4,239 Hits in 3.9 sec

Trustworthy AI: A Computational Perspective [article]

Haochen Liu, Yiqi Wang, Wenqi Fan, Xiaorui Liu, Yaxin Li, Shaili Jain, Yunhao Liu, Anil K. Jain, Jiliang Tang
2021 arXiv   pre-print
In this work, we focus on six of the most crucial dimensions in achieving trustworthy AI: (i) Safety & Robustness, (ii) Non-discrimination & Fairness, (iii) Explainability, (iv) Privacy, (v) Accountability  ...  For each dimension, we review the recent related technologies according to a taxonomy and summarize their applications in real-world systems.  ...  Graph Neural Networks (GNNs) have been developed for graph-structured data and can be used by many real-world systems, such as social networks and natural science.  ... 
arXiv:2107.06641v3 fatcat:ymqaxvzsoncqrcosj5mxcvgsuy

Link Prediction using Graph Neural Networks for Master Data Management [article]

Balaji Ganesan, Srinivas Parkala, Neeraj R Singh, Sumit Bhatia, Gayatri Mishra, Matheen Ahmed Pasha, Hima Patel, Somashekar Naganna
2020 arXiv   pre-print
Predicting links between people using Graph Neural Networks requires careful ethical and privacy considerations than in domains where GNNs have typically been applied so far.  ...  We introduce novel methods for anonymizing data, model training, explainability and verification for Link Prediction in Master Data Management, and discuss our results.  ...  Explainability Explainability methods in Graph Neural Networks tend to follow similar methods in text and images, namely identifying features that are most significant for the predictions.  ... 
arXiv:2003.04732v2 fatcat:qfak6f4265gerl7yvj36nbl444

Quantifying Privacy Leakage in Graph Embedding [article]

Vasisht Duddu, Antoine Boutet, Virat Shejwalkar
2021 arXiv   pre-print
For the first time, we quantify the privacy leakage in graph embeddings through three inference attacks targeting Graph Neural Networks.  ...  Graph embeddings have been proposed to map graph data to low dimensional space for downstream processing (e.g., node classification or link prediction).  ...  The blackbox setting considers the specific case of downstream node classification task for convolution kernel based graph embedding with neural network.  ... 
arXiv:2010.00906v2 fatcat:hqtdvzxncnbmpdx5sznqafecnu

Quantifying Privacy Leakage in Graph Embedding

Vasisht Duddu, Antoine Boutet, Virat Shejwalkar
2020 MobiQuitous 2020 - 17th EAI International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services  
For the first time, we quantify the privacy leakage in graph embeddings through three inference attacks targeting Graph Neural Networks.  ...  Graph embeddings have been proposed to map graph data to low dimensional space for downstream processing (e.g., node classification or link prediction).  ...  Deep Learning and more precisely Convolutional Neural Networks have shown tremendous performance over non-graph data such as images by capturing the spatial relation between pixels of an image and extracting  ... 
doi:10.1145/3448891.3448939 fatcat:gvgughaumrhkzo2pkoyfymklrm

Personalized Advertising Computational Techniques: A Systematic Literature Review, Findings, and a Design Framework

Iosif Viktoratos, Athanasios Tsadiras
2021 Information  
Finally, a design framework for personalized advertisement systems has been designed based on these findings.  ...  are highlighted and are pinpointed to help and inspire researchers for future work.  ...  and probability-based (Li and Lien, 2009) [82] Neural network (artificial neural network ANN), social graph-based (Qiu et al., 2009) [168] Weight-based and distance functions (sentiment analysis, tf-idf  ... 
doi:10.3390/info12110480 fatcat:53wmlsdlp5bmhffnv2fjst5vtq

TransMIA: Membership Inference Attacks Using Transfer Shadow Training [article]

Seira Hidano, Takao Murakami, Yusuke Kawamoto
2021 arXiv   pre-print
In this paper, we propose TransMIA (Transfer learning-based Membership Inference Attacks), which use transfer learning to perform membership inference attacks on the source model when the adversary is  ...  However, no prior work has pointed out that transfer learning can strengthen privacy attacks on machine learning models.  ...  Acknowledgment This work was supported by JSPS KAKENHI Grant Number JP19H04113, and by ERATO HASUO Metamathematics for Systems Design Project (No. JPMJER1603), JST.  ... 
arXiv:2011.14661v3 fatcat:kgfsrc7k6jhy5hm35jcracvdp4

Trustworthy Graph Neural Networks: Aspects, Methods and Trends [article]

He Zhang, Bang Wu, Xingliang Yuan, Shirui Pan, Hanghang Tong, Jian Pei
2022 arXiv   pre-print
Graph neural networks (GNNs) have emerged as a series of competent graph learning methods for diverse real-world scenarios, ranging from daily applications like recommendation systems and question answering  ...  In this survey, we introduce basic concepts and comprehensively summarise existing efforts for trustworthy GNNs from six aspects, including robustness, explainability, privacy, fairness, accountability  ...  In this survey, the former kind of GNNs are called interpretable graph neural networks, and the latter kind of methods for explainability of GNNs are called explainers for graph neural networks.  ... 
arXiv:2205.07424v1 fatcat:f3iul7soqvgzbgaeqw7nhypbju

