Research on Multidimensional Anomalous Behavior Baseline Characterization Technology Based on Deep Learning

In response to the stealth and difficulty in detecting the characteristics of Advanced Persistent Threat (APT) attacks, as well as the blind spots in internal network supervision, this paper proposes a multidimensional anomalous behavior baseline characterization technique based on deep learning. This technique integrates the frequency, time, business, and other multidimensional attributes of terminal behavior and network behavior, achieving comprehensive profiling and adaptive correlation analysis, and constructing a user network behavior baseline security model. By comparing the differences between current behavior and the baseline model, the model can assist security management personnel in prompt detection of abnormal behaviors that do not meet the security baseline requirements, thereby strengthening the security management personnel's capability to proactively identify and swiftly respond to emerging threats.