End-to-end security in service-oriented architecture

A service-oriented architecture (SOA)-based application is composed of a number of distributed and loosely-coupled web services, which are orchestrated to accomplish a more complex functionality. Any of these web services is able to invoke other web services to offload part of its functionality. The main security challenge in SOA is that we cannot trust the participating web services in a service composition to behave as expected all the time. In addition, the chain of services involved in an end-to-end service invocation may not be visible to the clients. As a result, any violation of client’s policies could remain undetected. To address these challenges in SOA, we proposed the following contributions. First, we devised two composite trust schemes by using graph abstraction to quantitatively maintain the trust levels of different services. The composite trust values are based on feedbacks from the actual execution of services, and the structure of the SOA application. To maintain the dynamic trust, we designed the trust manager, which is a trusted-third party service. Second, we developed an end-to-end inter-service policy monitoring and enforcement framework (PME framework), which is able to dynamically inspect the interactions between services at runtime and react to the potentially malicious activities according to the client’s policies. Third, we designed an intra-service policy monitoring and enforcement framework based on taint analysis mechanism to monitor the information flow within services and prevent information disclosure incidents. Fourth, we proposed an adaptive and secure service composition engine (ASSC), which takes advantage of an efficient heuristic algorithm to generate optimal service compositions in SOA. The service compositions generated by ASSC maximize the trustworthiness of the selected services while meeting the predefined QoS constraints. Finally, we have extensively studied the correctness and performance of the proposed security measures based on a realistic SOA case study. All experimental studies validated the practicality and effectiveness of the presented solutions