– The Internet of Things is susceptible to seepage of private information during the data sharing. To circumvent this problem, access control and secure data sharing have been familiarized in cloud- IoT which is prevalent now days due to it storage and data access services. An intrusion detection implementation defends the veracity of the data sharing confining record are stored in a data base by recognizing when something changed unpredictably so that security and privacy are major anxieties. Therefore, it is crucial to develop the security apparatus that perceive and mitigate any feasible attack in cloud security services. In this paper, we proposed secure access control method with two-fold revocation for ensuring cloud computing security by hybrid blockchain (AI-Hybrid chain). Initially, all the users and services are registered their credentials to Secure Agent (SA) and provides image-based user authentication. Then, SA provides a long-term secret key which is generated by Improved Advanced Encryption Standard (IAES) algorithm, both user and service credentials are encrypted and stored in cloud chain. After that, gateway verifies the incoming requests are legitimate or illegitimate. For verifying the incoming packets, the proposed work utilizes Di-Fuzzy Inference System (Di-FIS) based on effective metrics. Further, legitimate request is optimally classified as sensitive and non-sensitive request for achieving secure service allocation using Tomtit Flock Optimization algorithm (TFO) based on optimal features. Once the service is allocated to users, we perform novel access control mechanism based on role and attribute of the users, the trust can be validated using Transfer Learning based Double Deep Q Network (TL-DDQN). Furthermore, virtual firewall is deployed in the cloud environment for analyzing the user storage request traffic as normal validated using Transfer Learning based Double Deep Q Network (TL-DDQN). Furthermore, virtual firewall is deployed in the cloud environment for analyzing the user storage request traffic as normal and malicious based on Contractive Residual Network (Con-ResNet). The implementation of proposed work is conducted by Network Simulator – 3.26 and the performance of the proposed AI-Hybrid chain model is itemized based on various performance metrics in terms of attack detection rate, overhead, success ratio, throughput, traffic rate, delay, response time and unauthorized access.