Asymmetric Computing Architectures for Secure Embedded and Edge Systems in Cyber-Physical Applications

The growing deployment of embedded and edge computing systems in cyber-physical applications creates a need for computing architectures that balance security capabilities with the resource constraints of edge devices. This paper presents asymmetric computing architectures for secure embedded and edge systems, where different components of the security workload are distributed asymmetrically between resource-constrained edge devices and more capable edge or cloud infrastructure. We propose asymmetric autoencoder architectures where lightweight encoders perform feature extraction on edge devices while more complex decoders operate on edge servers or cloud infrastructure. The approach extends to asymmetric encryption, asymmetric authentication, and asymmetric intrusion detection. Experimental evaluation demonstrates that asymmetric architectures reduce edge device computational requirements by 55-68% compared to symmetric approaches while maintaining security effectiveness within 3% of symmetric baselines. We also present implementation guidelines for deploying asymmetric architectures in various cyber-physical application domains.