Abstract: With the rapid development of artificial intelligence technology, the deep learning-based automatic target recognition (ATR) method for synthetic aperture radar (SAR) has achieved remarkable results in SAR image interpretation. However, introducing subtle perturbations that are imperceptible to human eye into original SAR images can easily lead to recognition errors in SAR-ATR models, indicating that these models are affected by adversarial attacks. At present, adversarial attack techniques are increasingly being applied in the SAR-ATR field. Firstly, the current research status of SAR image adversarial attack both at home and abroad is systematically reviewed, and the principles, advantages, disadvantages, and applicable scenarios of different algorithms are deeply analyzed in the paper. Secondly, performance evaluation of typical SAR adversarial attack algorithms is carried out using publicly available datasets. Finally, considering the limitations of existing algorithms, five topics worthy of study in this field are proposed, which can provide theoretical reference for subsequent studies such as conducting research on adversarial attacks in complex domain, cross-data domain and cross-task scenarios, establishing a multi-dimensional connection mechanism between digital and physical domains, as well as enhancing model robustness and improving the evaluation system.