Improving Parameter Estimation and Defensive Ability of Latent Dirichlet Allocation Model Training Under Rényi Differential Privacy

Latent Dirichlet allocation(LDA)is a topic model widely used for discovering hidden semantics in massive text corpora.Collapsed Gibbs sampling(CGS),as a widely-used algorithm for learning the parameters of LDA,has the risk of privacy leakage.Specifically,word count statistics and updates of latent topics in CGS,which are essential for parameter estimation,could be employed by adversaries to conduct effective membership inference attacks(MIAs).Till now,there are two kinds of methods exploited in CGS to defend against MIAs:adding noise to word count statistics and utilizing inherent privacy.These two kinds of methods have their respective limitations.Noise sampled from the Laplacian distribution sometimes produces negative word count statistics,which render terrible parameter estimation in CGS.Utilizing inherent privacy could only provide weak guaranteed privacy when defending against MIAs.It is promising to propose an effective framework to obtain accurate parameter estimations with guaranteed differential privacy.The key issue of obtaining accurate parameter estimations when introducing differential privacy in CGS is making good use of the privacy budget such that a precise noise scale is derived.It is the first time that R′enyi differential privacy(RDP)has been introduced into CGS and we propose RDP-LDA,an effective framework for analyzing the privacy loss of any differentially private CGS.RDP-LDA could be used to derive a tighter upper bound of privacy loss than the overestimated results of existing differentially private CGS obtained byε-DP.In RDP-LDA,we propose a novel truncated-Gaussian mechanism that keeps word count statistics non-negative.And we propose distribution perturbation which could provide more rigorous guaranteed privacy than utilizing inherent privacy.Experiments validate that our proposed methods produce more accurate parameter estimation under the JS-divergence metric and obtain lower precision and recall when defending against MIAs.

A Review-Based Reputation Evaluation Approach for Web Services

Web services are commonly perceived as an environment of both offering opportunities and threats. In this environment, one way to minimize threats is to use reputation evaluation, which can be computed, for example, through transaction feedback. However, the current feedback-based approach is inaccurate and ineffective because of its inner limitations （e.g., feedback quality problem）. As the main source of feedback, the qualities of existing on-line reviews are often varied greatly from low to high, the main reasons include： （1） they have no standard expression formats, （2） dishonest comments may exist among these reviews due to malicious attacking. Up to present, the quality problem of review has not been well solved, which greatly degrades their importance on service reputation evaluation. Therefore, we firstly present a novel evaluation approach for review quality in terms of multiple metrics. Then, we make a further improvement in service reputation evaluation based on those filtered reviews. Experimental results show the effectiveness and efficiency of our proposed approach compared with the naive feedback-based approaches.

Partial Label Learning via Conditional-Label-Aware Disambiguation

Partial label learning is a weakly supervised learning framework in which each instance is associated with multiple candidate labels,among which only one is the ground-truth label.This paper proposes a unified formulation that employs proper label constraints for training models while simultaneously performing pseudo-labeling.Unlike existing partial label learning approaches that only leverage similarities in the feature space without utilizing label constraints,our pseudo-labeling process leverages similarities and differences in the feature space using the same candidate label constraints and then disambiguates noise labels.Extensive experiments on artificial and real-world partial label datasets show that our approach significantly outperforms state-of-the-art counterparts on classification prediction.