Research On Security Mechanism Of Cloud Computing Based On Hadoop

With the rapid development of Cloud computing, more and more users deposit thier data andapplication on the cloud. But the development of Cloud computing is hindered by the Cloudsecurity problem. Cloud computing has many characteristics, e.g. multi-user, virtualization,scalability and so on. Because of these new characteristics, traditional security technologies can notmake Cloud computing fully safe. Therefore, Cloud computing’s security becomes the currentresearch focus and is also this paper’s research direction.Firstly, the concept, characteristics, and solutions of Cloud computing are introduced andHadoop which is an open source Cloud computing system is manily introduced. On this basis, thispaer analyse the Cloud security problems from multiple angles, e.g. service mode, deploymentmode and makes a Cloud security problem cube (CSPC). Corresponding solutions of theseproblems are also introduced. This thesis comes up with a Cloud security model GCSM andtechnical architecture based on these problems and solutions, which provide support to achieveCloud security goal. Data remanence which is the weak of the Cloud security is also found fromthese.This paper improves the mechanism of deleting data in HDFS and comes up with amulti-grade safety data destruction mechanism of HDFS based on data overwrite algorithm, whichcan slove data remanence problem. This mechanism destroys data completely by overwritingoriginal data before deleting data. Moreover, the mechanism can balance the safety requirement andperformance requirement by using different overwrite algorithm according to different safetyrequirement. This thesis improves the heartbeat mechanism of HDFS and comes up with a newheartbeat mechanism, bidirectional heartbeat (BH). Based on BH, DSDBH which is dataself-destruction method of HDFS is designed to destroy data when the nodes are out of control. Thesimulation experiment shows that the improved mechanism can effectively overwrite the file ofBlock to destroy original data completely in HDFS. And the spending time of different overwritealgorithm is different, so it can keep efficiency and safety balance.