ProtectSIGN: Detection and Anonymization of Sensitive Non-Biometric Identifiers in Publicly Accessible Multimedia Documents

Introduction: In this Internet era, recent technological advancements have opened up vast opportunities for extracting sensitive information from online multimedia content, posing significant threats to user privacy. For instance, during the second quarter of 2022, cybercrime complaints saw a rise of 15.3%, totaling 237,658 complaints. Background: While the detection and anonymization of sensitive human signatures have become a critical social issue, existing literature mainly focuses on anonymization of biometric identifiers, neglecting non-biometric identifiers in the privacy preservation process. Methodology: The proposed framework-‘ProtectSIGN’, is the First of its kind, to address the detection and anonymization of non-biometric identifiers (human signatures) in authorized documents. Specifically, our framework employs Adaptive feature suppression strategies to optimize Scale Invariant Feature Transform (SIFT) feature extraction and uses an Anonymization algorithm to safeguard the anonymity of sensitive human signatures, which remains unexplored in literature. Results: Experimental evaluations on a real-world database of over 500+ government-issued passport documents demonstrate the effectiveness of this framework, in terms of various efficiency metrics. This framework can be employed to generate an anonymized database of signatures, which conceals sensitive information from unauthorized access, thereby ensuring secure handling of user data while maintaining privacy.