Developing an adaptive DCT-based steganography method using a genetic algorithm

Steganography is an appropriate approach to establish a secure connection between the sender and the receiver. Data embedding in Discrete Cosine Transform (DCT) coefficients for JPEG images is one of the most practical approaches nowadays. In this paper, a new method called GA-Shield is proposed, in which, instead of using fixed embedding capacity, embedding a different number of bits in the quantized DCT coefficients according to the magnitude of the coefficient is used to spread bits of secret message in the most suitable coefficients. In addition, this method uses a genetic algorithm to minimize the distortion due to embedding. This minimization is performed by deciding on the best formula to calculate coefficient value after embedding. In this phase, PSNR is used as the metric to measure the amount of distortion in the cover image to produce the stego image. As these changes decrease, the value of PSNR would be optimized, and the stego image would have better quality. The proposed method can embed 300 to 20,000 bits of data (on average) in the cover image and produce the stego image with a PSNR value in the range of 65 to 40 and a SSIM value of more than 0.985. The consequences of comparisons with the state-of-the-art show that despite the fact that the proposed technique has less embedding capacity than some of the current ones, the superiority of stego image quality and security of the proposed technique, mainly at low embedding levels, is significant.