A novel Low-D feature based generic steganalyzer to detect low volume payloads

Abstract

Data hiding techniques whenever used to hide mammoth payloads disturb statistical properties of the cover medium thus leaving a characteristic artifact. These artifacts can provide useful information to the watchful eyes of the steganalyst to identify potential carriers. But the probability of detection sharply declines when the amount of data getting embedded is reduced. Intelligent steganographers as a measure of evading significant artifacts hide only minimal amount of data. This work is an effort to differentiate stego images from innocuous cover images especially when they carry very minimal payloads. A novel low dimensional feature set has been used along with an ensemble classifier.