Differential Privacy (DP) ensures strong privacy guarantees while enabling meaningful data analysis and operates in two main settings: Local DP (LDP) and Central DP (CDP). In LDP, noise is added directly at the data owner’s level before sending it to a central server. LDP provides strong privacy guarantees because each individual user ensures their data is private before sharing it. However, the decentralized nature of noise addition can significantly reduce data utility, as the noise addition is independent across data owners. CDP, on the other hand, achieves higher data utility by centralizing noise addition on a trusted server to ensure privacy. However, this reliance raises a critical question: How can we ensure that the server strictly adheres to the agreed privacy guarantees ( \( \epsilon , \delta \) ) without blindly trusting the server? In fact, a malicious server may manipulate ( \( \epsilon , \delta \) ), alter the DP mechanisms or clip the generated noise in order to improve result accuracy at the expense of privacy. zk-SNARKs (succinct Non-interactive Arguments of Knowledge) offer a solution by enabling servers to prove compliance with privacy guarantees without revealing sensitive information. However, randomness in DP complicates zk-SNARK proofs, as proving correctness without leaking information requires a careful protocol design. In this paper, we propose a framework that combines zk-SNARKs and cryptographic hashes in CDP. zk-SNARKs verify the correctness of noise generation and addition, while hashes ensure data integrity. Experimental results demonstrate that our framework achieves verifiable privacy compliance with practical performance and minimal overhead, providing a foundation for verifiable and trustworthy data privacy-preserving.