Virginia Tech® home

Securing Multi-access Edge Computing using Artificial Intelligence

Principal Investigator: 
Zhi Tian, professor, electrical and computer engineering, George Mason University

Co-Principal Investigators: 
Yang (Cindy) Yi, associate professor, Electrical and Computer Engineering, Virginia Tech

Lingjia Liu, associate professor, electrical and computer engineering, Virginia Tech

Yue Wang, assistant professor, electrical and computer engineering, George Mason University

Securing Multi-access Edge Computing using Artificial Intelligence Project Description:  

The introduction of multi-access edge computing (MEC) as a paradigm for low-latency services has resulted in the emergence of new security and privacy concerns, particularly with regard to autonomous transportation and public infrastructure.

The CCI research team is addressing this challenge through four avenues. First, researchers will develop learning strategies for private virtual machine (VM)-based MEC to improve network security, deploying new reinforcement learning algorithms to personalize the allocation of resources/tasks between users and their VMs based on individual needs. Second, considering the high sensitivity of user data and privacy concerns, researchers will leverage distilled knowledge transfer and differential privacy in VM-based MEC to improve data distribution and increase data protection from attacks or privacy leakage. Third, given the unique challenges of the vehicular paradigm where users connect to edge servers dynamically while moving at high speeds, researchers will address the need for reliable wireless connectivity with high efficiency and low latency required by the distributed continuous learning (CL) models among smart vehicles. They will do so through the integration of continuous learning (CL) design with the underlying wireless transmission mechanism, as well as the incorporation of 1-bit compressed sensing into their CL solutions. Finally, researchers will develop decentralized CL techniques to take advantage of the built-in robustness to node and link failures in dynamic and heterogeneous vehicular AI environments.

The application of this research could contribute not only to the development of autonomous and driving networks, but could have implications for broader infrastructure in the development of smart cities and communities as well.