1. The document proposes using autoencoder neural networks for anomaly detection in wireless sensor networks (WSN) in an Internet of Things (IoT) system. 2. It has minimal communication and computation requirements compared to other methods, and the distributed nature is well-suited for WSN. 3. The paper presents a two-part algorithm using a single hidden layer autoencoder model and evaluates its performance on temperature and humidity sensor data from a WSN testbed, showing it can accurately detect anomalies over 90% of the time.