Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fair Anomaly Detection For Imbalanced Groups

Ziwei Wu, Lecheng Zheng, Yuancheng Yu, Ruizhong Qiu, John Birge, Jingrui He

TL;DR

FairAD, a fairness-aware anomaly detection method targeting the imbalanced scenario, is proposed, which consists of a fairness-aware contrastive learning module and a rebalancing autoencoder module to ensure fairness and handle the imbalanced data issue.

Abstract

Anomaly detection (AD) has been widely studied for decades in many real-world applications, including fraud detection in finance, and intrusion detection for cybersecurity, etc. Due to the imbalanced nature between protected and unprotected groups and the imbalanced distributions of normal examples and anomalies, the learning objectives of most existing anomaly detection methods tend to solely concentrate on the dominating unprotected group. Thus, it has been recognized by many researchers about the significance of ensuring model fairness in anomaly detection. However, the existing fair anomaly detection methods tend to erroneously label most normal examples from the protected group as anomalies in the imbalanced scenario where the unprotected group is more abundant than the protected group. This phenomenon is caused by the improper design of learning objectives, which statistically focus on learning the frequent patterns (i.e., the unprotected group) while overlooking the under-represented patterns (i.e., the protected group). To address these issues, we propose FairAD, a fairness-aware anomaly detection method targeting the imbalanced scenario. It consists of a fairness-aware contrastive learning module and a rebalancing autoencoder module to ensure fairness and handle the imbalanced data issue, respectively. Moreover, we provide the theoretical analysis that shows our proposed contrastive learning regularization guarantees group fairness. Empirical studies demonstrate the effectiveness and efficiency of FairAD across multiple real-world datasets.

Fair Anomaly Detection For Imbalanced Groups

TL;DR

FairAD, a fairness-aware anomaly detection method targeting the imbalanced scenario, is proposed, which consists of a fairness-aware contrastive learning module and a rebalancing autoencoder module to ensure fairness and handle the imbalanced data issue.

Abstract

Anomaly detection (AD) has been widely studied for decades in many real-world applications, including fraud detection in finance, and intrusion detection for cybersecurity, etc. Due to the imbalanced nature between protected and unprotected groups and the imbalanced distributions of normal examples and anomalies, the learning objectives of most existing anomaly detection methods tend to solely concentrate on the dominating unprotected group. Thus, it has been recognized by many researchers about the significance of ensuring model fairness in anomaly detection. However, the existing fair anomaly detection methods tend to erroneously label most normal examples from the protected group as anomalies in the imbalanced scenario where the unprotected group is more abundant than the protected group. This phenomenon is caused by the improper design of learning objectives, which statistically focus on learning the frequent patterns (i.e., the unprotected group) while overlooking the under-represented patterns (i.e., the protected group). To address these issues, we propose FairAD, a fairness-aware anomaly detection method targeting the imbalanced scenario. It consists of a fairness-aware contrastive learning module and a rebalancing autoencoder module to ensure fairness and handle the imbalanced data issue, respectively. Moreover, we provide the theoretical analysis that shows our proposed contrastive learning regularization guarantees group fairness. Empirical studies demonstrate the effectiveness and efficiency of FairAD across multiple real-world datasets.
Paper Structure (27 sections, 6 theorems, 24 equations, 4 figures, 11 tables)

This paper contains 27 sections, 6 theorems, 24 equations, 4 figures, 11 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Proposed Method
  4. Fairness-aware Contrastive Learning
  5. Re-balancing Autoencoder
  6. Theoretical Analysis
  7. Fairness Bounds
  8. Experiments
  9. Experimental Setup
  10. Effectiveness and Efficiency of FairAD (RQ1)
  11. Data Imbalance Study (RQ2)
  12. Ablation Study (RQ3)
  13. Conclusion
  14. Notations
  15. Related Work
  16. ...and 12 more sections

Key Result

Lemma 3.1

Let $\mathcal{L}_0^t$ denote the loss of the unfitted model on the subgroup $t \in$ {UN, PN, UA, PA}, and let $\mathcal{L}_1^t$ denote the loss of the fitted model on the subgroup $t$, and $\Delta^t = \mathcal{L}_0^t - \mathcal{L}_1^t > 0$ means the difference of loss between the fitted model and th

Figures (4)

  • Figure 1: Recall@1200 and absolute Recall difference of the existing methods on MNIST-USPS dataset.
  • Figure 2: Illustrations of uniformity. The blue dots and green dots denote the normal examples from the unprotected group and protected group respectively. The red and pink dots denote the anomalies from the unprotected group and protected group respectively. In (a), many existing AD methods overly flag the examples from the protected groups as anomalies. In (b), traditional contrastive regularization does not consider group fairness. In (c), our method ensures group fairness while maintaining proper uniformity.
  • Figure 3: Ablation Study on MNIST-USPS dataset.
  • Figure 4: Ablation Study on Compas dataset.

Theorems & Definitions (10)

  • Lemma 3.1
  • Definition 4.1
  • Definition 4.2
  • Lemma 4.3
  • Lemma 4.4
  • Theorem 4.5
  • Theorem 4.6
  • Definition C.1
  • Lemma E.1
  • proof