Abstract

With the growing use of cloud-based educational solutions, it is now possible to collect and analyse large volumes of data for personalized learning, performance assessment, and institutional decision-making. But, when sensitive educational information is outsourced and put in third-party cloud computing systems, there are significant issues of privacy and security, including the possibility of unauthorized access and data leakage during computation. While traditional encryption techniques work well for data at rest and in transit, they are not suitable for compute operations on encrypted data, thereby limiting their applicability to privacy-preserving analytics. To secure the outsourcing of educational data analytics, this paper presents a secure framework for outsourcing educational data analytics by using homomorphic encryption (HE) schemes to perform direct computation on encrypted datasets without revealing underlying sensitive information. The approach combines a partially homomorphic encryption scheme for linear statistical operations and an optimized levelled homomorphic encryption model for complex analytical operations like regression and classification. The proposed framework has been shown to provide a computation overhead of 37.5% with a privacy preservation rate of 98.3%. The system reduces the latency of the end-to-end processing by some 18-25% through optimized ciphertext operations and parallelization in the cloud. Furthermore, the accuracy of the analytical models is almost the same as that of the plaintext calculation, and the deviation is within 1.5%, which shows that the encrypted calculation is correct. The study concludes that although there is a challenge of computational overhead, proper selection of scheme and optimization strategies can be implemented to improve efficiency without compromising on the confidentiality of the data, thereby making it suitable for privacy-sensitive applications in the education domain in a cloud environment.