egoDetect: Visual Detection and Exploration of Anomaly in Social Communication Network.

Area of Science:

• Computer Science
• Network Security
• Data Visualization

Background:

• The internet's growth necessitates robust social network anomaly detection to combat fraud and privacy issues.
• Existing anomaly detection methods lack accuracy and require labeled data, hindering direct application.
• Human analysts are essential for interpreting anomaly detection results, highlighting the need for effective tools.

Purpose of the Study:

• To propose egoDetect, a novel visualization system for efficient and objective anomaly detection in social communication networks.
• To enable experts to analyze and explore anomaly detection results more effectively.
• To address the limitations of existing methods by providing an unsupervised approach.

Main Methods:

• Developed egoDetect, a visualization system employing unsupervised anomaly detection.
• Designed a novel glyph-based egocentric network visualization to explore user topology and relationships.
• Integrated rich user interactions for efficient navigation and further exploration by experts.

Main Results:

• The egoDetect system efficiently detects anomalies in social communication networks.
• The unsupervised approach allows for anomaly detection without prior training.
• The egocentric network visualization and interactive features facilitate objective analysis.

Conclusions:

• The proposed egoDetect system is effective for anomaly detection in social networks.
• Visualization and interactive exploration enhance the analysis of network anomalies.
• The system supports human analysts in making informed decisions regarding network security.