Groups of Surface Marine Craft with Intelligent Control: Bayesian Estimation against Adversarial Attacks

With increasing in-roads made by intelligent control into design of marine craft in recent years many challenges are posed as regards safety and security in the course of operation. In particular, when intelligent controlled marine craft are subject to adversarial attack, their vulnerability appears to be of an extreme nature. It also appears that mounting an adversarial attack has its own challenges, the most important being physical. It is currently technically difficult to reach attack surfaces of the marine craft systems which have limited physical accessibility to devices expected to be used while mounting the attack. However, it is plausible that this situation would be of paramount importance to designers in the future. The authors in a previous conference (ICMET 2019) discussed a situation when two surface marine craft are designed to ply in a leader-follower fashion under intelligent control and are subject to adversarial attack assuming a compromised attack surface. It was suggested therein that the nature of intelligent control being Bayesian estimation is preferable from the point of view of security. We take the exploration further and study how a number of such intelligently controlled marine craft behave under adversarial attack. If mounted in such a situation with many craft, the implications of an adversarial attack would be severe if counter measures are not available. Our study shows that vessels beyond a certain number could be well handled within a given area of operation by adjusting parameters of the Bayesian filter employed as a safeguard against adversarial attack.