Hybrid intrusion detection system with multi-factor dynamic task offloading in mobile cloud using RLnSRST-LSTM and CoAQ-FLS

Secure data transmission is enhanced in Mobile Cloud Computing (MCC) through dynamic offloading. However, prior studies have not fully considered multiple factors such as signal strength, Central Processing Unit (CPU) usage, memory, battery, and network bandwidth during offloading. This paper proposes a multi-factor-based Dynamic Task Offloading (DTO) mechanism using a Centroid of Area Q-learning-based Fuzzy Logic System (CoAQ-FLS). First, the mobile device is registered in the cloud, and a code is generated using the Chebyshev Polynomial Chaos Map-based Hash Message Authentication Code (CPCM-HMAC). After authentication, data is transferred, packet features are extracted using TShark, and suspicious activities are analyzed through Snort. The Local Model’s (LM) Intrusion Detection System (IDS) is trained using a Reinforcement Learning n-Step Return Sparsity Tikhonov-centric Long Short-Term Memory (RLnSRST-LSTM)-based attack classification. Data and model outputs are secured using Binet-Logistic Mapping Elliptic Curve Cryptography (BLM-ECC). Next, DTO results are recorded in the blockchain, ensuring authentication. Data from each LM is aggregated on the cloud server and processed by the Global IDS, and the result is transmitted back to the LM. A honeypot is implemented for counteracting attacks, and its information is securely recorded in the blockchain. The proposed system achieved a 99.13% intrusion.