An Effective Authenticating Method On The Compressed Image Data

(1)

An Effective Authenticating Method On The Compressed Image Data

Tsong-Yi Chen,

*

Da-Jinn Wang, Thou-Ho Chen and Chung-Yih Lee

Department of Electronic Engineering, National Kaohsiung University of Applied Sciences,

Kaohsiung, Taiwan 807, R.O.C.

*

Department of Information Management, National Kaohsiung Marine University,

Kaohsiung, Taiwan 811, R.O.C.

{chentso, thouho, 1093320118}@cc.kuas.edu.tw; wangdaj@mail.nkfu.edu.tw

Abstract

In this paper, we propose an effective watermarking method dedicated to authenticating the compressed image data. By exploiting the characteristics of JPEG compressed data, we compare the quantized with the non-quantized DCT coefficients for finding the embedding positions. A block-classification strategy is used for classifying 8x8 DCT blocks into key-blocks and flat-blocks in order to hide different watermarks into the positions mentioned above. Experimental results show that the efficacy of the proposed approach for JPEG images with about 75% to 90% compression ratio and the watermarked image still maintains a PSNR over 37dB, and only a slight overhead in bit-streams is introduced.

1. Introduction

In the past, several techniques and concepts based on data hiding or steganography have been designed as a means for tamper detection in digital images and for image authentication – fragile watermarks, semi-fragile watermarks, and self-embedding[1]-[13]. One class of authentication watermarks is formed by semi-fragile watermarks. Such watermarks are marginally robust and are less sensitive to pixel modifications. Thus, it is possible to use them for quantifying the degree of tamper and distinguish simple LSB shuffling from malicious changes, such as feature adding and removal. Lin and Chang[6] present a relation-based watermarking techniques for image authentication to extract a digital signature by using the invariant relation existing between any two DCT coefficients which were located at the same position of two different 8x8 blocks for making the image authentication system to be tolerable to JPEG

compression. They dedicated themselves to exploring the operation in a JPEG-based system.

The authentication errors may be divided into two types [12], false negative and false positive. However, the clipping-error owing to normalization in spatial domain is still an intractable problem. To cope with such a problem, an authentication method for lossy compressed video data by semi-fragile watermarking is developed in [13]. Although the clipping-error problem can’t be overcome completely.

In this paper, we focus on authenticating compressed image data. An effective technique which can detect malicious manipulations on the compressed image data (e.g. JPEG) and reduce the clipping-error problem is proposed. The basic idea behind the proposed method is firstly to classify DCT blocks into the smooth-blocks and the detail-blocks. To detect and locate alterations of the tampered image, simple features are embedded into the above classified blocks invisibly. By pre-processing DCT coefficients before quantization, the clipping-error problem resulted form normalization can be reduced significantly. For later authentication, the watermarked image is then put back into the compressed bit-streams.

2. Embedding authentication information

into the JPEG streams

The embedding position is one of important issue in watermarking techniques. Most watermarking systems embed watermark-bits into the least-significant bit (LSB) of the specified coefficients. Embed data into the LSB bits can reduce the damage induced by watermarking, but it’s also easy to be attacked since the embedded bits will be dropped by any simple ways such as set to zero. When hiding information into LSB bits, traditional image processing even if the lossless-transforms can destroy the embedded information. For

(2)

Table 1. The bit-streams after compress. Compared with water-marked and non-watermarked images. (unit: bytes)

Lena, size of 256x256 Motorcycle, size of 352x288 JPEG QF non- watermarked watermarked non- watermarked watermarked 75% 12,736 12,788 17,095 17,047 80% 15,301 15,058 19,412 19,124 85% 17,427 17,276 22,720 22,485 (a) (b) (c) (d)

Figure 3. (a) JPEG-image, Motorcycle, of size 352 x 288, (b) Watermarked JPEG image with QF=75%, authentication strength = 5 and PSNR = 37.00, (c) Illegal tampered image, (d) Authentication with detected blocks 35.

