Abstract

Blockchain-assisted cloud computing has emerged as a promising approach for ensuring data integrity and transparency; however, workflow scheduling within such environments remains complex due to resource heterogeneity, task dependencies, and latency constraints. This paper proposes a blockchain-based Dense Resolution Dilated Attention Forward Harmonic Network (Blockchain_DRDAFHNet) for secure, energy-efficient, multi-objective workflow scheduling in the cloud. The system model integrates a Task Pool Coordinator with blockchain miners and Virtual Machines (VMs) to anonymize and execute tasks securely via smart contracts. Scheduling is performed using two categories of parameters: VM parameters (CPU and bandwidth utilization) and task parameters (resource utilization, actual task running time, memory, makespan equivalent of total cost, reliability, task priority, Earliest Start Time, Earliest Finish Time, task length, trust, security level, security overhead, encryption speed, predicted energy via EA-LSTM, and secured data size). The proposed DRDAFHNet integrates a Dense Resolution Network (DRNet) with a Pyramid Dilated Attention Network (PDAN), with layers refined through forward harmonic analysis. Experiments are conducted in CloudSim using Python 3.9.11 with TensorFlow/Keras, across configurations of 10 PM/20 VM, 10 PM/40 VM, and 15 PM/50 VM. Under the 15 PM/50 VM configuration, Blockchain_DRDAFHNet achieves the best residual energy of 0.831 J, resource utilization of 0.576, and makespan of 0.546, outperforming seven existing methods including ALPSO, Actor-Critic, CGA, 2SD-GAT, HGALO-GOA, DNN-IGW-HHO, and RL-MOTS. These results demonstrate the model's robustness, security integration, and suitability for large-scale dynamic cloud scheduling.