subsequent analysis, patient health data is pre-processed using the one-hot encoding technique. We use the Modified Huber Independent Component Analysis based Squirrel Search Algorithm (MHICA-SSA) for dimensionality reduction to successfully reduce data size without compromising important information, overcoming the shortcomings of H-WSNs. The heart of the attack detection system is the CNN-LSTM model, which leverages the spatial feature extraction capabilities of CNNs and the temporal dependencies captured by LSTM networks. This powerful combination allows for accurate and efficient identification of malicious activities within the H-WSN. In order to verify the efficacy of HDML, we implement the suggested framework in the NS2 network simulator and assess its performance using a variety of performance metrics, such as packet delivery ratio, false alarm rate, network lifespan, energy consumption, throughput, and attack detection time and rate. This research contributes to enhancing the security and performance of H-WSNs, making them more robust and reliable for healthcare applications. In comparison to current methods, the suggested methodology increases throughput by up to 3.3 times and network lifetime by up to 4 times.