Digital Image Steganalysis for Detection of Concealed Information Using Machine Learning

Abstract—Steganography sneaks’ secret info into digitalimages so sneaky, the human eye misses it completely. It'sgreat for secure chats, but a nightmare forcybersecurity—think hidden data leaks or sneakytransmissions. Spotting that concealed stuff? Digitalforensics' holy grail.This paper rolls out a smart, automated steganalysissystem powered by machine learning to catch imageshiding data. We run a straightforward pipeline:preprocess images, extract features from all angles (noiseresiduals, LSB patterns, DCT frequencies, GLCMtextures—241 in total), trim with PCA (keeping 99%variance), then classify via SVM with RBF kernel.Trained on 4,000 images, tested on 800, it hits 83.4%accuracy, 80.5% precision, 88.0% recall, and 84.1% F1-score. That killer recall means it nabs most stegoimages—cybersecurity win. Bonus: Streamlit web appfor instant, realtime checks