Secure Approximate Deduplication for Forensic Images in Crowdsensing Vehicular Networks

The convergence of mobile crowdsensing and the Internet of Vehicles (IoV) has advanced image forensics in intelligent transportation systems. However, the large influx of homogenized sensing images during evidence collection causes a massive waste of communication and storage, hindering forensic analysis. Additionally, due to privacy concerns, these images are often encrypted and uploaded as different ciphertexts, introducing new challenges for deduplication. Hence, securely eliminating these near-duplicate images is essential to improve forensic efficiency and privacy preservation in crowdsensing vehicular networks. In this paper, we propose a <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</u>ecure <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">A</u>pproximate <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Dedup</u>lication scheme (SA-Dedup) for forensic images in fog-assisted crowdsensing vehicular networks. In the scheme, we divide geospatial grid cells based on different positional perspectives to improve the deduplication efficiency while obscuring precise vehicular positions. In each grid cell, we design a novel encrypted approximate matching method to detect similar images based on Function Secret Sharing (FSS) and Boneh-Goh-Nissim-based Homomorphic Encryption (BGN-based HE). In the deduplication process, vehicles' real identities are concealed to enhance identity privacy and can be traceable in case of malicious behavior. Further, the designed incentive mechanism can encourage vehicles to provide valuable sensing images for forensic tasks. Finally, the theoretical analysis and the experiment results demonstrate that SA-Dedup can implement secure and effective near-deduplication of forensic images.