Damageless Information Hiding Technique Using Neural Network
Kensuke Naoe, Yoshiyasu Takefuji · 2008
In this paper, we propose a new information hiding technique without embedding any information into the target content by using neural network trained on frequency domain. Proposed method can detect a hidden bit codes from the content by processing the selected feature subblocks into the trained neural network. Hidden codes are retrieved from the neural network only with the proper extraction key provided. The extraction key, in proposed method, are the coordinates of the selected feature subblocks and the network weights generated by supervised learning of neural network. The supervised learning uses the coefficients of the selected feature subblocks as set of input values and the hidden bit patterns are used as teacher signal values of neural network. With our proposed method, we are able to introduce a information hiding scheme with no damage to the target content.