Search

The issue of how to improve the usability of data publishing under differential privacy has become one of the top questions in the field of machine learning privacy protection, and the key to solving this problem is to allocate a reasonable privacy protection budget. To solve this problem, we design a privacy budget allocation algorithm based on out-of-bag estimation in random forest. The algorithm firstly calculates the decision tree weights and feature weights by the out-of-bag data under differential privacy protection. Secondly, statistical methods are introduced to classify features into best feature set, pruned feature set, and removable feature set. Then, pruning is performed using the pruned feature set to avoid decision trees over-fitting when constructing an ϵ -differential privacy random forest. Finally, the privacy budget is allocated proportionally based on the decision tree weights and feature weights in the random forest. We conducted experimental comparisons with real data sets from Adult and Mushroom to demonstrate that this algorithm not only protects data security and privacy, but also improves model classification accuracy and data availability. [ABSTRACT FROM AUTHOR]

Since Coppersmith proposed the use of the LLL algorithm to solve univariate modular polynomial equations at EUROCRYPT'96, it has sparked a fervent research interest in lattice analysis among cryptographers. Despite its polynomial-time nature, the LLL algorithm exhibits a high-order polynomial upper bound in terms of theoretical complexity, particularly with longer computation times when applied to high-dimensional lattices. In addressing this issue, we propose an improved algorithm based on block preprocessing, building on the original Coppersmith algorithm and thus providing proof of correctness for this algorithm. This approach effectively reduces the solution time of the algorithm, offering a maximum improvement of 8.1% compared to the original Coppersmith algorithm. Additionally, we demonstrate the compatibility of our algorithm with the rounding algorithm proposed at PKC 2014. The combined utilization of these approaches further enhances the efficiency of our algorithm. The experimental results show that the combined algorithm achieves a maximum improvement of 22.4% in solution time compared to the original Coppersmith algorithm. It also outperforms the standalone rounding algorithm with a maximum improvement of 12.1%. When compared to the improved Coppersmith algorithm based on row common factor extraction, our proposed algorithm demonstrates comparable or even superior performance in certain dimensions. The block preprocessing algorithm in our approach enables independent execution without data exchange, making it suitable for leveraging multi-processing advantages in scenarios involving higher degrees of modular polynomial equations. This offers a new perspective for achieving the parallel computation of the Coppersmith algorithm, facilitating parallel execution and providing valuable insights. [ABSTRACT FROM AUTHOR]