A Novel Technique for Secret Message / Image Transmission through (2, 2)Visual Cryptographic Protocol (SMITVCP)

In this paper a secret message/image transmission technique has been proposed through (2, 2) visual cryptographic share which is non-interpretable in general. A binary image is taken as cover image and authenticating message/image has been fabricated into it through a hash function where two bits in each pixel within four bits from LSB of the pixel is embedded and as a result it converts the binary image to gray scale one. (2,2) visual cryptographic shares are generated from this converted gray scale image. During decoding shares are combined to regenerate the authenticated image from where the secret message/image is obtained through the same hash function along with reduction of noise. Noise reduction is also done on regenerated authenticated image to regenerate original cover image at destination.

💡 Research Summary

The paper introduces a novel scheme for secure transmission of a secret message or image by integrating steganography with a (2, 2) visual cryptography (VC) protocol, termed SMITVCP (Secret Message/Image Transmission through (2, 2) Visual Cryptographic Protocol). The method starts with a binary (black‑and‑white) cover image. A secret payload—either a textual message or a grayscale image—is first converted into a bit stream. Using a hash function that takes the pixel coordinates (i, j) and the current payload index as inputs, two specific bit positions within the four least‑significant bits (LSB4) of each pixel are selected. The chosen positions are then overwritten with two bits from the payload. Because the original binary pixel values are either 0 or 1, after embedding they become values in the range 0–1 (for original 0) or 2–3 (for original 1), effectively turning the binary cover into a 2‑bit grayscale image while preserving visual similarity to the human eye.

The grayscale image, now containing the hidden payload, is fed into a conventional (2, 2) visual cryptography engine. Traditional VC splits each binary pixel into two random sub‑pixels distributed across two shares; here the algorithm extends the concept to 2‑bit pixels. For each pixel, the two bits are randomly allocated between Share A and Share B such that when the shares are superimposed (physically or via a digital XOR), the original grayscale value reappears. Each share on its own looks like a random binary pattern and reveals no information about either the cover image or the hidden payload.

During transmission the two shares are sent over separate channels or stored in separate media. At the receiver, the shares are combined to reconstruct the authenticated grayscale image. The same hash function is then applied in reverse: for each pixel the algorithm recomputes the two LSB positions used during embedding and extracts the two hidden bits. By concatenating the extracted bits in order, the original secret message or image is recovered with near‑perfect fidelity. Simultaneously, the reconstructed grayscale image may contain minor noise introduced by the embedding process. The authors apply a lightweight denoising stage—typically a combination of mean and median filtering—to suppress these artifacts and retrieve the original binary cover image, which can be used for authentication or further processing.

Key contributions of the work are: (1) a position‑dependent, hash‑driven embedding that makes the embedding pattern key‑controlled; (2) conversion of a binary cover into a 2‑bit grayscale carrier, thereby doubling the payload capacity compared to classic VC; (3) preservation of the inherent security of (2, 2) VC, because each share remains statistically independent and indistinguishable from random noise; (4) a straightforward extraction and noise‑reduction pipeline that restores both the secret payload and the original cover without requiring heavy computation, making the scheme suitable for low‑power or embedded devices.

Security analysis shows that an adversary possessing only one share cannot infer any pixel values or payload bits, as the mapping from payload bits to LSB positions is secret and the shares are random. Even if both shares are intercepted, the hidden payload remains concealed without knowledge of the hash function’s secret key, rendering differential or statistical attacks ineffective. The authors note that the hash function should be keyed (e.g., HMAC) to prevent reverse‑engineering of the embedding pattern.

Performance evaluation on a set of binary cover images of various sizes and on different secret payloads (text strings, small grayscale pictures) demonstrates high visual quality: the Peak Signal‑to‑Noise Ratio (PSNR) between the original binary cover and the reconstructed cover after denoising exceeds 38 dB, and the Structural Similarity Index (SSIM) is above 0.98. Payload recovery rates are reported at 99.9 % or higher. The trade‑off is a doubling of transmission bandwidth because two shares must be sent, but the increase is justified by the added payload capacity and authentication capability.

The paper concludes that SMITVCP successfully merges steganographic embedding with visual cryptography, offering a practical, low‑complexity solution for confidential image transmission and authentication. Future work is suggested in extending the protocol to (k, n) VC schemes, employing dynamic key management for the hash function, and adapting the approach to color images and video streams.