The Challenges of Rich Features in Universal Steganalysis Tom Pevn a - PowerPoint PPT Presentation

The Challenges of Rich Features in Universal Steganalysis Tomáš Pevný a and Andrew D. Ker b a Agent Technology Center, Czech Technical University in Prague, Czech Republic. a Department of Computer Science, Oxford University, England. 7th February 2013 T. Pevný and A. D. Ker Condensing rich features 7th February 2013 1 / 16

Batch universal steganalysis Internet Warden T. Pevný and A. D. Ker Condensing rich features 7th February 2013 2 / 16

Batch universal steganalyzer Extract features. Calculate distances between actors (MMD). Identify the steganographer(s). local outlier factor (LOF) T. Pevný and A. D. Ker Condensing rich features 7th February 2013 3 / 16

Batch universal steganalyzer Extract features. Calculate distances between actors (MMD). Identify the steganographer(s). local outlier factor (LOF) T. Pevný and A. D. Ker Condensing rich features 7th February 2013 3 / 16

Batch universal steganalyzer Extract features. Calculate distances between actors (MMD). Identify the steganographer(s). guilty local outlier factor (LOF) T. Pevný and A. D. Ker Condensing rich features 7th February 2013 3 / 16

Batch universal steganalyzer Extract features. Calculate distances between actors (MMD). Identify the steganographer(s). guilty local outlier factor (LOF) The method should work with any stego-sensitive features. T. Pevný and A. D. Ker Condensing rich features 7th February 2013 3 / 16

Accuracy with PF274 and CF ∗ features CF ∗ PF274 dimension 274 8750 F5 14.6 9.5 nsF5 10.7 23.1 JP Hide&Seek 7.8 16.2 OutGuess 1.9 5.7 Steghide 2.8 4.7 Average rank of one guilty actor (out of 100) emitting payload 0.1 bpnc T. Pevný and A. D. Ker Condensing rich features 7th February 2013 4 / 16

Curse of dimensionality Anomaly detection estimates density: more difficult in high dimensions. In unsupervised learning cannot discard noise in features. T. Pevný and A. D. Ker Condensing rich features 7th February 2013 5 / 16

Curse of dimensionality Our solution Supervised dimensionality reduction. Our aim Steganographic features should be sensitive to embedding changes, yet insensitive to image content. J. Fridrich, 2004 T. Pevný and A. D. Ker Condensing rich features 7th February 2013 5 / 16

Dimensionality reduction Prior art Principal component transformation Maximum covariance Ordinary least square regression Proposed Calibrated least-squares T. Pevný and A. D. Ker Condensing rich features 7th February 2013 6 / 16

Principal component transformation (PCT) 0 . 4 argmax w k Var ( X w k ) 100 2nd projection subject to 0 . 3 0 w k ⊥ w i , i ∈ { 1 ,..., k − 1 } . 0 . 2 − 100 0 . 1 X ∈ R n , d — matrix with features 0 − 200 − 100 0 100 w i — projections found 1st projection T. Pevný and A. D. Ker Condensing rich features 7th February 2013 7 / 16

Ordinary least square regression (OLS) 0 . 4 − 5 · 10 − 2 w k Cov ( X s w k , Y s ) − Var ( X s w k ) argmax 2nd projection 0 . 3 subject to − 0 . 1 0 . 2 w k ⊥ w i , i ∈ { 1 ,..., k − 1 } . − 0 . 15 0 . 1 − 0 . 2 0 − 0 . 3 − 0 . 2 − 0 . 1 0 0 . 1 0 . 2 0 . 3 0 . 4 X S ∈ R n , d — matrix with stego features 1st projection Y s ∈ R n , 1 — vector with payload w i — projections found T. Pevný and A. D. Ker Condensing rich features 7th February 2013 8 / 16

