نهان‌کاوی کور ویدئو با رویکرد یادگیری شبه‌ناظر برای الگوریتم‌های نهان‌نگاری ویدئوی مبتنی بر بردارهای حرکت
محورهای موضوعی : مهندسی برق و کامپیوترجواد مرتضوی مهریزی 1 , مرتضی خادمی 2 , هادی صدوقی یزدی 3
1 - دانشگاه فردوسی مشهد
2 - دانشگاه فردوسی مشهد
3 - دانشگاه فردوسی مشهد
کلید واژه: شبکه خودسازمان‌ده پویای شبه‌ناظر نهان‌کاوی کور ویدئو نهان‌نگاری ویدئو یادگیری شبه‌ناظر,
چکیده مقاله :
اکثر الگوريتم‌هایي که تا اين زمان در زمينه نهان‌کاوي کور ويدئو ارائه شده‌اند منطبق بر يادگيري باناظر بوده و هزینه تولید داده‌های برچسب‌دار در آنها بالا می‌باشد. تحت این شرایط در فرایند نهان‌کاوی کور ویدئو تنها می‌توان از تعداد محدودی از الگوریتم‌های نهان‌نگاری که کد آنها در دسترس است برای آموزش طبقه‌بند استفاده کرد. بنابراین نمی‌توان درباره کارایی نهان‌کاو برای شناسایی آن دسته از الگوریتم‌های نهان‌نگاری ویدئویی که کد آنها در دسترس نیست، مطمئن بود. همچنین الگوریتم‌های رایج عموماً برون‌خط می‌باشند و بنابراین آموزش مجدد سیستم زمان‌بر بوده و نمی‌توان سیستم را به صورت برخط بروز کرد. برای حل این مشکلات یک روش جدید نهان‌کاوی کور ویدئو با رویکرد یادگیری شبه‌ناظر در این مقاله ارائه شده است. در روش پیشنهادی با توجه به رفع محدودیت برچسب‌دار بودن داده‌های آموزشی، کارایی طبقه‌بند برای آن دسته از الگوریتم‌های نهان‌نگاری که کد آنها در دسترس نیست بهبود می‌یابد. همچنین ثابت می‌شود که روش پیشنهادی نسبت به روش‌های متداول در فرایند نهان‌کاوی کور ویدئو پیچیدگی زمانی کمتری داشته و یک روش برخط بهینه است. نتایج شبیه‌سازی بر روی پایگاه داده استاندارد نشان می‌دهد که روش پیشنهادی علاوه بر مزایای فوق، دارای دقت آشکارسازی قابل مقایسه با روش‌های متداول روز است.
Supervised learning algorithms are widely used in blind video steganalysis and the cost of generating labeled data in them is high. That is why only a limited number of steganography algorithms with accessible code can be used for the training the classifier. Therefore, we cannot be sure about the effectiveness of steganalyzer in identifying non-accessible video steganography algorithms. On the other hand, using offline classification methods in the blind video steganalysis causes the learning process be time consuming and the system cannot be updated online. To solve this problem, we propose a new method for the blind video steganalysis by semi-supervised learning approach. In the proposed method, by eliminating the limitation of labeled training dataset, the classifier performance is improved for video steganography algorithms with non-accessible code. It is also proved that the proposed method, compared to common classification methods for the blind video steganalysis, has less time complexity and it is an optimal online technique. The simulation results on the standard database show that in addition to the above advantages, this method has appropriate accuracy and is comparable to common methods.
[1] I. J. Cox, M. L. Miller, J. A. Bloom, J. Fridrich, and T. Kalker, Digital Watermarking and Steganography, Morgan Kaufmann Publishers, USA, 2008.
[2] Y. Deng, Y. Wu, H. Duan, and L. Zhou, "Digital video steganalysis based on motion vector statistical characteristics," Optik, vol. 124, pp. 1705-1710, 2013.
[3] X. Xu, J. Dong, W. Wang, and T. Tan, "Video steganalysis based on the constraints of motion vectors," in Proc. 20th IEEE Int. Conf. on Image Processing, ICIP'13, pp. 4422-4426, Melbourne, Australia, 15-18 Sep. 2013.
[4] K. Kancherla and S. Mukkamala, "Video steganalysis using spatial and temporal redundancies," in Proc. International Conf. on High Performance Computing & Simulation,., pp. 200-207, 21-24 Jun. 2009.
[5] B. Liu, F. Liu, and P. Wang, "Inter-frame correlation based compressed video steganalysis," in Proc. Congress on Image and Signal Processing, CISP'08, pp. 42-46, 27-30 May. 2008.
