Security protection measures for audit information systems using artificial intelligence

An audit information system is an important tool for enterprises and organizations to manage financial information and conduct financial audits. The existing security measures are difficult to effectively adapt to the dynamic and complex operating environment of the system, and there are limitations in risk detection. To enhance the reliability of the system and promote the smooth implementation of audit work, this paper studies the application of AI (artificial intelligence) in the security protection measures of audit information systems. Based on analyzing the system workflow and potential risks, an LSTM (Long and Short-Term Memory) security protection model was constructed. Real-time collection of various types of records in system components can be achieved by configuring a log collector, extracting semantic information using an abstract syntax tree and serializing it, converting text into vectors using a Skip-gram model, and conducting temporal dependency analysis on time series datasets. In the experimental study of detection rate, compared with the static code analysis model, the LSTM model in this paper has increased the average detection rate of abnormal events in user behavior, system events, and network traffic logs by 6.0%, 4.3%, and 5.3%, respectively. The conclusion indicates that the LSTM model under AI technology is helpful for real-time and accurate detection of potential risks in the system, providing reliable support for system security protection.