Geometric Erudition Intended for Variance Recognition within Shade Server Organization: A Multi-Order Markov Series Structure

Abstract

To enhance the security of IT infrastructure, especially in cloud computing platforms that host entire applications and data, anomaly detection plays a critical role. This paper introduces a multi-order Markov chain-based framework for anomaly detection, expanding upon the traditional Markov chain model. Our approach incorporates both high-order Markov chains and multivariate time series, embedded within algorithms designed under a statistical learning framework. To address time and space complexity, our algorithms employ non-zero value tables and logarithmic values in initial and transition matrices. For validation purposes, we utilized system calls and their return values from the DARPA intrusion detection evaluation dataset to construct a two-dimensional test input set. The test results indicate that our multi-order approach produces more effective indicators. Both the absolute values derived from single-order models and the changes in rankings across different-order models are strongly correlated with abnormal behaviors, thereby enhancing the detection of anomalies.