Data hiding in encrypted images based on predefined watermark embedding before encryption process

• We provide an approach of hiding a message into an encrypted image.
• We show that the embedded message can be read whether the image is encrypted or not.
• We present an implementation of our method based on LSB substitution scheme and RC4.
• We discuss the performances and the security of the proposed implementation.
• We precise how our approach can be used for verifying image reliability.

In this paper, we propose a novel approach which allows embedding a message into an encrypted image; a message that can be accessed whether the image is encrypted or not. To do so, the proposed solution relies on the insertion into the image of a predefined watermark, a “pre-watermark”, before the encryption process. Message insertion (resp. extraction) is then commonly conducted into (resp. from) the encrypted image. It is the impact of this data hiding process onto the “pre-watermark” that gives us access to the message into the spatial domain, i.e. after the decryption process. By doing so, the watermark processing process is independent of the knowledge of the encryption key and one only just has to know the watermarking key so as to embed the message and extract it from the encrypted or decrypted image. Reciprocally, encryption/decryption processes are completely independent from message embedding/extraction. We illustrate the feasibility of our approach considering the RC4 stream cipher algorithm and the Least Significant Bit substitution watermarking modulation. Experiments conducted on natural test images and ultrasound medical images demonstrate the general capabilities of our system to securely make available a message in both spatial and encrypted domains while minimizing image distortion. We further discuss the use of different encryption and watermarking algorithms and we illustrate how our system can be used for ensuring image integrity and authenticity control.