A Bag of Features and Blockchain-Based Efficient Cyber-Attack Detection in Healthcare Using Deep Convolutional Neural Networks

Abstract

This paper proposed novel model for detection of cyber-attacks in healthcare, merging a convolutional neural network (CNN) with a bag of features (BoF) to address challenges associated with low accuracy and class imbalance in existing cyber-attacks detection. Feature selection is accomplished through a bag of features, mitigating problems with feature duplication. A convolutional neural network is then used to extract spatial information, and average pooling and max pooling techniques are combined after that. By integrating attention processes, features are given different weights, which lowers overhead and improves model performance.Simultaneously, a bag of features is utilized to capture long-distance dependent information, facilitating comprehensive feature learning. Finally, classification is executed through a SoftMax function. Evaluation of Testing of the suggested attacks detection model is done using the CIC-IDS2017 and NSL-KDD datasets. Investigational outcomes demonstrate classification accuracies of 86.25%, 99.69%, and 99.65%, respectively, surpassing existing CNN models by margins of 1.95%, 0.47%, and 0.12%. The model effectively addresses issues of low accuracy and class imbalance, showcasing its robustness and efficacy in attacks detection tasks.