Secure and Imperceptible Image Steganography Using Dual Encryption with ACO-Guided Pixel Selection

Abstract

This paper presents a hybrid image steganography scheme that combines cryptography with intelligent embedding to improve imperceptibility, robustness, and security. The secret payload is first encrypted using AES-GCM for confidentiality and integrity, then further scrambled via a classical Rail Fence cipher. Next, an Ant Colony Optimization (ACO) algorithm intelligently selects an adaptive, non-sequential pixel embedding path, and the encrypted bits are hidden using a round-robin LSB substitution across the image’s RGB channels. Experimental results demonstrate that this multi-layer approach achieves high imperceptibility, with PSNR reaching ~74 dB at small payload of 64B and ~56 dB at 5 KB payload, while the Structural Similarity Index (SSIM) consistently stays above 0.999 across all payloads. The method also exhibits strong robustness, with extracted data showing a low BER (~0.25%) under common image processing attacks e.g., noise and compression. Compared to conventional LSB or single-layer steganography techniques, the proposed approach achieves significantly higher image fidelity e.g., ~30 dB PSNR gain over basic LSB at a similar payload and provides security advantages through dual-layer encryption absent in earlier ACO-LSB schemes. These results highlight the benefits of integrating encryption with ACO-based embedding to attain imperceptible, robust, and secure image steganography.