[6] K. Kancherla and S. Mukkamala, "Video steganalysis using motion estimation," in Proc. of Int. Joint Conf. on Neural Networks, pp. 1510-1515, Atlanta, GA, USA, 14-19 Jun. 2009.
[7] V. Pankajakshan, G. Doerr, and P. Kumar, "Detection of motion-incoherent components in video streams," IEEE Trans. on Information Forensics and Security, vol. 4, no. 1, pp. 49-58, Mar. 2009.
[8] Y. Cao, X. Zhao, and D. Feng, "Video steganalysis exploiting motion vector reversion-based features," IEEE Trans. on Signal Processing, vol. 19, no. 1, pp. 35-38, Jan. 2012.
[9] Q. Liu, A. H. Sung, and M. Qiao, "Video steganalysis based on the expanded markov and joint distribution on the transform domains-detecting MSU stego video," in Proc. 7th Int. Conf. on Machine Learning and Applications, pp. 671-674, San Diego, CA, USA, 11-13 Dec. 2008.
[10] S. Yuting, C. Zhang, and C. Zhang, "A video steganalytic algorithm against motion-vector-based steganography," Signal Processing, vol. 91, no. 8, pp. 1901-1909, Aug. 2011.
[11] V. Pankajakshan and A. T. S. Ho, "Improving video steganalysis using temporal correlation," in Proc. 3rd Int. Conf. on Intelligent Information Hiding and Multimedia Signal Processing, vol. 1, pp. 287-290, 26-28 Nov. 2007.
[12] U. Budhia, D. Kundur, and T. Zourntos, "Digital video steganalysis exploiting statistical visibility in the temporal domain," IEEE Trans. on Information Forensics and Security, vol. 92, no. 4, pp. 502-516, Dec. 2006.
[13] K. Kashyap and P. K. Bora, "Spatial averaging based steganalysis scheme to detect antipodal watermarks," in Proc. Int. Conf. Signal Processing and Communications, SPCOM'10, 5 pp., 18-21 Jul. 2010.
[14] H. G. Schaathun, Machine Learning in Image Steganalysis, John Wiley & Sons, United Kingdom, 2012.
[15] X. Zhang and S. P. Zhong, "An improved path-based TSVM algorithm for blind steganalysis classification," Artificial Intelligence and Computational Intelligence, vol. 5855, pp. 453-462, 2009.
[16] T. Joachims, "Transductive inference for text classification using support vector machines," in Proc. of The 26th Int. Conf. on Machine Learning, ICML'99, pp. 200-209, 1999.
[17] O. Chapelle and A. Zien, "Semi-supervised classification by low density separation," in Proc. of the 10th International Workshop on Artificial Intelligence and Statistics, AISTAT'05, pp. 57-64, 6-8 Jan. 2005.
[18] A. Hsu and S. K. Halgamunge, "Class structure visualization with semi-supervised growing self-organizing maps," Neurocomputing, vol. 71, no. 16-18, pp. 3124-3130, Oct. 2008.
[19] D. Alahakoon and K. Halgamunge, "Dynamic self-organizaing maps with controlled growth for knowledge discovery," IEEE Trans. on Neural Networks, vol. 11, no. 3, pp. 601-614, May 2000.
[20] C. Xu, X. Ping, and T. Zhang, "Steganography in compressed video stream," in Proc. IEEE Int. Conf. Innovative Computing, Information and Control, vol. 1, pp. 269-272, Beijing, China, 30 Aug.-1 Sep. 2006.
[21] D. Y. Fang and L. W. Chang, "Data hiding for digital video with phase of motion vector," in Proc. IEEE Int. Symp. Circuits and Systems, pp. 1422-1425, 21-24 May 2006.
[22] H. A. Aly, "Data hiding in motion vectors of compressed video based on their associated prediction error," IEEE Trans. on Information Forensics and Security, vol. 6, no. 1, pp. 14-18, Nov. 2011.
[23] http://www.mpeg.org/MPEG/MSSG/tm5/
[24] T. Vollmer, T. Soule, and M. Manic, "A distance measure comparison to improve crowding in multi-model optimization," in Proc. 3rd Int. Symp. on Resilient Control Systems, ISRCS'10, pp. 31-36, 2010.
[25] W. Tsang, T. Kwok, and P. Cheung, "Very large SVM training using core vector machines," J. of Machine Learning Research, vol. 6, no. 6, pp. 363-392, Apr. 2005.
[26] J. C. Bezdek, E. C. K. Tsao, and N. R. Pal, "Fuzzy kohonen clustering networks," Pattern Recognition, vol. 27, no. 5, pp. 757-764, May 1994.
