The attractive features of cloud platforms such as low cost, high availability and scalability are encouraging social networks, health and other service providers to outsource their client data to the cloud. Though there are many advantages of using cloud-based solutions, the privacy of the outsourced data is a major concern. Compromised cloud servers can leak sensitive information about users such as the incident of the iCloud celebrity data leakage. One practical solution to mitigate these concerns is to encrypt or anonymize the data before outsourcing to the cloud. Although encryption protects the data from unauthorized access, it increases the computation complexity to execute the required functions (., similarity or nearest neighbour search), which is the key requirement for different social discovery applications. On the other hand, anonymization supports privacy-preserving fast computation but inefficient anonymization result huge data utility loss. In this thesis, I have designed an efficient approach to perform the secure nearest neighbour search in high dimensional space. The proposed framework utilizes the advantages of Intel Software Guard Extensions (Intel SGX) architecture and efficient anonymization methods to perform the secure nearest neighbour search operation.