Network Security and Risk Assessment of Medical Information Systems Based on Artificial Intelligence

Abstract

Zero-day attacks exploit vulnerabilities that have not yet been made public or fixed to launch network attacks. Traditional medical information systems rely on signature libraries and rules of known vulnerabilities and cannot identify them, leading to slow system response and difficulty in preventing such attacks. Therefore, this paper uses artificial intelligence technology to detect zero-day attacks in real-time to improve medical information systems’ real-time detection and defense capabilities. First, the convolutional neural network (CNN) is used to extract the potential features of zero-day attacks from network traffic and system logs to identify unknown attack patterns automatically. Then, based on the support vector machine (SVM), the system’s normal behavior baseline is established, and abnormal activities in the system are quickly identified through real-time monitoring and anomaly detection. In addition, the reinforcement learning algorithm is used to design an intelligent defense system to automatically adjust the protection strategy according to the real-time security situation, including dynamically updating firewall rules and adjusting access control strategies. A multi-level artificial intelligence protection architecture is adopted, including comprehensive protection of the network layer, application layer, and data layer, to improve the system’s all-round defense capabilities against zero-day attacks. Experiments show that the average recognition rate of zero-day attacks reaches 94%; the average response time is only 3.15 seconds; the defense coverage rate in the medical information system reaches 90.5%, effectively improving the security of the medical information system.