SECURED IMAGE: A PRACTICAL SYSTEM FOR DETECTING MALICIOUS PAYLOADS HIDDEN IN IMAGES

Abstract

Digital image is one of the most actively trans- mitted types of data in modern communication systems which include email services, social media networks, cloud stores, and web-oriented applications. Although image files are generally considered harmless, they can also be used to cover silently hidden information by using steganography methods. Hidden information may include confidential information, unwarranted messages, or malicious software and pose a possible security threat.

The paper presents an empirical and explainable system, which is named as SecuredImage, a method to detect concealed data in digital images using statistical and image analysis approaches and controlled machine learning approach. The designed solution focuses on simplicity and realism with maintaining the efficient detection. The system is trained using a controlled set of regular images and LSB steganographic images. The inputs to a logistic regression classifier are extracted statistical features mean pixel intensity, variance, Shannon entropy, skewness, kurtosis, file size, among others. Experimental testing proves that SecuredImage can achieve reliable detection safety and computational efficiency, which prove that feature-based machine learning is a promising baseline approach to performing image steganalysis.

Index Terms—component, formatting, style, styling, insert