Research Article

Disguise of Steganography Behaviour: Steganography Using Image Processing with Generative Adversarial Network

Table 2

Testing error of the processed stego image generated from different image processing generation model against SRM and ensemble classifier.

Image processingAlgorithmPayload (bpp)
0.050.10.20.30.40.5

Histogram equalizationHILL0.01650.00180.00080.000100
DCGAN10.48000.47910.41830.39950.38030.3590
WGAN-GP10.48160.46980.41940.40390.39150.3765
S-UNIWARD0.01400.00130.0002000
DCGAN20.44820.44230.41050.39160.38770.3509
WGAN-GP20.46010.45140.41090.40120.38900.3663
WOW0.00950.00070.0001000
DCGAN30.43830.43280.40390.38060.37770.3508
WGAN-GP30.44440.44370.41920.39830.37960.3651

SharpeningHILL0.22490.08370.00820.00140.00040.0002
DCGAN10.48690.48440.41870.40020.39180.3673
WGAN-GP10.49520.49350.44670.41080.40030.3796
S-UNIWARD0.17600.04340.00230.00030.00010.0001
DCGAN20.44960.44830.41270.40090.39250.3824
WGAN-GP20.46980.44940.43270.40250.39880.3925
WOW0.17400.04270.00280.000200
DCGAN30.44770.44240.40890.40150.39750.3912
WGAN-GP30.46280.46220.41820.40210.39920.3962