<p>Protecting privacy in data mining has become a critical issue due to the increasing capabilities of data storage and analysis, particularly in domains involving personal data such as healthcare, banking, and commerce. Many techniques are used to protect sensitive data in data mining. However, these techniques often result in substantial information loss and reduced data utility. Therefore, a key challenge in privacy-preserving data mining (PPDM) is to develop techniques that can hide sensitive information with minimal impact on the original data. This paper introduces a new approach for protecting sensitive itemsets in association rule mining with transactions modifications rather transaction deletion. The proposed approach minimizes the impact on the original dataset while selectively hiding sensitive itemsets depending on the strategies of the gray wolf optimization (GWO) algorithm. The proposed approach introduces a novel algorithm termed GWOHSI (GWO for hiding sensitive itemset) to identify and hide sensitive itemsets with minimal side effects. The goal of this algorithm is to determine the optimal number of transactions should be modified for each item that contributes to sensitive itemsets, thus reducing their support to below the minimum support threshold. Comprehensive experiments are conducted to evaluate the performance of the proposed approach in terms of hiding failure, number of non-sensitive itemsets affected, data dissimilarity and execution time. Four datasets were used for evaluation, the results showed that the proposed approach effectively hides all sensitive itemsets (achieved a 100% hiding ratio), while minimizing the affected non-sensitive itemsets. The execution time of GWOHSI algorithm is considered satisfactory, and it is consistent across almost all datasets.</p>