SNGPLDP: Social network graph generation based on personalised local differential privacy

The social network graph (SNG) can display valuable information. Its generation needs vast amounts of users' data. However, conflicts arise between generating the SNG and protecting the sensitive data therein. To balance it, some SNG generation schemes are proposed by using local differential privacy (LDP) techniques while they do not consider the personalised privacy requirements of users. This paper proposes an SNG generation scheme by designing a personalised LDP method, named SNGPLDP. Specifically, we develop a personalised randomised perturbation mechanism that satisfies ∈_total- PLDP to perturb users' private data. A seed graph creation mechanism and an optimised graph generation mechanism (OGGM) are then designed to generate and optimise the SNG with the perturbed data. Experiments performed on four real datasets show the effectiveness of SNGPLDP in providing PLDP protection with general graph properties. Moreover, the proposed scheme achieves higher network structure cohesion and supports stronger privacy protection than the advanced methods.