Privacy-Preserving Coded Mobile Edge Computing for Low-Latency Distributed Inference

We consider a mobile edge computing scenario where number of devices want to perform linear inference ${W}{x} $ on some local data notation="LaTeX">$ {x}$ given network-side matrix {W}$ . The computation is performed at the network over servers. propose coding scheme that provides information-theoretic privacy against notation="LaTeX">$z$ colluding (honest-but-curious) servers, while minimizing overall latency—comprising upload, computation, download, and decoding latency—in presence straggling proposed exploits Shamir’s secret sharing yield straggler mitigation, combined with replication provide spatial diversity for download. also two variants further reduce latency. For considered 9 reduces latency by 8% compared nonprivate recently introduced Zhang Simeone, providing an honest-but-curious server.