From: Defeating data hiding in social networks using generative adversarial network
 | \( {T}_{\phi_{\mathrm{QIM}}}^{100} \) | \( {T}_{\phi_{\mathrm{SS}}}^{100} \) | \( {T}_{\phi_{\mathrm{ULPM}}}^{100} \) | ||||||
---|---|---|---|---|---|---|---|---|---|
Error rate | PSNR | SSIM | Error rate | PSNR | SSIM | Error rate | PSNR | SSIM | |
PQIM − SS − ULPM | 44.24% | 40.86 | 0.9834 | 29.34% | 43.30 | 0.9914 | 37.38% | 43.46 | 0.9902 |
PQIM − ULPM − SS | 44.23% | 40.89 | 0.9834 | 22.39% | 43.67 | 0.9919 | 37.37% | 43.59 | 0.9903 |
PSS − QIM − ULPM | 35.09% | 40.72 | 0.9822 | 38.00% | 43.16 | 0.9911 | 37.48% | 43.46 | 0.9902 |
PSS − ULPM − QIM | 49.95% | 40.69 | 0.9823 | 26.58% | 43.50 | 0.9916 | 37.38% | 43.61 | 0.9903 |
PULPM − QIM − SS | 36.19% | 40.74 | 0.9814 | 37.96% | 43.17 | 0.9911 | 37.41% | 43.48 | 0.9902 |
PULPM − SS − QIM | 38.84% | 40.66 | 0.9811 | 34.17% | 42.84 | 0.9901 | 37.39% | 43.61 | 0.9903 |
5 × 5 Wiener filtering | 9.21% | 29.22 | 0.8817 | 15.92% | 29.37 | 0.8930 | 19.93% | 29.22 | 0.8840 |
5 × 5 Gaussian filtering | 14.13% | 25.41 | 0.8479 | 19.26% | 25.48 | 0.8543 | 20.38% | 25.42 | 0.8443 |
5 × 5 median filtering | 3.93% | 22.63 | 0.7263 | 3.34% | 33.65 | 0.7291 | 18.20% | 22.63 | 0.7247 |
Salt and pepper noise_0.05 | 31.19% | 17.92 | 0.5156 | 19.45% | 17.92 | 0.5135 | 28.45% | 17.97 | 0.5183 |
Gaussian noise_0.05 | 22.61% | 19.45 | 0.5664 | 12.74% | 19.45 | 0.5640 | 24.21% | 19.46 | 0.5663 |
gamma correction_0.3 | 5.31% | 10.39 | 0.6747 | 7.28% | 10.39 | 0.6769 | 0.00% | 10.39 | 0.6752 |