Comparison of computation time

We compare the computation with different sample rate. The computation time is defined as the time to compute an AMAC tag. The data size we use is 800*600 bytes. Sample rate = 0.03 the computation time is 7ms, and when the sample rate is 0.06, the computation is 11ms, which is not as twice as 7ms. Moreover, consider the sample rate = 1, which means total data are used to compute the AMAC tag and no sampling was used, the computation time is only 15ms, not linear growth with the sample rate. We found that with random sam-pling, most of computation time is using on the random number generation for the position that random sampling need. If we eliminate the time used on random number generation, sample rate = 0.03 the computation time is 1ms, and when the sample rate is 0.06, the computation is 1ms, sample rate = 1 is 15 ms. The result computation time is close to linear with the sample rate, as we expected. We use the pseudo random function of c++, more effi-cient pseudo random function can increase the performance, still the benefit of random sampling is restrict by the pseudo random generation.

with random table

computation time for a AMAC tag, data size = 800*600 bytes

48

Chapter 6

Security analysis and discussion

The AMAC construction already satisfies some kind of security property. Assume that the construction is distance-preserving, for some distance functions and parameters. Then con-sider an adversary trying to produce two messages m1, m2 such that dm(m1, m2) > c2 , and the key k has been generated using the key-generation algorithm. An adversary tries to convince a verifier that the same tag is valid for both m1 and m2, which means Verification of m2 with the tag t1=T(m1) pass with higher probability than p2. However, since the k is generated ran-domly and the attacker has no information about the key, the attacker must cannot know which positions of data are chosen to compute the tag. Without any information of the posi-tion chooses, the attacker chooses the posiposi-tions seems randomly from the perspective of the verification. The only way to broken the security is to break the key or pseudo random num-ber generation.

For the perceptual data attack, the attacker changed the data within the acceptable amount but not randomly, the multimedia data is not the same from human perspective. In our AMAC, only the error number can detect, not the perceptual data error. If the attacker

49

changed the data with the amount of errors that blows the threshold, he will not be detect-ed by our AMAC. The error position and error distance are not measurdetect-ed in our tag function.

To enhance our AMAC for detecting attacker, preprocessing the multimedia data to extract the perceptual feature is helpful. There are many works that extract different types of mul-timedia data features, we make the assumption that the extract features are suitable for further histogram feature extraction, which means the errors in features of multimedia are not location and distance correlated. Thus, we can apply feature extraction of the type of multimedia data and then apply our AMAC, the final AMAC can detect the attacker.

50

Chapter 7 Conclusion

We proposed an AMAC with sensitivity control that can well adjust to different error thresholds, and consider the nature of AMAC using random sampling to reduce the compu-tation time of AMAC. We experiment our AMAC and compare to others, the results show that our AMAC has advantages than the others. However, to distinguish forgeries from inci-dental errors, we need to combine our AMAC with robustness preprocessing to keep the features of multimedia data. To enhance the robustness and combine with techniques of multimedia data preprocessing are future works.

51

References

[1] Emin Martinian and Gregory W. Wornell, “Multimedia content authentication:

fundamental limits,” in IEEE International Conference of Image Processing, Rochester, NY, Sep. 2002, vol. 2, pp. 22–25.

[2] Bin B. Zhu, Mitchell D. Swanson, and Ahmed H. Tewfik, “When seeing isn’t believing:

Current multimedia authentication technologies and their applications,”

in IEEE Signal Processing Magazine, pp. 40–49, Mar. 2004.

[3] Chai Wah Wu, “On the design of content-based multimedia authentication

systems,” in IEEE Transactions on Multimedia, vol. 4, no. 3, pp. 385–393, Sep. 2002.

[4] Richard Graveman and Kevin E. Fu, “Approximate message authentication codes,” in Ad-vanced Telecommunications & Information Distribution Research Program, vol. 1, Col-lege Park, MD, Feb. 1999.

[5] C Kailasanathan, R Safavi-Naini, and P Ogunbona, “Image authentication surviving acceptable modifications,” in IEEE-EURASIP Workshop on Nonlinear Signal Image Processing, Baltimore, MD, Jun. 2001.

