Lossless Data Hiding with Genetic-Based Hybrid Prediction

Abstract

Lossless data hiding is a newly developed topic in information security researches. With the term of 'lossless', secret information is embedded into original image with methods developed by researchers at the encoder, and marked image is produced. Correspondingly, at the decoder, users are capable of perfectly separating the embedded secret and original image from the marked image, based on the reasonable amount of side information, and it is the major reason for the name of 'lossless'. In this paper, we propose an optimized method based on hybrid prediction for lossless data hiding. With the characteristics of the original image, the predicted image can be produced firstly. Then, difference values between the two are utilized for the embedding of secret information. And finally, the marked image can be obtained by adding back the modified difference values. With the properly designed fitness function to control the error between original image and marked one, enhanced amount of secret information can be embedded. Besides, the amount of side information is reasonable for decoding. With the optimization of genetic algorithm, simulation results have revealed that with the same quality of marked images, increased amount of secret information can be observed with our algorithm. It also provides the flexibility in the design of fitness function to meet the needs for practical implementations.