An immunity-based anomaly detection system with sensor agents

Area of Science:

• Computer Science
• Cybersecurity
• Artificial Intelligence

Background:

• Traditional anomaly detection systems often rely on single sensors, limiting their accuracy.
• User behavior analysis is crucial for identifying security threats and system anomalies.

Purpose of the Study:

• To propose an immunity-based anomaly detection system utilizing multiple, diverse sensor agents.
• To develop an evaluation framework for distinguishing internal and external anomalies.
• To improve the accuracy of anomaly detection in user command sequences on UNIX-like systems.

Main Methods:

• Developed an immunity-based system with specialized sensor agents mimicking immune system principles.
• Agents are designed for specificity to individual user behavior.
• Employed multiple diverse agents for collaborative anomaly detection.
• Proposed a novel evaluation framework to differentiate internal and external anomalies.

Main Results:

• The immunity-based system demonstrated superior detection accuracy compared to conventional methods.
• The multi-agent approach significantly improved anomaly detection performance.
• Experimental results validated the effectiveness of the proposed system on UNIX-like systems.

Conclusions:

• Immunity-based anomaly detection with multiple, diverse agents offers a more robust and accurate solution.
• The proposed system effectively detects anomalies in user command sequences.
• The developed evaluation framework provides valuable insights into anomaly types.