The Risk and Opportunity of Adversarial Example in Military Field

Yuwei Chen; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2022, pp. 100-107

Artificial intelligence technology is increasingly widely used in the military field, and various countries have carried out a number of research and experiments, aiming to use artificial intelligence technology to shorten the closing time of their own kill chains, and obtain an advantage in the future battlefield, so as to increase the probability of victory in the battle. However, due to the vulnerability of deep learning models before adversarial examples, all systems or modules using artificial intelligence algorithms are at risk of being attacked, thereby delaying or hindering the closure of the opponent's kill chain and increasing the probability of combat victory from another aspect. Based on such risks, this paper proposes a conceptual scheme of military deception by attacking the AI modules of the combat units through adversarial examples, and proposes the challenges and prospects of the current technology. To the best of our knowledge, we are the first to analyze the impact of adversarial examples in the entire process of military operations, that is, the impact of each step and activity in the entire kill chain, and simulate the actual application of adversarial examples in combat through the wargame simulation platform. Ultimately, we found that when AI technology is really widely used in the military field, adversarial examples will have a subversive impact on several activities in several steps in the kill chain, which will directly lead to the interruption of the entire kill chain. This will lead to the failure of combat troops to successfully complete combat missions in accordance with the established objectives.