Differential private collaborative Web services QoS prediction

Abstract

Collaborative Web services QoS prediction has proved to be an important tool to estimate accurately personalized QoS experienced by individual users, which is beneficial for a variety of operations in the service ecosystem, such as service selection, composition and recommendation. While a number of achievements have been attained on the study of improving the accuracy of collaborative QoS prediction, little work has been done for protecting user privacy in this process. In this paper, we propose a privacy-preserving collaborative QoS prediction framework which can protect the private data of users while retaining the ability of generating accurate QoS prediction. We introduce differential privacy, a rigorous and provable privacy model, into the process of collaborative QoS prediction. We first present DPS, a method that disguises a user's observed QoS values by applying differential privacy to the user's QoS data directly. We show how to integrate DPS with two representative collaborative QoS prediction approaches. To improve the utility of the disguised QoS data, we present DPA, another QoS disguising method which first aggregates a user's QoS data before adding noise to achieve differential privacy. We evaluate the proposed methods by conducting extensive experiments on a real world Web services QoS dataset. Experimental results show our approach is feasible in practice.

Threshold‐oblivious on‐line web QoE assessment using neural network‐based regression model

“…Zhang et al [35] proposed a long-term QoS forecasting approach based on neural networks, but their work focused on long-term QoS prediction. Liu et al [36] proposed a collaborative QoS prediction framework for privacy preservation, which could guarantee the accurate QoS prediction while protecting users' privacy.…”

An L 1-and-L 2-Norm-Oriented Latent Factor Model for Recommender Systems

“…where p u,d and q i,d denote specific entries in P and Q, w u,i is given by w u,i =μ+b u +b i as μ denotes the global-average of R K , b u denotes the observed deviations on user u, and b i denotes the observed deviations on item I, respectively. Note that in (7), the regularization effect on each LF is specified with its relevant known rating count [17,19], thereby implementing a finely-grained control of the regularization effects. With (7), an L 3 F model reasonably aggregates the merits of L 1 norm-based and L 2 norm-oriented Losses, i.e., model robustness and stability.…”

“…Recommender system (RS) is highly useful in addressing this issue [5,6]. So far, various approaches are proposed to implement an RS, where collaborative filtering (CF) is highly popular [7][8][9][10][11][12][13].…”

Spatio-temporal top-k term search over sliding window

“…Third, we may also study how to integrate textual data to routing problems [32, 33, 40-42, 52, 56, 68] as well as spatio-textual routing problem. Fourth, we may study how to integrate streaming data sampling methods and pattern analysis [12,13,16,17,50,53,54,58,67] integrating with spatio-textual data.…”