The rapid growth of Internet of Things (IoT) devices has brought about major security challenges, highlighting the need for strong intrusion detection systems to safeguard the networks from cyber-attacks. The proposed system aims to develop an intrusion detection mechanism for Internet of Things (IoT) devices by leveraging a combination of category boosting algorithms and deep learning techniques (DL). The system integrates local and global model training while preventing personal data transfer and reducing network latency. Attack types like denial of service and distributed denial of service (DoS) can be effectively detected due to the dual focus of enhancing security and improving local training for minimizing poisoning. The system is implemented in a smart home IoT environment, utilizing devices such as Raspberry Pi and laptops as gateways. Data packets are analyzed using models trained on the BoT-IoT dataset, with Cisco Packet Tracer used to record packet transfers. Authentication mechanisms are used to secure device communication. The proposed IDS employs class-balancing constraints to address overfitting, ensuring robustness and generalization. This multi-faceted model provides tremendous insight into the challenges facing IoT security and has the potential for broader use beyond academic research.