Abstract:
Steganography-the method of concealing data within digital media-is an important approachtosecure communication because even the very existence of a secret message can be concealed. Inthis thesis, an experimental approach is adopted in using edge detection techniques inimagesteganography to investigate how embedding capacity can be optimized with high-qualityimages. The study examines and compares two popular methods: Canny and Prewitt. Canny's goodprecision and robustness against noise make it suitable for selective embedding with minimal perceptibility, while Prewitt provides computational efficiency and higher embedding capacitybecause of its wider range of edge detection. The proposed methodology integrates the AdvancedEncryption Standard encryption and Run-Length Encoding compression to secure and minimizethe size of the data. The encrypted and compressed data are embedded in the edge-detectedareasof a cover image through LSB substitution guided by a Linear Congruential Generator forenhanced security. Experimental results demonstrate that while Canny gives better image qualitywith higher PSNR and SSIM values, the embedding capacity for Prewitt is higher comparedtoCanny. On the other hand, Prewitt's computational efficiency and capacity make it suitableforreal-time or resource-constrained applications. Evidence that a pragmatic trade-off betweenembedding capacity and image quality does exist is given by this work; furthermore, Cannyturnsout to be more appropriate for security-critical applications, while Prewitt-for those wherecapacity and/or processing speed is in the first place. Future work could investigate hybridedgedetection or more sophisticated compression in pursuit of further optimizing steganographicperformance.
