SecureNLP: A System for Multi-Party Privacy-Preserving Natural Language Processing

Feng, Qi; He, Debiao; Liu, Zhe; Wang, Huaqun; Choo, Kim‐Kwang Raymond

doi:10.1109/tifs.2020.2997134

Cited by 65 publications

(13 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to solve the above problems, some scholars use encryption methods such as homomorphic encryption [2][3][4][5] and secure multiparty computing [6][7][8][9] to encrypt model parameters sent by participants to resist model inference attacks. Another part of scholars proposed the method of differential privacy [10,11], which makes it difficult for attackers to infer the original model parameters by adding disturbance noise to the model parameters. In order to further improve privacy security, Qu et al [12] designed a hybrid identity mechanism based on digital signature and encryption protocol to prevent attackers from stealing data information stored in the blockchain.…”

Section: Introductionmentioning

confidence: 99%

AFLPC: An Asynchronous Federated Learning Privacy-Preserving Computing Model Applied to 5G-V2X

Huang

et al. 2022

Security and Communication Networks

View full text Add to dashboard Cite

Federated learning can effectively protect local data privacy in 5G-V2X environment and ensure data protection in Internet of vehicles environment. The advantages of low delay of 5G network should be better utilized in the vehicle-road cooperative system. But the existing asynchronous federated learning obtains a local model through different node training and completes the update of the global model through the central server. There are problems such as single point of failure, privacy leakage, and deviation of aggregation parameters. In response to the above problems, we proposed a 5G-V2X-oriented asynchronous federated learning privacy-preserving computing model (AFLPC). We used an adaptive differential privacy mechanism to reduce noise while protecting data privacy. A weight-based asynchronous federated learning aggregation update method is proposed to reasonably control the proportion of parameters submitted by users with different training speeds in the aggregation parameters and actively update the aggregation parameters of lagging users, so as to effectively reduce the negative impact on the model caused by the different speed of finding you. Experiments show that the proposed method can effectively ensure the credibility and privacy of asynchronous federated learning in 5G-V2X scenarios and at the same time improve the utility of the model.

show abstract

Section: Introductionmentioning

confidence: 99%

AFLPC: An Asynchronous Federated Learning Privacy-Preserving Computing Model Applied to 5G-V2X

Huang

et al. 2022

Security and Communication Networks

View full text Add to dashboard Cite

show abstract

“…In the broader context of NLP, secure multiparty computation (Feng et al, 2020) and homomorphic encryption (Al Badawi et al, 2020) have been used to provide strong privacy guarantees. Since these cryptographic methods incur high performance penalties (see (Riazi et al, 2019) for an overview of their performance in deep learning), more recent proposals have focused on the careful use of simpler cryptographic primitives while training a model over encrypted text due to confidentiality reasons.…”

Section: Introductionmentioning

confidence: 99%

Multilingual Sentence-Level Bias Detection inWikipedia

Aleksandrova¹,

Lareau²,

Ménard³

2019

Proceedings - Natural Language Processing in a Deep Learning World

View full text Add to dashboard Cite

We propose a multilingual method for the extraction of biased sentences from Wikipedia, and use it to create corpora in Bulgarian, French and English. Sifting through the revision history of the articles that at some point had been considered biased and later corrected, we retrieve the last tagged and the first untagged revisions as the before/after snapshots of what was deemed a violation of Wikipedia's neutral point of view policy. We extract the sentences that were removed or rewritten in that edit. The approach yields sufficient data even in the case of relatively small Wikipedias, such as the Bulgarian one, where 62k articles produced 5k biased sentences. We evaluate our method by manually annotating 520 sentences for Bulgarian and French, and 744 for English. We assess the level of noise and analyze its sources. Finally, we exploit the data with well-known classification methods to detect biased sentences. Code and datasets are hosted at https://github.com/ crim-ca/wiki-bias.

show abstract

“…Therefore, when the exchange of documents among the members of these computation scenario is not an advisable option, multi-party computation (MPC) can be successfully used. MPC is a cryptographic primitive that computes aggregated functions over multiple sources of data, which cannot be revealed [23]. Formally, MPC assumes a set of inputs { 1 , 2 , .…”

Section: Solutionsmentioning

confidence: 99%

“…Several privacy-preserving mechanisms have been developed to safeguard personal data from threats, as attacks, disclosures, and unintended usages. Some of these mechanisms, such as encryption [19,30], differential privacy (DP) [24], multi-party computation (MPC) [23], and federated learning (FL) [25,31], are implemented alongside models for enabling applications to safeguard data privacy. There are also some attacking methods that aim at retrieving data samples used to train models, mostly neural networks, which can be seen as a threat or a safety checker, depending on the attacker's intention.…”

Section: Introductionmentioning

confidence: 99%

Privacy in Open Search: A Review of Challenges and Solutions

Sousa,

Guetl,

Kern

2021

Preprint

View full text Add to dashboard Cite

Privacy is of worldwide concern regarding activities and processes that include sensitive data. For this reason, many countries and territories have been recently approving regulations controlling the extent to which organizations may exploit data provided by people. Artificial intelligence areas, such as machine learning and natural language processing, have already successfully employed privacy-preserving mechanisms in order to safeguard data privacy in a vast number of applications. Information retrieval (IR) is likewise prone to privacy threats, such as attacks and unintended disclosures of documents and search history, which may cripple the security of users and be penalized by data protection laws. This work aims at highlighting and discussing open challenges for privacy in the recent literature of IR, focusing on tasks featuring user-generated text data. Our contribution is threefold: firstly, we present an overview of privacy threats to IR tasks; secondly, we discuss applicable privacy-preserving mechanisms which may be employed in solutions to restrain privacy hazards; finally, we bring insights on the tradeoffs between privacy preservation and utility performance for IR tasks.

show abstract

SecureNLP: A System for Multi-Party Privacy-Preserving Natural Language Processing

Cited by 65 publications

References 39 publications

AFLPC: An Asynchronous Federated Learning Privacy-Preserving Computing Model Applied to 5G-V2X

AFLPC: An Asynchronous Federated Learning Privacy-Preserving Computing Model Applied to 5G-V2X

Multilingual Sentence-Level Bias Detection inWikipedia

Privacy in Open Search: A Review of Challenges and Solutions

Contact Info

Product

Resources

About