The fog-assisted Internet-of-Things (IoT) is gaining interest due to its potential to cause duplicate data transmission. This paper proposes task distribution and secure deduplication over Cluster-based IoT using four layers: IoT Devices Layer, Fog Layer, Cloud Layer, and Service Layer. The IoT devices layer deploys devices to sense air pollutants, authenticates them to the cloud server, and uses Adaptive Rewards Optimized Deep Reinforcement Learning (ARO-DRL) for cluster-head selection. The fog layer uses SHA-3 for duplicate verification and the Emperor Penguin Optimization Algorithm for fog node selection. The Hybrid cloud environment protects organizations' data by combining private and public clouds. Experiments were conducted using NS3 with Java programming, and simulation results showed improvements in average latency, user satisfaction, network lifetime, energy consumption, and security strength.