CoLearn

Abstract

Edge computing and Federated Learning (FL) can work in tandem to address issues related to privacy and collaborative distributed learning in untrusted IoT environments. However, deployment of FL in resource-constrained IoT devices faces challenges including asynchronous participation of such devices in training, and the need to prevent malicious devices from participating. To address these challenges we present CoLearn, which build on the open-source Manufacturer Usage Description (MUD) implementation osMUD and the FL framework PySyft. We deploy CoLearn on resourceconstrained devices in a lab environment to demonstrate (i) an asynchronous participation mechanism for IoT devices in machine learning model training using a publish/subscribe architecture, (ii) a mechanism for reducing the attack surface in FL architecture by allowing only IoT MUD-compliant devices to participate in the training phases, and (iii) a trade-of between communication bandwidth usage, training time and device temperature (thermal fatigue). CCS CONCEPTS • Networks → Network algorithms; Network experimentation; Network privacy and anonymity; • Computer systems organization → Embedded and cyber-physical systems; • Computing methodologies → Neural networks; Machine learning.

The ownership in The Draft of Civil Code in Serbia

“…Another important point that needs clarification is that PTB-FLA is just a FL framework, and it is not a complete system such as CoLearn [16] and FedIoT [17]. CoLearn is an FL system based on the open-source Manufacturer Usage Description (MUD) implementation osMUD and the FL framework PySyft, whereas FedIoT is a system for realistic IoT devices (e.g., Raspberry PI) that comprises a specialized FL framework for IoT cybersecurity named FedDetect.…”

Artificial intelligence of things at the edge: Scalable and efficient distributed learning for massive scenarios

“…There is comparatively little research on ML applications that use ICN and Kafka. Feraudo et al have developed a pub/sub based selection mechanism for IoT devices to participate asynchronously in the FL process [25]. In [26], the researchers developed a method for training and inferring ML models that feed the continuous dataset directly into the ML model via the Kafka pipeline, as opposed to training ML models with static data.…”

Cohort-based Federated Learning Services for Industrial Collaboration on the Edge

“…Focusing on edge-based FL, Feraudo et al [42] provide an architecture that enables asynchronous FL for edge devices using a publish/subscribe pattern. For this, the client registers for FL by sending a message to a message broker, which notifies the server, that waits until the end of a predefined timespan to consider the start of FL rounds.…”