Abstract
In recent years, concern about information security is increasing. Image-based steganography is an information hiding technique for improving information security. Its purpose is to conceal the existence of secret data. Recently, a stego-image detection scheme was proposed in the literature, which uses the so called extreme learning machine and features generated by Markov chain. Verification results show that some steganography techniques are in fact not very secure. In this study, we investigate detection rates under different conditions and analyze the safety of steganography using support vector machines. Results show that the possibility of detecting the stego-images is very high if the embedding rate is large, and it is necessary to propose stronger steganography techniques in the future.