Analysis of Privacy Infringement Mechanisms by Data Lifecycle in AI-based Investigative Systems and Design of Integrated Control Governance
TL;DR
This study analyzes the cascading risk mechanisms across the data lifecycle in core AI investigative systems like facial recognition and predictive policing and proposes an integrated governance framework based on ‘Security by Design’ principles, combining technical defenses with policy controls.
Abstract
While AI-based investigative technologies enhance policing efficiency, their inherent architectural characteristics—such as extreme data dependency, deep learning black-boxes, and continuous feedback loops—introduce severe structural privacy infringement. This study analyzes the cascading risk mechanisms across the data lifecycle(collection-storage-analysis-utilization) in core AI investigative systems like facial recognition and predictive policing. By comparing with global regulatory trends in the EU and the US, this study identifies critical governance gaps within Korea's existing static legal frameworks. To address these challenges, we propose an integrated governance framework based on ‘Security by Design’ principles, combining technical defenses with policy controls. Tailored countermeasures are mapped to each lifecycle stage: edge masking and dynamic PIA for collection; federated learning and fusion audits for storage; explainable AI and security certification for analysis; and immutable logs with Human-in-the-Loop mandates for utilization. This study contributes a practical control model to ensure both judicial reliability and privacy in intelligent policing infrastructures.