Adversarial Deception Against SAR Target Recognition Network

Synthetic Aperture Radar (SAR) automatic target recognition (ATR) technology is one of the key technologies to achieve intelligent interpretation for SAR images. With rapid development deep learning, neural networks have been successively used in ATR and show priority comparison with conventional methods. Recently, more attention paid robustness learning based The reason that maliciously modified imperceptible adversarial images can deceive methods which are on networks. In this paper, we propose a novel deception algorithm, fully considers characteristics data. Our method obtain satisfactory perturbations higher success rate, confidence, smaller perturbation coverage than other state-of-the- art Experimental results using MSTAR dataset OpenSARShip demonstrate effectiveness our method. proposed be applications such as protection, sensor design image quality evaluation.