Comparative Analysis of Image Steganography Based on Convolutional Neural Network and Canny Edge Detection on Digital Images

Rapid development of information and communication technology, the need for data security in digital information exchange is becoming increasingly important. One of the techniques used to maintain confidentiality is steganography, which is a method of hiding confidential data in digital media such as images, audio, information, or video without causing quality from third parties. This study conducted a comparative study of the Least Significant Bit (LSB) steganography method that optimizes edge detection using the Canny edge detection Convolutional Neural Network (CNN). The developed model was tested using 26 gringsing ikat woven cloth motifs used as cover images and the weaver's palms used as secret data inserted into the cover image. The results showed that Canny-LSB and CNN-LSB could produce good stego images with a PSNR value of 58 dB; CNN-LSB is superior in terms of storage capacity and speed of insertion and extraction processes but is less vulnerable to technological attacks. The model developed in this study can insert 61 % more data than previous studies. In addition to inserting more data, this study produces better stego image quality; the PSNR value indicates this produced, 83% higher than similar studies.