Privacy-preserving federated neural network learning for disease-associated cell classification

Sav, Sinem; Bossuat, Jean-Philippe; Troncoso-Pastoriza, Juan Ramón; Claassen, Manfred; Hubaux, Jean-Pierre

doi:10.1016/j.patter.2022.100487

Cited by 17 publications

(8 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, we will explore HE-based training strategies which is hampered by the model complexity. Additionally, more advanced HE algorithms should be explored to mitigate potential risks of information leakage caused by intermediate model parameters exchanged in federated learning or distributed learning settings Sav et al [2022], Nasr et al [2019], Hitaj et al [2017], Wang et al [2019].…”

Section: Discussionmentioning

confidence: 99%

Secure Deep Learning on Genomics Data via a Homomorphic Encrypted Residue Activation Network

Song

Shi

2023

Preprint

View full text Add to dashboard Cite

Growing applications of deep learning on sensitive genomics and biomedical data introduce challenging privacy and secure problems. Homomorphic encryption (HE) is one of appropriate cryptographic techniques to provide secure machine learning evaluation by directly computing over encrypted data, so that allows the data owner and model owner to outsource processing of sensitive data to an untrusted server without leaking any information about the data. However, most current HE schemes only support limited arithmetic operations, which significantly hinder their applications to support secure deep learning algorithm. Considering the potential performance loss introduced for approximating activation function, in this paper, we develop a novel HE friendly deep network, named Residue Activation Network (ResActNet) to implement precise privacy-preserving machine learning algorithm with a non-approximating activation on HE scheme. We considered a residue activation strategy with a scaled power activation function in the deep network. In particular, a scaled power activation (SPA) function is set within the HE scheme, and so that can be directly deployed on HE computation. Moreover, we proposed a residue activation strategy to constrain the latent space in the training process for alleviating the optimization difficulty. We comprehensively evaluate ResActNet using diverse genomics datasets and widely-used image datasets. Our results demonstrate that ResActNet outperforms other alternative solutions to secure machine learning with HE and achieves low approximation errors in classification and regression tasks.

show abstract

Section: Discussionmentioning

confidence: 99%

Secure Deep Learning on Genomics Data via a Homomorphic Encrypted Residue Activation Network

Song

Shi

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…reported to be significant, indicating a large overlap despite differences in the analysis setting. In contrast, independently analyzing each center's private dataset led to 2,600 significant loci across all centers, considerably fewer than SF-GWAS (24,357). Moreover, given eMERGE as a study cohort and UKB as a larger validation cohort, meta-analyses of seven center-specific GWAS based on the eMERGE data resulted in fewer loci that are validated in UKB than SF-GWAS, illustrating the potential of the joint analysis enabled by SF-GWAS to increase statistical power (Supplementary Fig.…”

Section: /20mentioning

confidence: 94%

“…Improving upon existing works on MHE 8,[23][24][25] , we switch between the MHE and secret sharing representations of intermediate data, which enables efficient state-of-the-art MPC routines to be used in conjunction with MHE operations to carry out the global computations (Supplementary Note 3). We convert between the two schemes to perform key operations under the most efficient scheme for each of the computational steps of the GWAS pipeline.…”

Section: Our Approach: Combine He and Mpc To Enable Practical Secure ...mentioning

confidence: 99%

Secure and Federated Genome-Wide Association Studies for Biobank-Scale Datasets

Cho

Froelicher

Chen

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Sharing data across multiple institutions for genome-wide association studies (GWAS) would enable discovery of novel genetic variants linked to health and disease. However, existing regulations on genomic data sharing and the sheer size of the data limit the scope of such collaborations. Although cryptographic tools for secure computation promise to enable collaborative studies with formal privacy guarantees, existing approaches either are computationally impractical or support only simplified analysis pipelines. Here, we introduce secure and federated genome-wide association studies (SF-GWAS), a novel combination of secure computation frameworks that empowers efficient and accurate GWAS in a federated manner, i.e., on private data locally-held by multiple entities, while provably ensuring end-to-end data confidentiality. Another key advance is that we designed SF-GWAS to support the two most widely-used GWAS pipelines---those based on principal component analysis (PCA) or linear mixed models (LMMs). We ran SF-GWAS on five real GWAS datasets, including a large UK Biobank cohort of 410K individuals, thereby demonstrating the largest secure genetics collaboration to date. SF-GWAS achieves an order-of-magnitude runtime improvement over the prior art for PCA-based GWAS and newly allows secure LMM-based association tests, for which its runtime scales at a near-constant rate in cohort size. Our work realizes the power of secure, collaborative, and accurate GWAS at unprecedented scale and should be applicable to a broad range of analyses. Our open-source software is at: https://github.com/hhcho/sfgwas.

show abstract

“…The accuracy of our solution is comparable to that of the centralized non-secure method. PriCell enables data utility for effective multi-center investigations involving complicated healthcare data while guaranteeing patient privacy [22]. Veeramakali 2022 et al Choosing machine learning techniques should be based on solid results.…”

Section: Related Workmentioning

confidence: 99%

A Review on Privacy Preserving Using Machine learning and Deep Learning Techniques

Rajput¹,

Tiwari²

2023

IJRASET

View full text Add to dashboard Cite

In order to entice data custodians to provide precise documentation so that data mining can continue with confidence, protecting the confidentiality of healthcare data is crucial. Association rule mining has been extensively used in the past to analyze healthcare data. The majority of applications ignore the drawbacks of specific diagnostic procedures in favor of positive association criteria. Negative association criteria may provide more useful information when bridging disparate diseases and medications than positive ones. In the case of doctors and social groups, this is particularly accurate. Data mining for medical purposes must be done with patient identities protected, especially when working with sensitive data. However, it might be attacked if this information becomes public. In order to perform data mining research, technology that modifies data (data sanitization) that reconstructs aggregate distributions has recently addressed the importance of healthcare data privacy. This study examines data sanitization in healthcare data mining using metaheuristics in order to safeguard patient privacy. Studies on SHM have looked at the uses of IoT &/or Machine Learning (ML) within the field, as well as the architecture, security, & privacy issues. However, no studies have looked into how AI and ubiquity computing technologies have affected SHM systems. The objective of this research is to identify and map the primary technical concepts within the SHM framework.

show abstract

Privacy-preserving federated neural network learning for disease-associated cell classification

Cited by 17 publications

References 36 publications

Secure Deep Learning on Genomics Data via a Homomorphic Encrypted Residue Activation Network

Secure Deep Learning on Genomics Data via a Homomorphic Encrypted Residue Activation Network

Secure and Federated Genome-Wide Association Studies for Biobank-Scale Datasets

A Review on Privacy Preserving Using Machine learning and Deep Learning Techniques

Contact Info

Product

Resources

About