Tremendous growth in the availability of social media data and its collection and storage is moving towards the big data problems. Cloud-based data centres have adopted replication control mechanisms for the data distribution strategies to handle such huge volumes of data. These methods increase the query complexity and chance of the attacks. Parallel researches demonstrated that the large scale data demands in time complexity reduction and early attack detection. Most of the attacks depend on the network and application characteristics. Online personal information tracks deep insights about how people assemble numbers, rates, and attracted towards a particular substance or channel. This information of the Internet users can be gathered easily from their daily based activities over the Internet. However, the popular social media applications such as Twitter do not primarily signify the attacks or a chance of potential attack characteristics. With the added security concerns, it has been observed that the parallel research attempts have failed to justify the time complexity. Hence, this research proposes two novel algorithms as firstly, a secure distributed large volume data query algorithm using novel secure data discovery technique and secondly, a cognitive attack detection method using data characteristics analysis by deploying machine learning method. The final outcome of the research is to build a novel automated framework to detect the data centre clusters, which are potentially under attack.