FaviconStegoNet: Lightweight Deep Steganalysis for Tiny Images with Hidden Messages
Abstrak
A rising cybersecurity threat, steganography has made it possible to engage in covert communications in digital media, especially in very small image formats (32×32 favicons) where Least Significant Bit (LSB) methodologies can be used to insert malicious code in the form of URLs, PHP scripts, and so on. Such hidden data is hard to detect because of small pixel variations and low resolution. It has been suggested in this paper that FaviconStegoNet is a lightweight deep steganalysis framework, composed of a small CNN with an optional MLP classifier, and trained to learn bit-plane and pixel-level patterns more efficiently. Assessment of the model is conducted using the Stego-Favicons-Dataset, which comprises 360,000 images: 60,000 clean, 60,000 PHP-stego, and 240,000 URL-stego samples. The experimental findings indicate that the proposed method achieves an accuracy of 97.2 percent, a precision of 96.8 percent, and an F1-score of 97.0 percent, outperforming lightweight baseline models with low computational complexity. These results demonstrate the usefulness of FaviconStegoNet to perform real-time steganalysis in resource-constrained and cybersecurity-critical settings. Overall, the proposed method strikes a reasonable balance between performance and efficiency, making it a suitable choice for use in cybersecurity systems. Exploring additional image resolutions and more advanced steganography techniques could further enhance its effectiveness.
Identifier
- DOI: https://doi.org/10.62527/joiv.10.3.5579
- OpenAlex: https://openalex.org/W7163820942