Enhancing Privacy in Image Search: Secure Object-Based Retrieval from Encrypted Images Using BFV Homomorphic Encryption

As the digital world becomes increasingly vulnerable to breaches, ensuring the security of images has grown in prominence due to rising privacy, data protection and security issues. The encryption of images, coupled with storage on servers or cloud platforms, provides an effective approach in safeguarding these images, facilitating their safe recovery in the future. This paper presents an innovative technique for obj ect-based image detection from encrypted images. The technique utilizes the Brakerski/Fan- Vercauteren homomorphic encryption system, allowing computations on encrypted image data without the necessity for decryption. This enables secure pattern matching, crucial for locating images containing the targeted object. Such a method is particularly essential in areas that uphold image privacy, including but not limited to autonomous cars, healthcare, finance, and law enforcement. The introduced technique works by encrypting the image, generating multiple encrypted object patterns, and employing homomorphic subtraction to confirm the presence of the object. Experimental results advocate for the feasibility of this method, displaying satisfactory accuracy and the assurance of maintaining privacy preservation.