This work designed and implemented a security framework for running MapReduce tasks across different clusters in a distributed environment. The security framework provides users with a single-sign-on process to submit jobs to G-Hadoop. In addition, it applies various security mechanisms to protect the G-Hadoop system from attacks as well as abusing or misusing. These security mechanisms are based on some current security solutions, for example SSL and cryptographic algorithm, or the concepts of other security solutions such as GSI. Some concepts, for example, proxy credentials, user session, and user instance, are applied in this security framework as well to provide the functionalities of the framework.
With these security mechanisms the designed security framework has the ability to prevent the most common attacks, such as MITM attack, replay attack, and delay attack, and ensures a secure communication of G-Hadoop over public net- works. In addition, it adopts different mechanisms to protect the resources of G-Hadoop from abusing or misusing. On the whole, it provides a trustful and complete solution of the single-sign-on process for the user to access G-Hadoop. For further improvement, job execution in Phase V as well as the encryption algorithms and keys will be designed as changeable to increase the difficulty of cryptographic analysis by an attacker.