This research introduces a pioneering approach, termed Hierarchical Collaborative Botnet Detection, leveraging Federated Deep Learning to address the escalating security concerns within the Internet of Things (IoT) ecosystems characterized by heterogeneous devices. The proposed framework establishes a hierarchical structure facilitating efficient collaboration among devices at different levels, enabling scalable and distributed botnet detection. Federated Deep Learning ensures model training without centralizing sensitive data, respecting privacy constraints inherent in IoT environments. The methodology involves the development of a collaborative learning model capable of analyzing diverse data sources across the IoT landscape, utilizing the N-BaIoT dataset for comprehensive evaluation. Comprehensive simulations and experiments, conducted with the N-BaIoT dataset, showcase the robustness and efficiency of the proposed approach in detecting botnet activities across diverse IoT devices. Based on experimental results, the proposed method can identify botnets with an average accuracy of 98,97, precision of 98,75, recall of 99,41, and an F1-score of 99,11. The hierarchical and federated nature of the model contributes to a more resilient and scalable botnet detection system for large-scale IoT deployments, laying the foundation for a secure and collaborative IoT landscape in the face of evolving cyber threats. © 2024 ACM.