The above sections introduce the definitions and methods of two types of differential privacy: centralized and local. The essence of both approaches is to perturb data to ensure that the computation results satisfy the quantified privacy guarantees of differential privacy. However, perturbation inevitably impacts data usability. Local differential privacy (LDP), which perturbs data on the user’s side, introduces more noise, resulting in lower usability compared to centralized differential privacy (CDP), which only perturbs the final computation results. The usability of LDP methods is typically reduced by a factor of \(O(\sqrt{n})\) .