Bayesian Selection Policies for Human-in-the-Loop Anomaly Detectors with Applications in Test Security

Michael Fauss, Xiang Liu, Chen Li, Ikkyu Choi, H Vincent Poor · 2025

Read Bayesian Selection Policies for Human-in-the-Loop Anomaly Detectors with Applications in Test Security by Michael Fauss, Xiang Liu, Chen Li, Ikkyu Choi, H Vincent Poor (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13122414.txt.

Description

Michael Fauss, Xiang Liu, Chen Li, Ikkyu Choi, and Vincent Poor published this 2025 study on choosing which test takers a human expert should review for cheating. Tests can flag odd answer patterns automatically, but expert time is scarce, so each administration can only scrutinize a manageable few. Reviews then teach the system, yet only the reviewed are ever labeled, a semi-supervised tangle.

The authors cast the choice in Bayesian terms and derive the optimal selection policy in principle. Because exact computation is infeasible, they offer a variational approximation plus three heuristic policies that trade exploration against exploitation in different proportions. Simulations and real testing data benchmark the approximations against standard algorithms and theoretical bounds. The aim is a flagging queue that improves itself, catching more true anomalies per expert hour.

Catalog record

CREATORMichael Fauss, Xiang Liu, Chen Li, Ikkyu Choi, H Vincent Poor
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE93 KB
VERSION1.0
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC13122414

Browse the full library · About the Tower of Babel