52

[6] M. Kıvanc Mıhcak and Ramarathnam Venkatesan, “A tool for robust audio information hiding: A perceptual audio hashing algorithm,” in International Information Hiding Work-shop, Philadelphia, PA, Nov. 2001.

[7] Liehua Xie, Gonzalo R. Arce, and Richard F. Graveman, “Approximate Image Message Authentication Codes,” in Collaborative Tech Alliance Communications and Networks, College Park, MD, pp. 217–221, Apr. 2003.

[8] Karen M. Bloch and Gonzalo R. Arce, “Analyzing protein sequences using signal analysis techniques,” in Computational and Statistical Approaches to Genomics,

pp. 113-124, 2003.

[9] Ashwin Swaminathan, Yinian Mao, and Min Wu,” Robust and Secure Image Hashing,”

in IEEE Transactions on Information Forensics and Security, Vol. 1, No. 2, June 2006 [10]Marc Schneider and Shih-Fu Chang, “A robust content based digital signature for image

authentication,” in IEEE International Conference on Image, pp. 227-230, Sep. 1996.

[11]Der-Chyuan Lou and Jiang-Lung Liu, “Fault resilient and compression tolerant digital signature for image authentication,” in IEEE Transactions on Consumer Electron, vol. 46, no. 1, pp. 31–39, Feb. 2000.

[12]Ee-Chien Chang, Mohan S. Kankanhalli, Xin Guan, Zhiyong Huang, Yinghui Wu, “Robust image authentication using content based compression,” in ACM Multimedia System, vol. 9, pp. 121–130, Aug. 2003.

[13]Ching-Yung Lin, and Shih-Fu Chang, “A robust image authentication method

distinguishing JPEG compression from malicious manipulation,” in IEEE Transactions on Circuits System Video Technology, vol. 11, no. 2, pp. 153–168, Feb. 2001.

53

[14]Chun-Shien Lu, and Hong-Yuan Mark Liao, “Structural digital signature for image authentication: An incidental distortion resistant scheme,” in IEEE Transactions on Multimedia, vol. 5, no. 2, pp. 161–173, Jun. 2003.

[15]Nasir D. Memon, Poorvi L. Vora, Boon-Lock Yeo, Minerva M. Yeung, “Distortion bounded authentication techniques,” in Society for Imaging Science and Technology Security and Watermarking of Multimedia Contents, vol. 3971, San Jose, CA, pp. 164–174, Feb. 2000.

[16]M. Kıvanc Mıhcak and Ramarathnam Venkatesan, “New iterative geometric methods for robust perceptual image hashing,” in ACM Workshop on Security and Privacy in Digital Rights Management, Philadelphia, PA, Nov. 2001.

[17]R. Venkatesan, M. H. Jakubowski, and P. Moulin, “Robust image hashing,”

in IEEE International Conference on Image Processing, VOL. 3, NO. 2, pp. 664-666, Jun. 2001.

[18]Renwei Ge, Gonzalo R. Arce, and Giovanni Di Crescenzo, “Approximate Message

Authentication Codes for N-ary Alphabets,” in IEEE Transactions on Information Forensics and Security, VOL. 1, NO. 1, Mar. 2006

[19]Sujoy Roy, Qibin Sun, “Performance Analysis of Locality Preserving Image Hash,”

in IEEE International Conference of Image Processing, pp. 12-15 Oct. 2008.

[20]Venu Satuluri and Srinivasan Parthasarathy,” Bayesian Locality Sensitive Hashing for Fast Similarity Search,” in VLDB Endowment, Vol. 5, No. 5, pp. 430-441, 2012.

[21]Dahua Xie, Tzung-Her Chent and C.-C. Jay Kuo, ”Distortion-Aware Multimedia Authenti-cation using An Enhanced Hash Function,” in IEEE Multimedia Signal Processing, Shanghai, China,pp. 1-4, Oct. 2005.

[22]Mayur Datar,” Locality-Sensitive Hashing Scheme Based on p-Stable Distributions,”

in ACM Symposium on Computational Geometry, pp. 253-262, Jun. 2004.1