Edge-Centric Intelligent Computing: Intelligence, Privacy and Applications

Xiangyang LiNaNA 2020 · OpenReview

Prof. Xiangyang Li, keynote speaker at NaNA 2020.

Abstract

Nowadays, we have witnessed the explosion of data and devices in the areas of IoT and mobile computing. Moreover, we are facing more and more complicated computation tasks, serious privacy and security problems, and critical delay requirements. All these challenges call for the applications of edge computing to deploy computing resource closed to mobile users and the data source. In this talk, based on our own research results, we will mainly focus on the theory and platform implementation of edge intelligence, including online intelligence, distributed intelligence, as well as the privacy and security issues involved. We will also elaborate our current progress in the key applications of our edge platform such as Automatic Driving and AIoT Security.

1. Introduction: The Need for Edge-Centric Computing

The rapid proliferation of Internet of Things (IoT) devices and mobile computing has created a "data explosion" that traditional centralized cloud architectures are ill-equipped to handle. As outlined by Prof. Xiangyang Li in his keynote at NaNA 2020, three primary factors necessitate a shift toward the network edge:

  • Latency Constraints: Modern applications, such as autonomous vehicles, have "critical delay requirements" where network round-trip times to distant servers can be catastrophic.
  • Computational Complexity: Tasks are becoming too sophisticated for individual mobile devices but require faster processing than the cloud can offer.
  • Privacy and Security Risks: Sending sensitive raw data to the cloud creates massive security vulnerabilities.

Edge-centric intelligent computing addresses these by deploying resources closer to the data source, enabling real-time processing and localized data handling.

2. Frameworks for Edge Intelligence

Transitioning from theory to a functional edge platform requires a multi-faceted research approach. Prof. Li’s work focuses on three core pillars:

  • Theoretical Foundations: Developing the mathematical and logical frameworks necessary for decentralized intelligence.
  • Distributed and Online Intelligence: Implementing systems where AI models can learn and perform inference in real-time across a network of heterogeneous edge devices.
  • Platform Implementation: Creating dedicated edge platforms designed to handle specific AIoT (AI + IoT) security protocols and high-stakes computational workloads.

3. Key Insights and Strategic Synergy

A fundamental shift in this paradigm is the realization that intelligence is no longer a centralized service but a distributed feature of the network itself. This shift makes security a foundational element rather than an afterthought; by processing data locally, the attack surface is fundamentally altered, enhancing privacy by design.

Furthermore, the synergy of AI and IoT (AIoT) is the primary proving ground for these theories. The goal is not merely "faster internet," but the creation of "smarter things" that can operate autonomously and securely within a distributed ecosystem.

4. Applications and Academic Impact

The research group has focused on two primary application domains: Automatic Driving and AIoT Security. These fields represent the most demanding environments for edge computing, requiring both high-speed processing and robust security.

The significance of this work is reflected in its academic reception. Prof. Li’s research has garnered over 22,000 citations and has been recognized with seven "Best Paper" awards and multiple "Best Demo" awards at premier venues like ACM MobiCom.

5. Conclusion and Future Outlook

As computation tasks grow increasingly complex, the edge must evolve to handle higher workloads without sacrificing its low-latency advantages. Future research will continue to address the moving target of AIoT security and the scaling of edge-centric theories to broader domains within the mobile computing ecosystem.

Summary updated

Abstract

Nowadays, we have witnessed the explosion of data and devices in the areas of IoT and mobile computing. Moreover, we are facing more and more complicated computation tasks, serious privacy and security problems, and critical delay requirements. All these challenges call for the applications of edge computing to deploy computing resource closed to mobile users and the data source. In this talk, based on our own research results, we will mainly focus on the theory and platform implemenation of edge intelligence, including online intelligence, distributed intelligence, as well as the privacy and security issues involved. We will also elaborate our current progress in the key applications of our odge platform such as Automatic Driving and AIoT Security.
BibTeX
@inproceedings{li2020edge,
  author = {Xiangyang Li},
  title = {{Edge-Centric Intelligent Computing: Intelligence, Privacy and Applications}},
  booktitle = {NaNA},
  year = {2020},
  url = {https://openreview.net/forum?id=B5xUXa5BwF},
}
Figure