5. Conclusions

In this paper, we present an effective technique which can detect malicious manipulations on compressed image data (e.g. JPEG) and reduce the clipping-error problem. The quantized coefficients are more suitable than original coefficients for data embedding. In this paper, we can reduce the problem of clipping-errors which are caused by normalization in spatial domain. The results show that the embedded authentication bits will not raise too many bit-streams when compressing the watermarked image with standard JPEG compression operation.

Besides, we cannot embed too many authentication-bits into the flat-blocks due to the value and number of the non-zero quantized AC coefficients is small and less. So we use a specific approach to embed the watermark into the flat-block. Further, we can try to recover the tampered image according to the few authentication keys.

6. References

[1] M. Yeuang and F. Mintzer, “An Invisible Watermarking Technique for Image Verification,” Proceedings of IEEE

International Conference on Image Processing, Santa

Barbara, Oct 1997.

[2] M. Wu, and B. Liu, “Watermarking for Image Authentication,” Proceedings of IEEE International Conference on Image Processing, Chicago Illinois, USA, vol.

2, pp. 437-441, Oct. 4-7, 1998.

[3] M. Schneider and S-F. Chang, “A Robust Content Based Digital Signature for Image Authentication,” Proceedings of

IEEE International Conference on Image Processing,

Lausane, Switzerland, Oct. 1996.

[4] S. Bhattacharjee and M. Kutter, “Compression Tolerant Image Authentication,” Proceedings of IEEE International

Conference on Image Processing, Chicago, October 1998.

[5] Kwang-Fu Li, Tung-Shou Chen, and Seng-Chen Wu, “Image tamper detection and recovery system based on discrete wavelet transformation,” Communications,

Computers and Signal Processing, 2001 IEEE Pacific Rim

Conference on, pp. 164-167, vol. 1, Aug. 2001.

[6] C.Y. Lin, S.F. Chang, “SARI: Self-Authentication-and-Recovery Image watermarking system,” ACM Multimedia

2001 Workshops - 2001 Multimedia Conference, 628-629,

2001.

[7] Chung-Shien Lu, and Hong-Yuan Mark Liao, “Structural Digital Signature for Image Authentication: An Incidental Distortion Resistant Scheme,” IEEE Transactions

on Multimedia, vol. 5, No. 2, June 2003.

[8] Phen Lan Lin, Chung-Kai Hsieh, Po-Whei Huang, “A Hierarchical Digital Watermarking Method for Image tamper Detection and Recovery,” Pattern Recognition, vol. 38, Issue: 12, pp. 2519-2529, Dec. 2005.

[9] L. M. Marvel, G. Hartwig, and C. G. Boncelet Jr., “Compression Compatible Fragile and Semi-Fragile Tamper Detection,” SPIE International Conf. on Security and

Watermarking of Multimedia Contents II, vol. 3971, No. 12,

EI `00, San Jose, USA, Jan 2000.

[10] Frank Hartung, Bernd Girod, “Digital Watermarking of Raw and Compressed Video,” Proceedings SPIE Digital

Compression Technologies and Systems for Video Communication, vol. 2952, pp.205-213, 1996.

[11] M.D. Swanson, Bin Zhu, A.H. Tewfik, “Multi-resolution Scene-based Video Watermarking Using Perceptual Models,” IEEE Journal on Selected Areas in

Communications, 16(4), pp. 540-550, 1998.

[12] I. J. Cox, J. Kilian, T. Leighton, and T. Shamoon, “Secure Spread Spectrum Watermarking for Multimedia,”

Proceedings of IEEE International Conference on Image Processing, vol. 3, pp. 243-246, Sep. 1996.

[13] T. Y. Chen, et al., “Authentication of Lossy Compressed Video Data by Semi-Fragile Watermarking,” Proceedings of

IEEE International Conference on Image Processing,

Singapore, pp. 2637-2640, Oct. 2004.