RSIRW-DCF: Dual-Channel Feature-Guided Robust Watermarking Against Dual Attacks for Remote Sensing Image
Wenying Wen, Jiameng Hu, Xiangli Xiao, Baolin Qiu, Yushu Zhang · IEEE Transactions on Consumer Electronics · 2025
Remote sensing image (RSI) involves sensitive information such as military and national. If these data are attacked or stolen during transmission, intellectual property rights could be violated. In some cases, national security might also be jeopardized. Therefore, protecting the copyright of RSI data is essential. However, the existing digital watermarking schemes do not effectively defend RSI against the screen-shooting attack on top of other geometric and non-geometric attacks, i.e., dual attacks. To defend against the threats posed by dual attacks, this paper proposes a dual-channel feature-guided robust watermarking scheme against dual attacks for RSI, called RSIRW-DCF. RSI contains rich texture information. However, it is vulnerable to screen-shooting attack during transmission. So, we employ texture and scale-invariant feature transform (SIFT) features as dual-channel feature to guide watermark embedding in RSI. Specifically, we extract texture complex points from the grayscale covariance matrix and keypoints from SIFT. The intersection of these points determines where the watermark will be embedded. The watermarked RSI generated by this scheme can withstand dual attacks during transmission, protecting the copyright of RSI data. Experiments exhibit that after dual attacks, the bit error rate (BER) of our method is significantly lower than that of the traditional robust watermarking. It is also, on average, 1.67% lower than the BER of the screen-shooting resilient watermarking.