Accessibility of covariance information creates vulnerability in Federated Learning frameworks.

MOTIVATION: Federated Learning (FL) is gaining traction in various fields as it enables integrative data analysis without sharing sensitive data, such as in healthcare. However, the risk of data leakage caused by malicious attacks must be considered. In this study, we introduce a novel attack algorithm that relies on being able to compute sample means, sample covariances, and construct known linearly independent vectors on the data owner side. RESULTS: We show that these basic functionalities, which are available in several established FL frameworks, are sufficient to reconstruct privacy-protected data. Additionally, the attack algorithm is robust to defense strategies that involve adding random noise. We demonstrate the limitations of existing frameworks and propose potential defense strategies analyzing the implications of using differential privacy. The novel insights presented in this study will aid in the improvement of FL frameworks. AVAILABILITY AND IMPLEMENTATION: The code examples are provided at GitHub (https://github.com/manuhuth/Data-Leakage-From-Covariances.git). The CNSIM1 dataset, which we used in the manuscript, is available within the DSData R package (https://github.com/datashield/DSData/tree/main/data).

Nucleocapsid-specific T cell responses associate with control of SARS-CoV-2 in the upper airways before seroconversion.

Despite intensive research since the emergence of SARS-CoV-2, it has remained unclear precisely which components of the early immune response protect against the development of severe COVID-19. Here, we perform a comprehensive immunogenetic and virologic analysis of nasopharyngeal and peripheral blood samples obtained during the acute phase of infection with SARS-CoV-2. We find that soluble and transcriptional markers of systemic inflammation peak during the first week after symptom onset and correlate directly with upper airways viral loads (UA-VLs), whereas the contemporaneous frequencies of circulating viral nucleocapsid (NC)-specific CD4+ and CD8+ T cells correlate inversely with various inflammatory markers and UA-VLs. In addition, we show that high frequencies of activated CD4+ and CD8+ T cells are present in acutely infected nasopharyngeal tissue, many of which express genes encoding various effector molecules, such as cytotoxic proteins and IFN-γ. The presence of IFNG mRNA-expressing CD4+ and CD8+ T cells in the infected epithelium is further linked with common patterns of gene expression among virus-susceptible target cells and better local control of SARS-CoV-2. Collectively, these results identify an immune correlate of protection against SARS-CoV-2, which could inform the development of more effective vaccines to combat the acute and chronic illnesses attributable to COVID-19.

Using machine learning to predict antibody response to SARS-CoV-2 vaccination in solid organ transplant recipients: the multicentre ORCHESTRA cohort.

OBJECTIVES: The study aim was to assess predictors of negative antibody response (AbR) in solid organ transplant (SOT) recipients after the first booster of SARS-CoV-2 vaccination. METHODS: Solid organ transplant recipients receiving SARS-CoV-2 vaccination were prospectively enrolled (March 2021-January 2022) at six hospitals in Italy and Spain. AbR was assessed at first dose (t0), second dose (t1), 3 ± 1 month (t2), and 1 month after third dose (t3). Negative AbR at t3 was defined as an anti-receptor binding domain titre <45 BAU/mL. Machine learning models were developed to predict the individual risk of negative (vs. positive) AbR using age, type of transplant, time between transplant and vaccination, immunosuppressive drugs, type of vaccine, and graft function as covariates, subsequently assessed using a validation cohort. RESULTS: Overall, 1615 SOT recipients (1072 [66.3%] males; mean age±standard deviation [SD], 57.85 ± 13.77) were enrolled, and 1211 received three vaccination doses. Negative AbR rate decreased from 93.66% (886/946) to 21.90% (202/923) from t0 to t3. Univariate analysis showed that older patients (mean age, 60.21 ± 11.51 vs. 58.11 ± 13.08), anti-metabolites (57.9% vs. 35.1%), steroids (52.9% vs. 38.5%), recent transplantation (<3 years) (17.8% vs. 2.3%), and kidney, heart, or lung compared with liver transplantation (25%, 31.8%, 30.4% vs. 5.5%) had a higher likelihood of negative AbR. Machine learning (ML) algorithms showing best prediction performance were logistic regression (precision-recall curve-PRAUC mean 0.37 [95%CI 0.36-0.39]) and k-Nearest Neighbours (PRAUC 0.36 [0.35-0.37]). DISCUSSION: Almost a quarter of SOT recipients showed negative AbR after first booster dosage. Unfortunately, clinical information cannot efficiently predict negative AbR even with ML algorithms.