A novel reversible data embedding method was reported in a recent IEEE journal article. The method was based on difference expansion (DE) technique. It used redundancy in digital images to achieve a high embedding capacity, while keeping visual distortion of the stego-image low. In this thesis, this technique was studied and experimentally evaluated. An effective steganalysis scheme for this DE-based reversible data embedding method was proposed, which used 12-dimensional feature vectors and a Bayes Classifier. The proposed steganalysis scheme steadily achieved a correct classification rate of 99%.