基于人工智能的图像指挥控制（C2）框架：利用人工智能模型在C2图像中隐藏和提取命令.pdf

1、#BHAS BlackHatEventsAI-Powered Image-Based Command and Control(C2)Framework:Utilizing AI Models to Conceal and Extract Commands in C2 ImagesQian Feng,Chris NavarretePalo Alto Networks#BHAS BlackHatEventsBlind Image Steganography#BHAS BlackHatEventsBlind Image Steganography(BIS)in AttacksBlind Image

2、Steganography(BIS)in Attacks Metadata manipulation Image pixel manipulationLeast significant bits(LSB)manipulationF5Steghide The encoder or decoder is binary codeOceanLotus APTOilRig#BHAS BlackHatEventsDeep Blind Image Steganography Neural Networks for encoder and decoderAI model for the encoder and

3、 decoderImage content manipulation#BHAS BlackHatEventsAI-Stega Model OverviewAI-Stega Model Overview#BHAS BlackHatEventsImageToTensor Image transformationConvert to Tensortransforms.ToTensor()Normalizationtransforms.Normalize(mean=0.5,0.5,0.5,std=0.5,0.5,0.5)Normalization#BHAS BlackHatEventsSecretTo

4、Tensormessages in png format001100100100000100110111010001100011010101000101001110010100010000111000010000100011000101000011001101000011010101110010011001010110001001101111011011110111010000100000001000000010000000100000001000000010000000100000001000002A7F5E9D8B1C45,reboot Secret TransformationConve

5、rt String to bitsConvert Bits to Tensor torch.tensor(np.array(hack,dtype=np.float32)#BHAS BlackHatEventsEncoder Secret Transformation Feature Learning ComponentConvBnReLU2d Concatenation EncodingConvBnReLU2d#BHAS BlackHatEventsDecoder Inputstega Image Output:secret tensor#BHAS BlackHatEventsMessage

6、Reconstruction#BHAS BlackHatEventsLoss Function Image Reconstruction loss(MSE loss)nn.MSELoss(stega_image,cover_image).to(device)Secret Reconstruction Loss(L1 Loss)np.sum(np.abs(decoded_rounded-secret.numpy()/(batch_size*secret.shape1)#BHAS BlackHatEventsTraining Tasks Train the model for generic da