Edge computing has revolutionized distributed architectures, enabling workloads to be strategically positioned at the network’s edge, where data is generated and actions are initiated. This paradigm facilitates various applications, including the Internet of Things (IoT), smart grids, and smart homes, by providing cloud services directly to end-users. In particular, delay-sensitive applications benefit from the mobility, low latency, bandwidth, and location awareness offered by edge computing. Edge computing is an ideal platform for efficient data processing in smart homes. However, the security of edge gateways and nodes within smart home environments remains susceptible to security threats. This paper presents a novel approach employing a Deep Forest (DF)-based Intrusion Detection System (IDS) for Edge Intelligence. By embedding an IDS in edge nodes, this approach leverages edge computing to enhance security in smart home environments. The Deep Forest Classifier (DF) is innovative because it effectively addresses data imbalance in training datasets, a crucial factor in anomaly detection for binary-class classification. This integration enables real-time, low-latency threat detection and significantly improves security metrics, demonstrating substantial performance enhancements over traditional machine learning models. Experiments conducted on the UNSW-NB15 dataset demonstrate that the proposed binary classifier improves Accuracy by 26.24% over Naïve Bayes, 11.68% over Logistic Regression and 6.20% over KNN models.