2020 Index IEEE Transactions on Knowledge and Data Engineering Vol. 32

2021 IEEE Transactions on Knowledge and Data Engineering  
., +, TKDE July 2020 1378-1392 Directed graphs DAG: A General Model for Privacy-Preserving Data Mining.  ...  ., +, TKDE Jan. 2020 188-202 Convolutional neural nets Flow Prediction in Spatio-Temporal Networks Based on Multitask Deep Learning.  ... 
doi:10.1109/tkde.2020.3038549 fatcat:75f5fmdrpjcwrasjylewyivtmu

On the Privacy Risks of Model Explanations [article]

Reza Shokri, Martin Strobel, Yair Zick
2021 arXiv   pre-print
We extensively evaluate membership inference attacks based on feature-based model explanations, over a variety of datasets.  ...  We investigate the privacy risks of feature-based model explanations using membership inference attacks: quantifying how much model predictions plus their explanations leak information about the presence  ...  This is based on the method of Shokri et al. [35] , who formulate membership inference as a learning problem for the attacker and train a neural network to predict membership.  ... 
arXiv:1907.00164v6 fatcat:jn7ju6gzpvevffguye2sctr2iu

A Survey of Trustworthy Graph Learning: Reliability, Explainability, and Privacy Protection [article]

Bingzhe Wu, Jintang Li, Junchi Yu, Yatao Bian, Hengtong Zhang, CHaochao Chen, Chengbin Hou, Guoji Fu, Liang Chen, Tingyang Xu, Yu Rong, Xiaolin Zheng (+8 others)
2022 arXiv   pre-print
to robustness, explainability, and privacy.  ...  In this survey, we provide a comprehensive review of recent leading approaches in the TwGL field from three dimensions, namely, reliability, explainability, and privacy protection.  ...  Trustworthy Graph Learning Accuracy Reliability Explainability Privacy Protection Recent few years have seen deep graph learning (DGL) based on graph neural networks (GNNs) making remarkable progress in  ... 
arXiv:2205.10014v2 fatcat:aobv34rwg5ehpka4fsuar2gm7i

LPGNet: Link Private Graph Networks for Node Classification [article]

Aashish Kolluri, Teodora Baluta, Bryan Hooi, Prateek Saxena
2022 arXiv   pre-print
In this paper, we present a new neural network architecture called LPGNet for training on graphs with privacy-sensitive edges.  ...  Deep neural networks are increasingly being used for node classification on graphs, wherein nodes with similar features have to be given the same label.  ...  INTRODUCTION Graph neural networks (GNN) learn node representations from complex graphs similar to how convolutional neural networks do from grid-like images.  ... 
arXiv:2205.03105v1 fatcat:np4psn4nofbczpobr2cttqw5iu

Secure Image Inference using Pairwise Activation Functions

Jonas T. Agyepong, Mostafa Soliman, Yasutaka Wada, Keiji Kimura, Ahmed El-Mahdy
2021 IEEE Access  
Polynomial approximation has for the past few years been used to derive polynomials as an approximation to activation functions for use in image prediction or inference employing homomorphic encryption  ...  INDEX TERMS Exploratory analysis, homomorphic encryption scheme, homomorphic image inference, pairwise functions, polynomial approximation, privacy-preserving machine learning.  ...  the neural network for image classification.  ... 
doi:10.1109/access.2021.3106888 fatcat:5jqu6yjkl5hb3l2igp5e6a3nim

Adversary for Social Good: Protecting Familial Privacy through Joint Adversarial Attacks

Chetan Kumar, Riazat Ryan, Ming Shao
For example, implicit social relation such as family information may be simply exposed by network structure and hosted face images through off-the-shelf graph neural networks (GNN), which will be empirically  ...  Second, to protect family privacy on social networks, we propose a novel adversarial attack algorithm that produces both adversarial features and graph under a given budget.  ...  what additional information will be inferred from the social networks.  ... 
doi:10.1609/aaai.v34i07.6791 fatcat:lspxwtunjfhpjjov2vao63rbgu

Membership Inference Attack on Graph Neural Networks [article]

Iyiola E. Olatunji, Wolfgang Nejdl, Megha Khosla
2021 arXiv   pre-print
Graph Neural Networks (GNNs), which generalize traditional deep neural networks on graph data, have achieved state-of-the-art performance on several graph analytical tasks.  ...  We introduce two realistic settings for performing a membership inference (MI) attack on GNNs.  ...  We will publish our code at the time of publication. 2 Background and Related Works Graph Neural Networks Graph Neural Networks popularized by graph convolutional networks (GCNs) and their variants,  ... 
arXiv:2101.06570v3 fatcat:czknpvcdsvdwtkdlcrbk37dvdm

Cross-domain fault localization: A case for a graph digest approach

William D. Fischer, Geoffrey G. Xie, Joel D. Young
2008 2008 IEEE Internet Network Management Workshop (INM)  
We present an inference-graph-digest based formulation of the problem.  ...  The formulation not only explicitly models the inference accuracy and privacy requirements for discussing and reasoning over cross-domain problems, but also facilitates the re-use of existing fault localization  ...  ACKNOWLEDGMENTS We thank the anonymous reviewers and our shepherd Keisuke Ishibashi for their constructive comments. Srikanth Kandula provided valuable input for our background research.  ... 
doi:10.1109/inetmw.2008.4660328 fatcat:4xjuzwoqfrfv7l7hiit5z2vhiy
« Previous Showing results 1 — 15 out of 4,239 results