Maximum covariance (MCV) · 10 6 1 w k Cov ( X s w k , Y s ) argmax 0 . 4 0 . 8 subject to 2nd projection 0 . 6 0 . 3 w k ⊥ w i , i ∈ { 1 ,..., k − 1 } . 0 . 4 0 . 2 0 . 2 0 . 1 0 X S ∈ R n , d — matrix with stego features − 0 . 2 Y s ∈ R n , 1 — vector with payload 0 − 1 . 4 − 1 . 2 − 1 − 0 . 8 − 0 . 6 − 0 . 4 − 0 . 2 0 0 . 2 w i — projections found · 10 8 1st projection T. Pevný and A. D. Ker Condensing rich features 7th February 2013 9 / 16

Calibrated least squares (CLS) 0 0 . 4 w k Cov ( X s w k , Y s ) − Var ( X c w k ) argmax − 2 2nd projection 0 . 3 subject to − 4 0 . 2 w k ⊥ w i , i ∈ { 1 ,..., k − 1 } . 0 . 1 − 6 0 X S ∈ R n , d — matrix with stego features 0 5 10 15 20 Y s ∈ R n , 1 1st projection — vector with payload X c ∈ R n , d — matrix with cover features w i — projections found T. Pevný and A. D. Ker Condensing rich features 7th February 2013 10 / 16

Experimental settings 3000 users of leading social network, 100 images from each ◮ 1000 users for supervised feature reduction ◮ 2000 users used for testing Guilty actor emits payload 0.1 bpnc ◮ linear (in the paper) or greedy strategy ◮ one of following algorithms: F5, nsF5, JPHide&Seek (JP), OutGuess (OG), Steghide (SH) Steganalyst uses reduced CF ∗ features. Accuracy is measured by average rank of guilty actor. ◮ 1 . 0 = perfect, 50 . 5 = random guessing. T. Pevný and A. D. Ker Condensing rich features 7th February 2013 11 / 16

Results PCT MCV OLS CLS F5 40.3 23.4 22.2 1.6 (4) (1) (1) nsF5 38.0 26.6 5.8 2.1 (4) (1) (1) JP 38.4 27.2 6.9 1.7 (5) (1) (1) OG 26.5 31.6 2.4 1.2 (4) (1) (1) SH 23.0 2.6 1.3 1.1 (6) (1) (1) T. Pevný and A. D. Ker Condensing rich features 7th February 2013 12 / 16

Robustness CLS trained on PCT F5 nsF5 JP OG SH F5 40.3 1.6 1.9 8.8 6.6 4.5 (1) (1) (1) (4) (3) nsF5 38.0 1.8 2.1 10.1 10.9 10.5 (1) (1) (1) (4) (3) JP 38.4 8.9 7.2 1.7 15.5 10.5 (1) (2) (1) (2) (2) OG 26.5 3.7 3.0 11.8 1.2 1.1 (1) (6) (2) (1) (1) SH 23.0 5.2 3.2 9.1 1.2 1.1 (1) (6) (2) (1) (1) T. Pevný and A. D. Ker Condensing rich features 7th February 2013 13 / 16

Optimal number of projections F5 30 nsF5 JP OG average rank SH 20 10 0 2 4 6 8 10 # of projections T. Pevný and A. D. Ker Condensing rich features 7th February 2013 14 / 16

Conclusion High dimensional features are not compatible with unsupervised steganalysis. Investigated dimensionality reduction to improve SNR of rich features. Validated the approach in universal batch steganalysis. The proposed method, CLS, exhibits robustness to embedding method. T. Pevný and A. D. Ker Condensing rich features 7th February 2013 15 / 16

F5 phenomenon F5 0 . 25 nsF5 estimated change rate JP 0 . 2 OG SH 0 . 15 0 . 1 5 · 10 − 2 0 0 5 · 10 − 2 0 . 1 0 . 15 0 . 2 0 . 25 true payload T. Pevný and A. D. Ker Condensing rich features 7th February 2013 16 / 16