Software-defined networking for the Internet of Things (SDN-IoT) has recently emerged as a promising to enhance IoT networks’ performance, reliability, and security. SDN-IoT can significantly improve network management by providing greater network traffic control and faster responses to evolving requirements or threats. However, the reliability of low-cost sensors that degrades over time, makes them susceptible to anomalies that can lead to produce erroneous sensed data. Detecting such anomalies in SDN-IoT networks is a challenging topic. This paper proposes a multi-timescale anomaly detection approach designed to monitor short-term, periodic, and long-term data patterns produced by sensors in SDN-IoT systems. This method is built on a modified and adapted version of previous work [1] to detect outliers in data. The proposed framework introduces a novel SDN-IoT Anomaly Detection System based on Piecewise Median Anomaly Detection Decomposition and the Generalized ESD outliers test. Additionally, it leverages P4 as a network programming language to deploy the anomaly detection mechanism across IoT gateways. Finally, preliminary results that evaluate the anomaly detection system’s impact on service quality parameters are presented, specifically transmission delay and throughput.