Skip to main content

Table 6 Robustness(%) with respect to the different CNN model in Caltech-101

From: Coverless image steganography based on DenseNet feature mapping

Attack

Parameter

InceptionResNetV2

ResNet50

InceptionV3

DenseNet121

Compression

Q(10)

32.10

27.80

34.44

80.59

Gauss noise

σ(0.001)

57.17

47.32

55.90

92.10

Gauss filtering

(3×3)

71.41

62.83

67.02

95.32

Scaling

3

95.02

93.85

91.61

99.12

C-H-E

–

29.37

21.56

32.88

81.27

Gramma

0.6

55.61

48.10

53.27

89.85

Centered cropping

20%

5.56

15.90

15.71

72.88

Edge cropping

20%

41.95

23.90

39.41

76.98

Rotation

20 ∘

12.10

6.24

16.78

71.02

Translation

(45,30)

23.61

21.66

24.59

75.02