Search

Recent works have demonstrated the great performance of forgery image forensics based on deep learning, but there is still a risk that detectors could be susceptible to unknown illegal attacks, raising growing security concerns. This paper starts from the perspective of reverse forensics and explores the vulnerabilities of current image manipulation detectors to achieve targeted attacks. We present a novel reverse decision aggregate gradient attack under low-frequency constraints (RevAggAL). Specifically, we first propose a novel pixel reverse content decision-making (PRevCDm) loss to optimize perturbation generation with a specific principle more suitable for segmenting manipulated regions. Then, we introduce the low-frequency component to constrain the perturbation into more imperceptible details, significantly avoiding the degradation of image quality. We also consider aggregating gradients on model-agnostic features to enhance the transferability of adversarial examples in black-box scenarios. We evaluate the effectiveness of our method on three representative detectors (ResFCN, MVSSNet, and OSN) with five widely used forgery datasets (COVERAGE, COLUMBIA, CASIA1, NIST 2016, and Realistic Tampering). Experimental results show that our method improves the attack success rate (ASR) while ensuring better image quality. [ABSTRACT FROM AUTHOR]

The fusion of hyperspectral (HS) and multispectral (MS) images with complementary high spectral and high spatial resolution information has been successfully applied to improve the low spatial resolution limitation of current satellite sensors with high spectral resolution. Fusion methods based on spectral unmixing have led to state-of-the-art results with the advantage of obtaining pure spectral signatures and their proportion in the image of analysis as a result of the fusion process. However, pure spectral signatures observed in HS images are affected by variations in atmospheric, environmental and illumination conditions which are significant sources of errors in HS image analysis. Conventional unmixing-based fusion methods neglect the spectral variability of HS images introducing and propagating errors through the fusion process, affecting further inference tasks such as classification, target detection, and change detection. Therefore, this paper proposes HS-MS image fusion algorithm based on spectral unmixing accounting for spectral variability through an augmented linear mixing model (ALMM). An alternating optimization strategy is combined with the alternating direction method of multipliers (ADMM) to efficiently minimize the underlying cost function of the fusion problem. Simulation results on realistic data demonstrate that the proposed method outperforms state-of-the-art fusion methods in different metrics. Additionally, the fusion of crop images acquired with DESIS and Sentinel-2 sensors in a region of Colombia confirms the advantages of the proposed fusion method. [ABSTRACT FROM AUTHOR]