A Study on Privacy-Focused Machine Learning in IoT Networks via Federated Learning

The proliferation of Internet-connected Things (Iot)
devices has ushered in an era of unprecedented data generation
at the network’s edge. Leveraging this data for the artificial
intelligence applications presents privacy challenges. Federated
learning (FL) offers a viable solution. This study investigates the
implementation and enhancement of FL within IoT
environments, focusing on communication efficiency, model
aggregation, and device compatibility. The methodology is
grounded in analytical principles. employing deductive reasoning
and descriptive design. Secondary data is gathered from
published research and technical documentation. The findings
underscore the significance of secure communication protocols
like secure socket layer (SSL) for robust data encryption and
message Queue Telemetry Transport (MQTT) for efficient
messaging. Additionally, the paper examines how aggregation
strategies influence model convergence, with Federated
Averaging providing efficient convergence and secure
Aggregation ensuring anonymity when privacy is paramount.
Further, the research evaluates algorithm optimization
techniques- Including model pruning, Quantization, and
Lightweight Cognitive Architectures- that enhance model
performance on resource constrained IoT devices.