GeoRedact: towards automated redaction of privacy-sensitive geo-spatial information in full-motion videos

Dharar, Samir (2024) GeoRedact: towards automated redaction of privacy-sensitive geo-spatial information in full-motion videos. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (44MB)

Abstract

In recent years, the defense sector has seen significant advancements with the integration of drones that capture full-motion video (FMV) along with geospatial metadata. Byte-level analysis of these videos can disclose confidential mission information, highlighting the importance of protecting sensitive data from unauthorized access. While current redaction techniques often focus on visual elements, such as faces and license plates, the redaction of geospatial metadata has received less attention. This dissertation presents a systematic investigation into FMV redaction by introducing tools for metadata inspection and transformation, a novel approach for redacting geospatial metadata, and a new evaluation method, the Privacy-Utility Redaction Score (PURS), for assessing object detection models. The container parsers developed can efficiently extract and narrow down bytes of interest with minimal memory and CPU usage, and crucially, without memory leaks, making them well-suited for security-focused applications. The metadata redaction module enables selective redaction of user-specified metadata elements in Motion Imagery Standards Board (MISB)-compliant FMV. Additionally, the Privacy-Utility Redaction Score provides a metric for evaluating models based on their ability to redact private objects while ensuring public objects remain visible.

Actions (login required)

View Item