Search for: All records

Ensuring fairness in anomaly detection models has received much attention recently as many anomaly detection applications involve human beings. However, existing fair anomaly detection approaches mainly focus on association-based fairness notions. In this work, we target counterfactual fairness, which is a prevalent causation-based fairness notion. The goal of counterfactually fair anomaly detection is to ensure that the detection outcome of an individual in the factual world is the same as that in the counterfactual world where the individual had belonged to a different group. To this end, we propose a counterfactually fair anomaly detection (CFAD) framework which consists of two phases, counterfactual data generation and fair anomaly detection. Experimental results on a synthetic dataset and two real datasets show that CFAD can effectively detect anomalies as well as ensure counterfactual fairness. 

Malicious insiders cause significant loss to organizations. Due to an extremely small number of malicious activities from insiders, insider threat is hard to detect. In this article, we present a Dirichlet Marked Hawkes Process (DMHP) to detect malicious activities from insiders in real-time. DMHP combines the Dirichlet process and marked Hawkes processes to model the sequence of user activities. The Dirichlet process is capable of detecting unbounded user modes (patterns) of infinite user activities, while, for each detected user mode, one set of marked Hawkes processes is adopted to model user activities from time and activity type (e.g., WWW visit or send email) information so that different user modes are modeled by different sets of marked Hawkes processes. To achieve real-time malicious insider activity detection, the likelihood of the most recent activity calculated by DMHP is adopted as a score to measure the maliciousness of the activity. Since the majority of user activities are benign, those activities with low likelihoods are labeled as malicious activities. Experimental results on two datasets show the effectiveness of DMHP.