Microservice Security Detection via Metric Sensitivity

The patent replaces traditional rule-based and signature-based detection mechanisms with a machine learning-based anomaly detection system. The system uses unsupervised learning algorithms to analyze resource utilization metrics and behavioral patterns, substituting mechanical detection rules with adaptive statistical models that can identify novel attack patterns without requiring predefined execution paths.

The patent transforms the detection approach by changing from analyzing code execution paths to analyzing resource utilization parameters. It monitors metrics such as CPU usage, memory consumption, network traffic, and disk I/O, converting the detection problem from code-flow analysis to statistical parameter analysis, thereby overcoming the limitation of limited code path variation.

The patent implements preliminary action by continuously collecting and pre-processing resource utilization metrics in the background before attacks occur. It maintains baseline behavioral profiles and pre-computes statistical parameters, so that when an attack occurs, the system can perform rapid anomaly detection by comparing current metrics against pre-established patterns, reducing real-time computational burden.

The system employs self-service through automated baseline learning and adaptive threshold adjustment. The machine learning models automatically update their parameters based on observed normal behavior, eliminating the need for manual configuration and reducing ongoing computational complexity. The system serves itself by continuously refining its detection capabilities without external intervention.