Embedding a message in media files, also known as steganography, is a common approach to hide secret information. It has been exploited by some criminals to confidentially exchange messages. As a countermeasure, tools have been developed in order to detect hidden information form digital media such as text, image, audio or video files. However the efficiency and performance of previous approaches still have room for improvement. In this research, we focus on algorithm design for better efficiency of hidden message detection from PNG files. We employ three classic AI approaches including neural network, fuzzy logic, and genetic algorithm and evaluate their efficiency and performance in controlled experiments. Finally we introduce our message detection system for PNG files based on LSB approach and present its usability in different case scenarios.