Automated Anonymization of Video Data
Md Arifur Rahman
The Open Science movement encourages scientists to share their research data publicly, thereby enabling data reuse, facilitating research reproducibility, and ultimately improving research transparency. However, the consideration of ethical guidelines and legal requirements is crucial before publishing research data. Privacy is a key aspect and may require obfuscation or deletion of personal data unless informed consent has been obtained. This aspect is particularly important when dealing with video recordings of public places, which may inadvertently contain personal data of pedestrians or road users. However, manual anonymization is error-prone and time-consuming.
The objective of this thesis project is to develop an automated approach for anonymizing video data within the context of research data management. First, a suitable motivation scenario highlighting existing challenges associated with sharing research data has to be outlined. Particular emphasis should be given to ethical considerations and legal requirements. A requirements analysis has to be conducted, relevant requirements must be identified and described. A literature review has to be conducted, existing approaches and strategies for anonymizing video data must be classified and assessed. Based on these findings, a concept for an automated anonymization approach has to be developed. The feasibility of the concept must be demonstrated through a proof-of-concept implementation. Finally, the proposed approach must be evaluated against the specified requirements.