A5100378902- Privacy-Preserving Technologies in Data
- Privacy, Security, and Data Protection
- Algorithms and Data Compression
- Mobile Crowdsensing and Crowdsourcing
- Network Packet Processing and Optimization
- Geochemistry and Geologic Mapping
- Advanced Graph Neural Networks
- Human Mobility and Location-Based Analysis
- Natural Language Processing Techniques
- Geochemistry and Geochronology of Asian Mineral Deposits
- Stochastic Gradient Optimization Techniques
- Complex Network Analysis Techniques
- Geological and Geochemical Analysis
- Genomics and Phylogenetic Studies
- Cryptography and Data Security
- Vehicular Ad Hoc Networks (VANETs)
- Ethics and Social Impacts of AI
- Recommender Systems and Techniques
- Blockchain Technology Applications and Security
- Geoscience and Mining Technology
- Internet Traffic Analysis and Secure E-voting
- Speech and dialogue systems
- Topic Modeling
- Speech Recognition and Synthesis
- DNA and Biological Computing
Hong Kong Polytechnic University
2022-2025
Sun Yat-sen University
2024
Southeast University
2019-2023
Ministry of Education of the People's Republic of China
2021-2022
SUNY Oneonta
2007-2014
Purchase College
2008
We present, in this paper, two efficient algorithms for linear time suffix array construction. These achieve their complexities, using the techniques of divide-and-conquer, and recursion. What distinguish proposed from other construction (SACAs) are variable-length leftmost S-type (LMS) substrings fixed-length d-critical sampled problem reduction, simple sorting these substrings: induced algorithm LMS radix substrings. The very mechanisms render our an elegant design framework, and, turn,...
The objective of privacy-preserving synthetic graph publishing is to safeguard individuals' privacy while retaining the utility original data. Most existing methods focus on neural networks under differential (DP), and yet two fundamental problems in generating graphs remain open. First, current research often encounters high sensitivity due intricate relationships between nodes a graph. Second, DP usually achieved through advanced composition mechanisms that tend converge prematurely when...
Graph embedding generation techniques aim to learn low-dimensional vectors for each node in a graph and have recently gained increasing research attention. Publishing enables various analysis tasks, such as structural equivalence link prediction. Yet, improper publication opens backdoor malicious attackers, who can infer sensitive information of individuals from the vectors. Existing methods tackle this issue by developing deep learning models with differential privacy (DP). However, they...
We present a new suffix array construction algorithm that aims to build, in external memory, the for an input string of length n measured magnitude tens Giga characters over constant or integer alphabet. The core this is adapted from framework original internal memory SA-DS samples fixed-size d-critical substrings. This external-memory algorithm, called EM-SA-DS, uses novel cache data structures construct sequential scanning manner with good spatial locality: read written disk sequentially....
Graph embedding maps a graph into low-dimensional vectors, i.e., matrix, while preserving the structure, solving high computation and space cost for analysis. Matrix factorization (MF) is an effective means to achieve since maintaining utility of structure. The personalized structure features implied in matrix can identify individual, which potentially breaches individual sensitive information original graph. Currently, protecting privacy without compromising key sharing matrix. Differential...
The goal of privacy-preserving social graph publishing is to protect individual privacy while preserving data utility. Community structure, which an important global pattern nodes, a crucial utility as it serves fundamental operations for many analysis tasks. Yet, most existing methods with differential (DP) commonly fall in edge-DP sacrifice security exchange Moreover, they reconstruct graphs from the local feature-extraction resulting poor community preservation. Motivated by this, we...
Let S be an n-character string terminated with unique smallest sentinel, its suffix array SA(S) is of pointers for all the suffixes in sorted lexicographically ascending order. Specially, Burrows-Wheeler transform building efficient compression solutions can quickly computed by fast sorting based on construction algorithms (SACAs). The existing well-known practical linear SACAs are those two contemporarily reported 2003 Karkkainen and Sanders (KS) (J. Karkkaiinen P. Sanders, 2003) Ko Aluru...
The goal of privacy-preserving social graph publishing is to protect individual privacy while preserving utility. Node nearest neighbor structure a crucial utility as it the basis for many analysis tasks. Most existing methods with differential focus on degree distribution yet neglect maintenance connections between nodes' neighbors. Moreover, they require massive noise added mask change single edge, thereby rendering poor structure. As result, tough preserve high under privacy. To tackle...
Many real-world networks can be represented as weighted graphs, where weights represent the closeness or importance of relationships between node pairs. Sharing these graphs is beneficial for many applications while potentially leading to privacy breaches. Variants deep learning approaches have been developed synthetic graph publishing, but privacy-preserving (especially graph) publishing has not fully addressed. To bridge this gap, we propose WDP-GAN, a generative adversarial network (GAN)...
Supplying data augmentation to conversational question answering (CQA) can effectively improve model performance. However, there is less improvement from single-turn datasets in CQA due the distribution gap between and multi-turn datasets. On other hand, while numerous are available, we have not utilized them effectively. To solve this problem, propose a novel method convert The proposed consists of three parts, namely, QA pair Generator, Reassembler, Rewriter. Given sample consisting...
The goal of privacy-preserving social graph publishing is to protect individual privacy while preserving data utility. Community structure, which an important global pattern nodes, a crucial utility as it serves fundamental operations for many analysis tasks. Yet, most existing methods with differential (DP) commonly fall in edge-DP sacrifice security exchange Moreover, they reconstruct graphs from the local feature-extraction resulting poor community preservation. Motivated by this, we...