Practical Identification of Specific Emitters Used in the Automatic Identification System

Abstract

There are increasing demands for radio systems to work in more hostile conditions these days. Communication protocols used already by many users, however, need considerable efforts to be modified to increase resistance against deception. Identification technologies of emitters are expected to offer an additional layer to prevent fraudulent devices from accessing wireless systems. They have been developed to distinguish emitters even of the same product based only on emitted signals as well as to handle a large variety of signals efficiently. In this paper a hierarchical identification method consists of a classification of signals into subclasses and an identification of signals in the same subclass is presented; modulation symbol sequences are utilized to classify Automatic Identification System signals into subclasses efficiently and an identifier of a set of binary support vector machines is trained on one of the hardest subclass of signals emitted by the six similar emitters mounted on the six boats. These are the largest number of boats servicing in the same regular line in Japan. Experimental results of identification show practical mean accuracy of 97.6% under a controlled S/N, which corresponds to that of signals sampled over a distance of 100 km.