Steganography anomaly detection using simple one-class classification

Benjamin M. Rodriguez; Gilbert L. Peterson; Sos S. Agaian

doi:10.1117/12.717979

2 May 2007 Steganography anomaly detection using simple one-class classification

Benjamin M. Rodriguez, Gilbert L. Peterson, Sos S. Agaian

Author Affiliations +

Proceedings Volume 6579, Mobile Multimedia/Image Processing for Military and Security Applications 2007; 65790E (2007) https://doi.org/10.1117/12.717979
Event: Defense and Security Symposium, 2007, Orlando, Florida, United States

Abstract

There are several security issues tied to multimedia when implementing the various applications in the cellular phone and wireless industry. One primary concern is the potential ease of implementing a steganography system. Traditionally, the only mechanism to embed information into a media file has been with a desktop computer. However, as the cellular phone and wireless industry matures, it becomes much simpler for the same techniques to be performed using a cell phone. In this paper, two methods are compared that classify cell phone images as either an anomaly or clean, where a clean image is one in which no alterations have been made and an anomalous image is one in which information has been hidden within the image. An image in which information has been hidden is known as a stego image. The main concern in detecting steganographic content with machine learning using cell phone images is in training specific embedding procedures to determine if the method has been used to generate a stego image. This leads to a possible flaw in the system when the learned model of stego is faced with a new stego method which doesn't match the existing model. The proposed solution to this problem is to develop systems that detect steganography as anomalies, making the embedding method irrelevant in detection. Two applicable classification methods for solving the anomaly detection of steganographic content problem are single class support vector machines (SVM) and Parzen-window. Empirical comparison of the two approaches shows that Parzen-window outperforms the single class SVM most likely due to the fact that Parzen-window generalizes less.

Citation Download Citation

Benjamin M. Rodriguez, Gilbert L. Peterson, and Sos S. Agaian "Steganography anomaly detection using simple one-class classification", Proc. SPIE 6579, Mobile Multimedia/Image Processing for Military and Security Applications 2007, 65790E (2 May 2007); https://doi.org/10.1117/12.717979

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

;

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE