Semantic Cluster based Classification for Data Leakage Detection for the Cloud Security

novel approach for the data leak detection in the cloud environment is discussed in this paper. The paper uses the semantic based clustering for the anomaly detection to find the data leak. The Clustering is further used for the classification to add up for the semi supervised classification. After classification the threat patterns are stored in the database for further preventive actions in the data transmission. The necessary theory is discussed and the proposed approach is discussed with the results obtained. Keywordsleak prevention, semantic clustering, and semi supervised classification 1. INTRODUCTIONleakage is a problem which occurs intentionally or in an unintentional way. The intentional data leakage is the problem where the hackers try to disclose the data for fun or for any profit motivation. The unintentional damage is based on an accidental happening. Either intentional or unintentional act it involves in the loss of the data which leads to the loss and the reputation of the enterprise. This is common to every enterprise possessing the data and facing the trouble of leaking. It is also a global problem to think through since the data sharing is the mandate of the day for the communication. In the resource sharing enterprise which moves to the cloud computing for the distributed nature also severely suffer from the problem of data leakage. A survey says that by 2014 public IT services will exceed the traditional IT services by nearly five times (1). The forecasting based on the cloud computing predicts that worldwide revenue from the public IT cloud services will reach 55.5 billion dollars in 2014 (2). The cloud is prone to the data leakage because of its operational characteristics and its architecture, the chance is more because of the huge transactions which involves risks and challenges (3). The preventive mechanism is discussed in this paper which deploys the semantic clustering with the semi supervised classification for the updating of the threat history. The threat repository could be further used as the detection database to avoid the data breaches. This paper addresses the issue of proactive management through the learning from the previous data. The idea fills the gap of semantic based threat identification, which is lacking in the existing methods.