Anomaly Detection in Networked Bandits

2025-08-27

AI Summary

This work addresses the joint challenge of anomaly detection and user preference learning in social networks. Methodologically, we propose the first online learning framework that unifies anomaly detection with networked bandits: a collaborative contextual bandit model is developed, incorporating network regularization to capture user preferences, while anomalies are identified via residual-based modeling of deviations from structural consistency within the network. Theoretically, we establish a rigorous upper bound on the cumulative regret. Empirically, extensive experiments on both synthetic and real-world datasets demonstrate that our approach significantly outperforms state-of-the-art methods in both recommendation accuracy and anomaly detection rate. The core contribution lies in the novel integration of structure-aware anomaly detection into the networked bandit paradigm, enabling simultaneous optimization of personalized recommendation and robust anomaly discovery.

Technology Category

• Machine Learning: Online Learning & Bandits
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Economics, Online Markets and Human Computation: Social networks and social learning
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

The nodes' interconnections on a social network often reflect their dependencies and information-sharing behaviors. Nevertheless, abnormal nodes, which significantly deviate from most of the network concerning patterns or behaviors, can lead to grave consequences. Therefore, it is imperative to design efficient online learning algorithms that robustly learn users' preferences while simultaneously detecting anomalies. We introduce a novel bandit algorithm to address this problem. Through network knowledge, the method characterizes the users' preferences and residuals of feature information. By learning and analyzing these preferences and residuals, it develops a personalized recommendation strategy for each user and simultaneously detects anomalies. We rigorously prove an upper bound on the regret of the proposed algorithm and experimentally compare it with several state-of-the-art collaborative contextual bandit algorithms on both synthetic and real-world datasets.

Problem

• Detecting abnormal nodes in social networks with deviant behaviors
• Designing robust online learning algorithms for user preference learning
• Developing personalized recommendation strategies while identifying anomalies

Innovation

• Network-based bandit algorithm for anomaly detection
• Learning user preferences and feature residuals
• Personalized recommendation with simultaneous anomaly identification