IStego100K: Large-scale Image Steganalysis Dataset, mixed with various steganographic algorithms, embedding rates, and quality factors.
This dataset is proposed in:
@inproceedings{yangzl2019IStego100K,
title = {IStego100K: Large-scale Image Steganalysis Dataset},
author = {Yang, Zhongliang and Wang, Ke and Ma, Sai and Huang, Yongfeng and Kang, Xiangui and Zhao, Xianfeng},
booktitle = {International Workshop on Digital Watermarking},
year = {2019},
organization = {Springer}
}
Full PDF can be downloaded from arxiv
100,000 pairs of cover and stego images (200K in total), origin images were downloaded from Unsplash
Marked as SS-Test in the paper. 8104 images with cover/stego labels (not in pair), origin images were downloaded from Unsplash
Marked as DS-Test in the paper.10000 images with cover/stego labels (not in pair), origin images were shot on different mobile devices.
Note: The number of images is 11809 in the paper, but we removed some low quality images before uploading.
We use the following steganographic algorithms for our dataset:
- nsF5: J. Fridrich, T. Pevný, and J. Kodovský, Statistically undetectable JPEG steganography: Dead ends, challenges, and opportunities. In J. Dittmann and J. Fridrich, editors, Proceedings of the 9th ACM Multimedia & Security Workshop, pages 3–14, Dallas, TX, September 20–21, 2007. [code] [pdf]
- J-UNIWARD: V. Holub, J. Fridrich, T. Denemark, Universal Distortion Function for Steganography in an Arbitrary Domain, EURASIP Journal on Information Security, (Section:SI: Revised Selected Papers of ACM IH and MMS 2013), 2014(1). [code] [pdf]
- UERD: L. Guo, J. Ni, W. Su, C. Tang, and Y.Q. Shi. Using statistical image model for jpeg steganography: uniform embedding revisited. IEEE Transactions on Information Forensics & Security, 10(12), 2669-2680, 2015. [code] [pdf]
We apply the following steganalysis algorithms for dataset evaluation:
- DCTR: V. Holub and J. Fridrich, Low Complexity Features for JPEG Steganalysis Using Undecimated DCT, IEEE Transactions on Information Forensics and Security, to appear. [code] [pdf]
- GFR: X. Song, F. Liu, C. Yang, X. Luo and Y. Zhang, Steganalysis of Adaptive JPEG Steganography Using 2D Gabor Filters, Proceedings of the 3rd ACM Workshop on Information Hiding and Multimedia Security. ACM, 2015. [code] [pdf]
- SRNet: M. Boroumand,M. Chen,and J. Fridrich. Deep Residual Network for Steganalysis of Digital Images, IEEE Transactions on Information Forensics and Security. PP. 1-1. 10.1109/TIFS.2018.2871749, 2018. [code] [pdf]
- XuNet: G. Xu. Deep convolutional neural network to detect j-uniward, Proceedings of the 5th ACM Workshop on Information Hiding and Multimedia Security. ACM, 2017. [code] [pdf]
Overall Results
Dataset | Methods | Acc(%) | P(%) | R(%) | F1(%) |
---|---|---|---|---|---|
SS-Test | DCTR GFR SRNet XuNet |
71.34 66.26 - - |
79.72 69.58 - - |
57.23 57.97 - - |
66.63 63.25 - - |
DS-Test | DCTR GFR SRNet XuNet |
56.95 59.12 - - |
55.50 61.61 - - |
70.11 48.42 - - |
61.95 54.22 - - |
Note: We trained SRNet and XuNet on a single GPU (GTX 1080Ti), and found that they are hardly to converge on IStego100K.
Results for Different Steganography Algorithms
Test Set | Steganalysis | Steganography | Acc(%) | P(%) | R(%) | F1(%) |
---|---|---|---|---|---|---|
SS-Test | DCTR | UERD nsF5 J-uniward |
71.77 84.44 57.73 |
79.75 85.10 67.58 |
58.36 83.51 29.71 |
67.40 84.30 41.27 |
SS-Test | GFR | UERD nsF5 J-uniward |
68.47 71.61 58.81 |
71.34 72.72 62.91 |
61.75 69.18 42.92 |
66.20 70.91 51.02 |
DS-Test | DCTR | UERD nsF5 J-uniward |
53.96 62.28 51.67 |
53.35 60.56 51.43 |
63.06 87.59 59.83 |
57.80 71.61 55.31 |
DS-Test | GFR | UERD nsF5 J-uniward |
56.05 67.24 54.59 |
58.40 68.21 56.62 |
42.09 64.58 39.26 |
48.92 66.35 46.37 |
Results for Different Steganography Algorithms
Test Set | Steganalysis | Payload | Acc(%) | P(%) | R(%) | F1(%) |
---|---|---|---|---|---|---|
SS-Test | DCTR | 0.1 0.2 0.3 0.4 |
58.55 71.43 76.30 79.55 |
67.84 80.19 82.22 83.74 |
32.51 56.90 67.11 73.35 |
43.96 66.57 73.90 78.20 |
SS-Test | GFR | 0.1 0.2 0.3 0.4 |
55.87 63.51 70.83 75.71 |
59.40 67.98 72.04 74.89 |
37.10 51.08 67.89 76.75 |
45.67 58.33 69.95 72.05 |
DS-Test | DCTR | 0.1 0.2 0.3 0.4 |
52.86 56.21 58.56 60.17 |
52.42 54.99 56.53 57.72 |
61.90 68.40 74.11 76.05 |
56.77 60.97 64.13 65.63 |
DS-Test | GFR | 0.1 0.2 0.3 0.4 |
52.29 56.66 62.15 65.40 |
53.42 58.87 64.65 67.18 |
35.79 44.19 53.65 60.22 |
42.86 50.49 58.63 63.51 |
Results for Different Quality Factors on SS-Test
Steganalysis | QF | Acc(%) | P(%) | R(%) | F1(%) |
---|---|---|---|---|---|
DCTR | 75 80 85 90 95 |
75.23 71.50 74.09 69.04 62.12 |
85.63 86.48 84.34 76.09 66.41 |
60.64 61.56 59.18 55.54 49.05 |
71.00 71.82 69.55 64.21 56.43 |
GFR | 75 80 85 90 95 |
70.08 69.91 68.42 64.67 58.30 |
75.06 74.98 71.54 67.02 59.76 |
60.15 59.75 61.17 57.75 50.82 |
66.78 66.50 65.95 62.04 64.93 |
For more details such as pre-processing, data distribution, and steganalysis baselines, please take a look at the paper.