Enhancing impersonation fraud detection on smart buildings physical access control systems: an anomaly-based approach using social groups trajectories data.

• Ano de defesa: 2022
• Autor(a) principal: Silva, Gabriel Mariano de Castro
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Anomaly detection; Controle de acesso físico; Data mining; Detecção de anomalias; Detecção de grupos sociais; Edifícios inteligentes; Fraude de identidade; Impersonation fraud; Mineração de dados; Mineração de padrões de trajetória; Physical access control; Smart buildings; Social groups detection; Trajectory pattern mining
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/3/3141/tde-24022023-082742/

Anomaly-based impersonation detection consists of constructing profiles based on users frequent behaviors and comparing them with new data. The underlying idea is that a diverse behavior may indicate possible fraud, i.e., someone trying to impersonate the user. Most research in the area aims to use spatio-temporal data broadly available from ubiquitous location sensors, like GPS, mobile telephony, beacons, and physical access control systems. On the other hand, many studies achieved good performance in finding social bonds among users. In the present work, we combined concepts from previous research and proposed a new model of profiles called Group-T-Patterns, originally published in (SILVA; SICHMAN, 2022), that uses social groups to construct mobility profiles and enhance anomaly detection. In particular, we developed an algorithm to mine Group-T-Patterns named GTPM (Group Trajectory Pattern Mining) and implemented a fully functional impersonation fraud detector for physical access control systems. We conducted an empirical analysis using data from two real-world datasets, and the results show that adding companion activities information to mobility profiles enhances anomaly-based impersonation attack detection.