Efficient intrusion detection model based on health care sector using lstm based rnn

Abstract

Almost all real-world operations have moved online in recent years, with computers interacting with one another over the Internet. Consequently, there is an increase in network security vulnerabilities, making it difficult for network managers to protect their networks against all types of cyberattacks. Numerous methods for detecting network intrusions have also been created. However, they face critical difficulties from the continuous rise of new weaknesses that are outside the ability to understand of existing frameworks. We present an astute and effective Profound Learning (DL)- based network interruption discovery framework (NIDS),motivated by deep learning's outstanding performance in a variety of detection and identification tasks. We investigate an RNN-based prediction model for the detection of intrusions in industrial IoT networks. For intrusion detection, we use anomaly detection algorithms to identify if a packet is normal or abnormal. These methods quantify and assess the distance measurement in actual packets, as well as predict the following packet. The cyber security community has access to a wide range of malware datasets for use in public domain research. Furthermore, to the best of our knowledge, no study has offered a thorough evaluation of how well different machine learning techniques perform across a range of publicly accessible datasets. In this paper, we investigate novel hybrid deep learning model, with the aim of building an adaptable and efficient intrusion detection system that can identify and categorise unexpected and cyber-attacks. The results of this type of research make it easier to select the optimal algorithm for use in anticipating and stopping impending cyberattacks. Finally, to perform anomaly identification, a cosine similarity boundary that is thought of as a typical packet was provided. Then, a scoring function based on cosine similarity was applied.