Research and Development of Applying Machine Learning Technology to Information Hiding Detection and Analysis

The objective of this project is research and development of applying machine-learning technology to information hiding detection and analysis. The steps of machine learning can divided into three parts: the preparation and selection of input data, the selection and establishment of learning models, and the accuracy and suitability tests of the trained models.Preparation of input data for this project will use three types of digital images for training, namely: general digital images, pixel value histograms, and wavelet transform images. The digital image is processed by various information hiding technologies and put into training, so that important parts can be extracted from various features of the image to improve the efficiency and accuracy of information hiding detection and analysis.This project selects three learning models for analysis: decision tree, support vector machine, and ensemble learning.Training modes of mechanical learning can be divided into four categories: supervised learning, unsupervised learning, semi-supervised learning, and reinforcement learning. The main difference between the various models is the amount of "degrees" given to the answers, so that models with different characteristics can be obtained, and the applicable scenarios are different.Selection and establishment of the learning model of this plan will be based on the training of the above four training modes, analyze the feature proportions with high accuracy of various combinations, and repeat the training after combining again to train a universal detection analysis system.The aim of this project is to improve social security. Through mechanical learning to quickly find all kinds of secret and hidden information existing in the Internet, we look forward to preventing violations of laws and disciplines and preventing crises like the 911 terrorist attacks.

Project ID:PB10907-2446
External Project ID:MOST109-2221-E182-041