Attack-Resistant Uniform Fairness for Linear and Smooth Contextual Bandits

2026-02-04

AI Summary

This work addresses the vulnerability of contextual bandits to strategic manipulation that undermines fairness in deployment. It proposes the first robust algorithm that simultaneously guarantees $(1 - \widetilde{O}(1/T))$ uniform fairness and minimax-optimal regret under $C$-budget adversarial attacks. By introducing a corruption-adaptive exploration strategy and an error-compensation threshold mechanism, the method effectively balances the trade-off between fairness and safety for both linear and smooth reward functions. Theoretical analysis shows that the fairness cost becomes asymptotically negligible, while experiments demonstrate that the algorithm maintains high fairness and learning efficiency even under low-budget adversarial attacks in realistic settings. This study is the first to reveal the fragility of performance-based fairness mechanisms to subtle signal manipulation, thereby filling a critical theoretical gap in fair online learning under adversarial conditions.

Technology Category

• Machine Learning: Adversarial Learning & Robustness
• Game Theory and Economic Paradigms: Adversarial Learning
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Modern systems, such as digital platforms and service systems, increasingly rely on contextual bandits for online decision-making; however, their deployment can inadvertently create unfair exposure among arms, undermining long-term platform sustainability and supplier trust. This paper studies the contextual bandit problem under a uniform $(1-\delta)$-fairness constraint, and addresses its unique vulnerabilities to strategic manipulation. The fairness constraint ensures that preferential treatment is strictly justified by an arm's actual reward across all contexts and time horizons, using uniformity to prevent statistical loopholes. We develop novel algorithms that achieve (nearly) minimax-optimal regret for both linear and smooth reward functions, while maintaining strong $(1-\tilde{O}(1/T))$-fairness guarantees, and further characterize the theoretically inherent yet asymptotically marginal"price of fairness". However, we reveal that such merit-based fairness becomes uniquely susceptible to signal manipulation. We show that an adversary with a minimal $\tilde{O}(1)$ budget can not only degrade overall performance as in traditional attacks, but also selectively induce insidious fairness-specific failures while leaving conspicuous regret measures largely unaffected. To counter this, we design robust variants incorporating corruption-adaptive exploration and error-compensated thresholding. Our approach yields the first minimax-optimal regret bounds under $C$-budgeted attack while preserving $(1-\tilde{O}(1/T))$-fairness. Numerical experiments and a real-world case demonstrate that our algorithms sustain both fairness and efficiency.

Problem

• contextual bandits
• fairness
• strategic manipulation
• attack resistance
• uniform fairness

Innovation

• attack-resistant fairness
• uniform contextual bandits
• corruption-adaptive exploration
• minimax-optimal regret
• fairness-specific attacks