Blockchain-based distributed cloud deployments  generally use static methods to model encryption, hashing,  consensus and sharding techniques. Existing dynamic security  models for blockchains either have higher complexity, or lower  performance efficiency when applied to real-time deployments. To  overcome these issues, this text proposes design of an incremental  bioinspired-learning model for integrating dynamic security in  blockchain-based distributed cloud deployments. The proposed  model uses Q-Learning to dynamically select different encryption  & hashing. This assists in dynamically modifying security  performance for different data-level attacks. Model proposes use  of a novel Proof-of-Distributed-Cloud-Performance-Trust  (PoDCPT), which integrates temporal execution performance of  distributed Virtual Machines (VM) for selection of optimal miner  nodes during consensus. These miner nodes are selected based on  computational delay, execution efficiency of VMs, and their  temporal mining performance under different attacks. The  selected miner nodes assist in adding new blocks to the sharded  chains. Sharding configurations of these chains are selected using hybrid Ant Lion Optimization (ALO) with Teacher-Learning?based-Optimization (TLbO) process. This hybrid combination  assists in formation of delay & complexity-aware sidechain  configurations. This enables the model to reduce computational  delay by 3.5%, reduce energy consumption by 5.4%, improve  throughput by 8.3%, while enhancing mining efficiency by 4.5%  under different cloud attacks.