Cloud computing has played a very important role in driving the flourish of modern artificial intelligence and data analytical applications during the past decade. However, due to the explosion of mobile devices and data traffic, it becomes very difficult to meet the delay-sensitive and context-aware service requirements by using cloud computing alone. Edge computing emerges to be a potential solution by pushing the intelligent services from the cloud to the network edges that are in closer proximity to end devices and data sources. It provides a complement for cloud computing to make it be only responsible for delay non-sensitive, resource-intensive or computationally complex tasks. Meanwhile, fueled by the Moore’s law, the computing power of mobile devices (like smartphones and vehicles) has experienced exponential increase and been augmented enough to support local intelligence and data analytics. Consequently, a new research trend is to orchestrate the differentiated computing capabilities/resources of end devices, edge servers and the cloud for achieving real-time, flexible and pervasive services and intelligence. Although many existing works are dedicated to improving the performance of end devices by computation offloading and enabling real-time intelligence at the edge, tailored technologies are still missing for end-edge-cloud orchestrated computing systems to fully exploit the potential advantages of the hierarchical architecture.
The purpose of this special issue is to provide a platform for sharing the state-of-the-art research and development on intelligence-enabled end-edge-cloud orchestrated computing. Through an open call for papers and rigorous peer-review, we select 5 papers as representatives of ongoing research and development activities in the following subjects.
End-edge-cloud orchestrated computing applications
The paper, titled “Cooperative Abnormal Sound Event Detection in End-Edge-Cloud Orchestrated Systems”, presents a cooperative abnormal sound event detection framework for city surveillance by end-edge-cloud orchestrated computing. The end-edge-cloud orchestrated audio processing system consists of two processing phases: pre-processing (feature extraction) and post-processing (sound source localization and sound event classification). In the pre-processing phase, the log-mel spectrogram and time of arrival information are first extracted from the audio waveform captured by the distributed acoustic sensors and then sent to the computation entity. In the post-processing phase, the sound source is localized through least-square minimization, while spectrograms are fed into the pre-trained neural networks and the result aggregation algorithm for further classification. In order to fully exploit the hierarchical computational capabilities of different computing devices, the paper also proposes an offloading decision-making scheme to speed up the detection process.
Blockchain for end-edge-cloud orchestrated computing
The paper, titled “Blockchain-Driven Anomaly Detection Framework on Edge Intelligence”, propose an end-edge-cloud orchestrated computing framework for online anomaly detection in IoT systems using blockchain and smart contracts. An efficient feature extractor is developed for the end devices to preprocess the raw log data to simultaneously reduce the data size and keeps sufficient information for anomaly detection model. At the cloud layer of the framework, a well-designed anomaly detection model is deployed to use the processed data to output normal workflow patterns. The edge layer is responsible for leveraging a permissioned blockchain and smart contracts to guarantee data integrity and achieve automatic anomaly detection based on the model output from the cloud layer. The cooperation among the three layers can significantly improve the performance of anomaly detection.
The paper, titled “Towards Fair and Efficient Task Allocation in Blockchain-based crowdsourcing”, proposes a blockchain-based task allocation framework for crowdsourcing with the idea of end-edge-cloud computing. Combining the characteristics of end-edge-cloud orchestrated computing, the proposed framework overcomes the drawback of centralization. A distributed reverse and blind auction-based task allocation mechanism, named RbatAlloc, is presented to realize fair and efficient task allocation in the proposed blockchain-based end-edge-cloud framework, which contains the processes from task release, task matching to solution submission and solution evaluation. In the proposed framework, the cloud layer is responsible for the task release and solution evaluation, and the end layer is responsible for task execution and solution submission. At the edge layer of the framework, a smart contract-based distributed reverse and blind auction method is designed to achieve fair and efficient task matching.
Privacy preservation in end-edge-cloud orchestrated computing systems
The paper, titled “ECAS: An Efficient and Conditional Privacy Preserving Collision Warning System in Fog-based Vehicular Ad Hoc Networks”, designs a collision warning system based on fog computing for vehicular networks. It utilizes fog nodes to collect and compute the speed violation reports from vehicles and broadcast the real-time information to nearby entities. In the proposed system, an ordinary one-way hash function is adopted to improve the performance of signature and authentication, while lightweight encryption is employed in traffic violation reports to assure the integrity and reliability as well as preserving the privacy of received information. Theoretical analysis and experiment results demonstrate the effectiveness and efficiency of the proposed scheme.
The paper, titled “A Trusted Recommendation Scheme for Privacy Protection Based on Federated Learning”, builds an end-edge-cloud orchestrated recommendation system using federated learning. The geographically distributed edge servers form different CDN clusters. In the same CDN cluster, the edge servers can update the local recommendation model through point-to-point interactions and eventually converge to a unified set of model parameters. The cloud server is responsible for data aggregation, global information analysis, massive data storage, and large-scale data calculation. Based on the end-edge-cloud orchestrated computing architecture, the authors design a differentially private recommendation system by adding Laplace noises to the model training for preserving the data privacy. Moreover, by saving the trained model and recommended news to the licensed blockchain network, the proposed recommendation system can achieve permanent preservation and real-time traceability service.
We believe that this special issue covers the state-of-the-art research on end-edge-cloud orchestrated computing/intelligence, and can attract the attention of the community to devote continuous efforts in addressing remaining open challenges. Finally, we would like to sincerely thank Prof. Jianping Wu and Prof. K. K. Ramakrishnan, the Editor-in-Chief of CCF ToN, for their support of the special issue, and Dora Liu, for his guidance and great help during the whole process. We would also like to thank Prof. Wei Li, Southeast University, and Prof. Mingwei Xu, Tsinghua University, for their support on initializing the special issue.
About this article
Cite this article
Ren, J., Jiang, H., Shen, X. et al. Editorial of CCF transactions on networking: special issue on intelligence-enabled end-edge-cloud orchestrated computing. CCF Trans. Netw. 3, 155–157 (2020). https://doi.org/10.1007/s42045-020-00048-5