Recently, cloud computing services have found widespread use. Offloading computations to public clouds has many benefits albeit harming the privacy of users and data. Homomorphic encryption (HE) facilitates cloud computing services that can do computations over encrypted data in the cloud without requiring decryption and this enables privacy-preserving applications. In this work, it is proposed to offload delay-tolerant tasks of the Network Data Analytics Function (NWDAF) which plays a key role in autonomous network management and security in 5G networks using heuristic and machine learning (ML) techniques. However, not all methods such as deep neural networks (DNN), and complex analysis algorithms can be deployed in the core network due to their high computational requirements. Besides, telecommunication companies may opt to outsource the analytics to experienced third-party providers rather than implementing it themselves. By offloading Network Data Analytics Function (NWDAF) tasks to the cloud, telecommunication companies can concentrate on expanding their core capabilities, and reduce their capital and operational expenditures.
To support the proposed idea, two of the Network Data Analytics Function (NWDAF) use cases are selected and it is shown that homomorphic encryption can be employed for preserving the data privacy in both use cases. Selected use cases are user equipment (UE) abnormal behavior or anomaly detection, and user equipment (UE) mobility analytics and expected behavior prediction. For the former use case, various DL models are trained with plaintext data and it is demonstrated that inferencing on encrypted data is as accurate and precise as plaintext inferencing. For the latter use case, an approach for confidentially finding islands (connected components) in a graph is proposed and shown that how this approach can be used for mobility analytics and expected behavior prediction. Various performance evaluation results are quantified and it shows that privacy-preservation can be achieved with a cost of computation overhead.