Anomaly Detection and Countermeasures for Edgeclouds
The accelerated growth of the Internet of Things (IoT) and emerging 5G infrastructure has opened up opportunities to develop intelligent applications that transform data into business and societal value for plenty of application domains such as public services, intelligent transportation, augmented reality, industrial automation, healthcare automation.
Oracle cloud estimates the data and IP traffic produced by IoT devices, people, and machines to exceed 15.3 ZB by 2020. Most the many billions of devices will be located at the edge of the Internet and have higher bandwidth and QoS requirements than today. These requirements urge us to move forward to the era of the Internet of Everything (IoE), which not only adds facilities but also processes a massive amount of data at the edge of the network.
However, the centralized cloud computing model has shown to have inherent problems when it comes to meeting the requirement of multisources or decentralized data processing at the edge of the network. For bandwidth hungry or response time critical applications, network bandwidth and latency have reached a bottleneck due to largescale user access that needs to serve endusers with high latency networks and reliability. Thus, centralized clouds cannot provide services with high performance and reliability for such applications.
Edgeclouds are federations of edge devices, fog nodes, and distant clouds, where the massive amount of data moves back and forth between the edge and distant cloud datacenters that concern data privacy issues. For edge clouds, decentralized autonomous anomaly detection and countermeasures is key for ensuring performance and security, as these problems cannot be expected to be perfectly controlled.
We propose to take a coordinated approach to security and performance analysis detection as we for these closely related problems see great potential for new finding that may be overlooked when studied in isolation. Our plan is to propose a design for anomaly detection in decentralized multilayered architectures and employ machine learning, statistical learning, and information theoretic learning for detection and mitigation of performance and security anomalies in edgeclouds.