**8. Discussion**

The egoDetect, a method proposed in this paper, mainly uses visualization to capture and analyze the anomalies in online social network from the perspective of ego network. The analysis of communication data has been verified by experts, which proves the effectiveness and usefulness of the method. When designing the corresponding visual coding and layout algorithm, scalability is our special concern. Our method starts from the macroscopic level, looks for suspicious anomalies

by comparing the status of all nodes in the network. Then, it analyzes suspicious objects by ego network from the mesoscopic level. Finally, from the microscopic level, it analyzes suspicious objects using time series data. In other words, through the high-level abstraction and generalization of the objects, we ensure the scalability of the method. Based on this, we can focus more on the objects themselves, rather than how many attributes there are and what each attribute means. Therefore, this method has good scalability and generality, and can perform well in different fields of time series data. For example, in the field of IOT and cyber-physical social systems, sensors, the basis for data collection, are the most critical part. If the sensors function abnormally or are attacked maliciously, the validity of the data cannot be guaranteed, and the processing and analysis cannot be carried out. Therefore, it is very important to detect abnormal sensors. Just as people can be compared to sensors, and vice versa. Besides, the data collected by sensors can be compared to the call data of people. Through anomaly detection and visual analysis of all sensors, we can quickly target dubious sensors. Then we can analyze dubious sensors based on the visualization of ego network and time series data, and finally we can find out abnormal sensors to ensure the security of the whole network and systems. Therefore, our method has strong scalability, and has grea<sup>t</sup> value for sensor networks, cyber-physical social systems and IOT.
