The advent of big data has significantly advanced the development of various applications and services by leveraging the extensive informational benefits of data publishing. However, the sensitive information contained within these data requires robust privacy protection prior to sharing and publishing. Effective privacy protection demands a clear definition of privacy standards and guarantees. While differential privacy (DP) and its variants are widely considered the gold standard for privacy preservation, their practical implementation often faces challenges that can lead to undesirable outcomes. This paper highlights the limitations of DP, such as the misuse of sequential composition, the presence of dishonest entities, and a privacy threat known as stigmatization, which falls outside the scope of DP’s guarantees. To address these challenges, we propose the \(\eta \) -inference model, which ensures information limited privacy and effectively mitigates the problems encountered with DP. Additionally, the \(\eta \) -inference model exhibits data incrementally invariant properties, making it particularly suitable for dynamic and distributed data publishing scenarios.