An Efficient Lightweight LSB Steganography with Deep Learning Steganalysis
Dipnarayan Das, Asha Durafe, Vinod Patidar
Abstract
Active research is going on to securely transmit a secret message or so-called steganography by using data-hiding techniques in digital images. After assessing the state-of-the-art research work, we found, most of the existing solutions are not promising and ineffective against machine learning-based steganalysis. In this paper, a lightweight steganography scheme is presented through graphical key embedding and obfuscation of data through encryption. By keeping a mindset of industrial applicability, to show effectiveness of proposed scheme, we emphasized mainly on deep learning based steganalysis. The proposed steganography algorithm containing two schemes withstands not only statistical pattern recognizers but also machine learning steganalysis through feature extraction using a well-known pretrained deep learning network Xception. We provided a detailed protocol of the algorithm at different scenarios and implementation details. Furthermore, different performance metrics are also evaluated with statistical and machine learning performance analysis. The results were quite impressive respect to state of the arts. We received 2.55% accuracy through statistical steganalysis and machine learning steganalysis gave maximum 49.93~50% correctly classified instances at good condition.