Differential Privacy Protection Against Membership Inference Attack on Machine Learning for Genomic Data

Chen, Junjie; Wang, Wendy Hui; Shi, Xinghua

doi:10.1101/2020.08.03.235416

Cited by 16 publications

(18 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DP is often obtained by applying a procedure that introduces randomness into the data. DP has been the most widely used method, according to Chen et al [30], to assess privacy exposure relating to persons. In addition, Chen et al [30] has evaluated DP uses and its efficiency as solution to MIA in genomic data.…”

Section: Discussionmentioning

confidence: 99%

“…Cryptographic techniques can also be used to secure ML models [28,29]. Chen et al [30] assess the effectiveness of using Differential Privacy as a genomic data protection mechanism to minimize the danger of membership inference attacks.…”

Section: Defence Techniques Against Attacks In the Testing/inferring Phasementioning

confidence: 99%

“…Nasr et al [14] Differential privacy (DP) Chen et al [30] Homomorphic encryption is a technique to provide data privacy via data encryption Biggio et al [32] We presented the main existing countermeasures against machine learning attacks, as shown in Tab. 4.…”

Section: Min-max Privacy Gamementioning

confidence: 99%

See 2 more Smart Citations

Towards Securing Machine Learning Models Against Membership Inference Attacks

Hamida¹,

Mrabet²,

Belguith³

et al. 2022

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

From fraud detection to speech recognition, including price prediction, Machine Learning (ML) applications are manifold and can significantly improve different areas. Nevertheless, machine learning models are vulnerable and are exposed to different security and privacy attacks. Hence, these issues should be addressed while using ML models to preserve the security and privacy of the data used. There is a need to secure ML models, especially in the training phase to preserve the privacy of the training datasets and to minimise the information leakage. In this paper, we present an overview of ML threats and vulnerabilities, and we highlight current progress in the research works proposing defence techniques against ML security and privacy attacks. The relevant background for the different attacks occurring in both the training and testing/inferring phases is introduced before presenting a detailed overview of Membership Inference Attacks (MIA) and the related countermeasures. In this paper, we introduce a countermeasure against membership inference attacks (MIA) on Conventional Neural Networks (CNN) based on dropout and L2 regularization. Through experimental analysis, we demonstrate that this defence technique can mitigate the risks of MIA attacks while ensuring an acceptable accuracy of the model. Indeed, using CNN model training on two datasets CIFAR-10 and CIFAR-100, we empirically verify the ability of our defence strategy to decrease the impact of MIA on our model and we compare results of five different classifiers. Moreover, we present a solution to achieve a trade-off between the performance of the model and the mitigation of MIA attack.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Defence Techniques Against Attacks In the Testing/inferring Phasementioning

confidence: 99%

See 1 more Smart Citation

Towards Securing Machine Learning Models Against Membership Inference Attacks

Hamida¹,

Mrabet²,

Belguith³

et al. 2022

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

show abstract

“…Therefore, devising PPTs with fewer parameters for configuration, or possibly self-learning parameter values based on data, is still an open issue in the privacy domain. In recent years, ML has been extensively used to address various privacy concerns emerging from digitization [404]- [407]. Therefore, employing different ML concepts and techniques to secure personal data is an interesting area of research.…”

Section: B Promising Future Research Directionsmentioning

confidence: 99%

A Comprehensive Analysis of Privacy Protection Techniques Developed for COVID-19 Pandemic

Majeed

Hwang

2021

IEEE Access

View full text Add to dashboard Cite

Since the emergence of coronavirus disease-2019 (COVID-19) outbreak, every country has implemented digital solutions in the form of mobile applications, web-based frameworks, and/or integrated platforms in which huge amounts of personal data are collected for various purposes (e.g., contact tracing, suspect search, and quarantine monitoring). These systems not only collect basic data about individuals but, in most cases, very sensitive data like their movements, spatio-temporal activities, travel history, visits to churches/clubs, purchases, and social interactions. While collection and utilization of person-specific data in different contexts is essential to limiting the spread of COVID-19, it increases the chances of privacy breaches and personal data misuse. Recently, many privacy protection techniques (PPTs) have been proposed based on the person-specific data included in different data types (e.g., tables, graphs, matrixes, barcodes, and geospatial data), and epidemic containment strategies (ECSs) (contact tracing, quarantine monitoring, symptom reports, etc.) in order to minimize privacy breaches and to permit only the intended uses of such personal data. In this paper, we present an extensive review of the PPTs that have been recently proposed to address the diverse privacy requirements/concerns stemming from the COVID-19 pandemic. We describe the heterogeneous types of data collected to control this pandemic, and the corresponding PPTs, as well as the paradigm shifts in personal data handling brought on by this pandemic. We systemically map the recently proposed PPTs into various ECSs and data lifecycle phases, and present an in-depth review of existing PPTs and evaluation metrics employed for analysis of their suitability. We describe various PPTs developed during the COVID-19 period that leverage emerging technologies, such as federated learning, blockchain, privacy by design, and swarm learning, to name a few. Furthermore, we discuss the challenges of preserving individual privacy during a pandemic, the role of privacy regulations/laws, and promising future research directions. With this article, our aim is to highlight the recent PPTs that have been specifically proposed for the COVID-19 arena, and point out research gaps for future developments in this regard.

show abstract

“…DP is employed to prevent the leakage of personal information during the stage in a deep learning model [40][41][42]. Moreover, because healthcare data contain privacy-sensitive information, DP is adopted in various deep learning and artificial intelligence systems for healthcare system [43][44][45][46][47].…”

Section: Local Differential Privacy For De-identificationmentioning

confidence: 99%

Blockchain-Based Privacy-Preserving System for Genomic Data Management Using Local Differential Privacy

2021

View full text Add to dashboard Cite

The advances made in genome technology have resulted in significant amounts of genomic data being generated at an increasing speed. As genomic data contain various privacy-sensitive information, security schemes that protect confidentiality and control access are essential. Many security techniques have been proposed to safeguard healthcare data. However, these techniques are inadequate for genomic data management because of their large size. Additionally, privacy problems due to the sharing of gene data are yet to be addressed. In this study, we propose a secure genomic data management system using blockchain and local differential privacy (LDP). The proposed system employs two types of storage: private storage for internal staff and semi-private storage for external users. In private storage, because encrypted gene data are stored, only internal employees can access the data. Meanwhile, in semi-private storage, gene data are irreversibly modified by LDP. Through LDP, different noises are added to each section of the genomic data. Therefore, even though the third party uses or exposes the shared data, the owner’s privacy is guaranteed. Furthermore, the access control for each storage is ensured by the blockchain, and the gene owner can trace the usage and sharing status using a decentralized application in a mobile device.

show abstract

Differential Privacy Protection Against Membership Inference Attack on Machine Learning for Genomic Data

Cited by 16 publications

References 39 publications

Towards Securing Machine Learning Models Against Membership Inference Attacks

Towards Securing Machine Learning Models Against Membership Inference Attacks

A Comprehensive Analysis of Privacy Protection Techniques Developed for COVID-19 Pandemic

Blockchain-Based Privacy-Preserving System for Genomic Data Management Using Local Differential Privacy

Contact Info

Product

Resources

About