Surveillance of objects in modern cities through IoT and Edge Computing such as automobile traffic monitoring for public safety has caught attentions recently. However, we observe two primary problems that may hinder scalable deployment of surveillance systems. First, high definition video consumes high bandwidth, although not all the data transferred have enough valid information for object recognition. Second, the centralized architecture often adopted today enforces us to transfer all the data to the central server, which makes public concerns on privacy, since most of the data captured for object recognition may include a great deal of privacy. To solve these problems, we propose this Low Bandwidth and High Privacy Surveillance System Architecture that integrates edge and cloud computing. In our proposed architecture, all the data generated from the local equipment such as surveillance cameras stay stored on the Edge Computing servers to secure them. Only filtered meta data free from privacy are transferred to the Cloud Computing server that can be used for points for the queries for the real raw data. Therefore, both the network bandwidth and the privacy requirements of the data are mitigated. More importantly, the owner of the data can control the access to their data since their data are intact in their hands.

Jiaxing Lu, Kouichirou Amemiya, Kazuhito Matsuda, Makoto Kubota, Akihiro Nakao. “Low Bandwidth and High Privacy Surveillance System Architecture Integrating Edge and Cloud Computing”. 信学技報, vol. 118, no. 371, NS2018-171, pp. 71-76, 2018年12月. copyright©2018 IEICE