A Agent Learning Model For Service Composition In Mobile Cloud Computing

Abstract

Mobile cloud computing has challenges like limited resources, openness, and uncertainty, affecting service quality and security. This paper presents a three-layer trust-based service composition model using fuzzy evaluation to manage trust. It includes a learning module to understand user preferences, even with incomplete service requests. A prototype built on the JADE platform shows improved success rates and user satisfaction. Mobile cloud computing (MCC) offers significant opportunities for delivering scalable and flexible services, but it faces critical challenges, including limited computational resources, the open and dynamic nature of mobile environments, and inherent uncertainties that impact service quality, reliability, and security. To address these issues, this paper proposes a novel three-layer trust-based service composition model that leverages fuzzy logic-based evaluation to effectively manage trust in MCC systems. The proposed model consists of three distinct layers: a service discovery layer that identifies and filters available services based on functional and non-functional requirements; a trust evaluation layer that employs fuzzy logic to assess the trustworthiness of services by considering multiple trust attributes, such as reliability, security, and performance; and a service composition layer that optimally combines services to meet user demands while maximizing trust and quality of service (QoS). A key feature of the model is its integrated learning module, which utilizes machine learning techniques to analyse historical user interactions and infer user preferences, even in cases of incomplete or ambiguous service requests. This enhances the model's ability to deliver personalized and context-aware service compositions. To validate the proposed approach, a prototype was developed using the JADE (Java Agent Development Framework) platform, enabling agent-based service interactions in a simulated MCC environment. Experimental results demonstrate that the proposed model significantly improves the success rate of service composition, enhances user satisfaction, and outperforms existing approaches in terms of trust management and adaptability to dynamic mobile cloud environments. The model’s ability to handle uncertainty and resource constraints makes it a promising solution for advancing the reliability and efficiency of MCC systems.