Spoofing attack on doppler radar based human classification systems

This thesis focuses on spoofing attacks in micro-Doppler radar-based human-animal classification systems. The fundamental theory is that radar systems can be modeled as a communication channel. Therefore, the impossibility of algorithmically detecting spoofing or jamming attacks in a communication channel, as shown in the literature, should also apply to radar systems. The main reason for this is that the problem of detecting jamming attacks in a communication channel is analogous to the halting problem. Hence, this thesis investigates intuitive methods for detecting and resisting such attacks. This theoretical framework is expressed mathematically, followed by experimental demonstrations to validate it. To conduct these demonstrations, an unpublished dataset was designed, implemented, and used to develop a micro-Doppler radar-based human-animal classification system, where jamming attacks were conducted on the system. Detection and diagnosis mechanisms for the attacks were designed, and human/animal classification algorithms were updated to enhance their ability to function under attack conditions.The findings obtained in this thesis are as follows: First, it was demonstrated that radar communication systems can be modeled as a communication layer from the physical layer perspective to establish the theoretical basis. Next, an innovative dataset was collected, which, for the first time in the literature, includes various animal species, providing a foundation for future research. Then, a classical radar signature classifier was implemented, and it was subjected to a classical jamming attack. In the final stage, an innovative artificial intelligence system was designed to detect these attacks and resist them.