A sensor data generation system based on the fusion of countermeasure network and event graph

• 1、Key Laboratory of Trustworthy Distributed Computing and Service, Ministry of Education
• 2、School of Cyberspace Security, BUPT

Abstract：It has been proved that the generation model of sensor data generation system for the fusion of countermeasure network and event graph into countermeasure network and variational self-encoder can generate synthetic data with high simulation. and it has achieved incredible results in text, image, audio and other fields. However, so far, there is not a mature method to generate sensor data of mobile terminals. This is because the data generated by the sensor is high-dimensional, high-complexity, and contains a lot of noise and changes, which makes it difficult for the generation model to learn the real distribution from these data. In addition, there are different characteristics for different users to collect sensor data of the same action, which brings greater challenges to how to generate fine-grained sensor data. In order to solve the above problems, this paper proposes a personalized sensor data generation system for the fusion of countermeasure network and behavior graph, which creatively introduces the logic graph. The purpose of this scheme is to use the logic and time sequence relationship derived from the logic graph to reduce the probability that the personalized sensor data set does not conform to the logic, and then improve the accuracy of the data generation of the user behavior sensor. The results show that the accuracy of the sensor data generated by the proposed method is up to 90% under the behavior and action recognition task, and the personalized sensor data is generated for a long time through the designed state transition template.

Keywords： Artificial intelligence generative adversarial network variational self-encoder Markov chain event graph