arXiv subjects

Ruyuan Zhang

Publications and source records attributed to Ruyuan Zhang.

6 recordsLinked to original sources

arXiv2025

Flexible Threshold Multi-client Functional Encryption for Inner Product in Federated Learning

Federated learning (FL) is a distributed machine learning paradigm that enables multiple clients to collaboratively train a shared model without disclosing their local data. To address privacy issues of gradient, several privacy-preserving machine-learning schemes based on multi-client functional encryption (MCFE) have been proposed. However, existing MCFE-based schemes cannot support client dropout or flexible threshold selection, which are essential for practical FL. In this paper, we design a flexible threshold multi-client functional encryption for inner product (FTMCFE-IP) scheme, where multiple clients generate ciphertexts independently without any interaction. In the encryption phase, clients are able to choose a threshold flexibly without reinitializing the system. The decryption can be performed correctly when the number of online clients satisfies the threshold. An authorized user are allowed to compute the inner product of the vectors associated with his/her functional key and the ciphertext, respectively, but cannot learning anything else. Especially, the presented scheme supports clients drop out. Furthermore, we provide the definition and security model of our FTMCFE-IP scheme,and propose a concrete construction. The security of the designed scheme is formally proven. Finally, we implement and evaluate our FTMCFE-IP scheme.

cs.CR↗

arXiv2024

Multi-client Functional Encryption for Set Intersection with Non-monotonic Access Structures in Federated Learning

Federated learning (FL) based on cloud servers is a distributed machine learning framework that involves an aggregator and multiple clients, which allows multiple clients to collaborate in training a shared model without exchanging data. Considering the confidentiality of training data, several schemes employing functional encryption (FE) have been presented. However, existing schemes cannot express complex access control policies. In this paper, to realize more flexible and fine-grained access control, we propose a multi-client functional encryption scheme for set intersection with non-monotonic access structures (MCFE-SI-NAS), where multiple clients co-exist and encrypt independently without interaction. All ciphertexts are associated with an label, which can resist "mix-and-match" attacks. Aggregator can aggregate ciphertexts, but cannot know anything about the plaintexts. We first formalize the definition and security model for the MCFE-SI-NAS scheme and build a concrete construction based on asymmetric prime-order pairings. The security of our scheme is formally proven. Finally, we implement our MCFE-SI-NAS scheme and provide its efficiency analysis.

cs.CR↗

arXiv2020

Dissociable neural representations of adversarially perturbed images in convolutional neural networks and the human brain

Despite the remarkable similarities between convolutional neural networks (CNN) and the human brain, CNNs still fall behind humans in many visual tasks, indicating that there still exist considerable differences between the two systems. Here, we leverage adversarial noise (AN) and adversarial interference (AI) images to quantify the consistency between neural representations and perceptual outcomes in the two systems. Humans can successfully recognize AI images as corresponding categories but perceive AN images as meaningless noise. In contrast, CNNs can correctly recognize AN images but mistakenly classify AI images into wrong categories with surprisingly high confidence. We use functional magnetic resonance imaging to measure brain activity evoked by regular and adversarial images in the human brain, and compare it to the activity of artificial neurons in a prototypical CNN-AlexNet. In the human brain, we find that the representational similarity between regular and adversarial images largely echoes their perceptual similarity in all early visual areas. In AlexNet, however, the neural representations of adversarial images are inconsistent with network outputs in all intermediate processing layers, providing no neural foundations for perceptual similarity. Furthermore, we show that voxel-encoding models trained on regular images can successfully generalize to the neural responses to AI images but not AN images. These remarkable differences between the human brain and AlexNet in the representation-perception relation suggest that future CNNs should emulate both behavior and the internal neural presentations of the human brain.

cs.CV↗

arXiv2019

A visual encoding model based on deep neural networks and transfer learning

Background: Building visual encoding models to accurately predict visual responses is a central challenge for current vision-based brain-machine interface techniques. To achieve high prediction accuracy on neural signals, visual encoding models should include precise visual features and appropriate prediction algorithms. Most existing visual encoding models employ hand-craft visual features (e.g., Gabor wavelets or semantic labels) or data-driven features (e.g., features extracted from deep neural networks (DNN)). They also assume a linear mapping between feature representation to brain activity. However, it remains unknown whether such linear mapping is sufficient for maximizing prediction accuracy. New Method: We construct a new visual encoding framework to predict cortical responses in a benchmark functional magnetic resonance imaging (fMRI) dataset. In this framework, we employ the transfer learning technique to incorporate a pre-trained DNN (i.e., AlexNet) and train a nonlinear mapping from visual features to brain activity. This nonlinear mapping replaces the conventional linear mapping and is supposed to improve prediction accuracy on brain activity. Results: The proposed framework can significantly predict responses of over 20% voxels in early visual areas (i.e., V1-lateral occipital region, LO) and achieve unprecedented prediction accuracy. Comparison with Existing Methods: Comparing to two conventional visual encoding models, we find that the proposed encoding model shows consistent higher prediction accuracy in all early visual areas, especially in relatively anterior visual areas (i.e., V4 and LO). Conclusions: Our work proposes a new framework to utilize pre-trained visual features and train non-linear mappings from visual features to brain activity.

cs.CV↗

arXiv2018

LAMOST observations in the Kepler field. II. Database of the low-resolution spectra from the five-year regular survey

The LAMOST-Kepler (LK-) project was initiated to use the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) to make spectroscopic follow-up observations for the targets in the field of the Kepler mission. The Kepler field is divided into 14 subfields that are adapted to the LAMOST circular field with a diameter of 5 degrees. During the regular survey phase of LAMOST, the LK-project took data from 2012 June to 2017 June and covered all the 14 subfields at least twice. In particular, we describe in this paper the second Data Release of the LK-project, including all spectra acquired through 2015 May to 2017 June together with the first round observations of the LK-project from 2012 June to 2014 September. The LK-project now counts 227 870 spectra of 156 390 stars, among which we have derived atmospheric parameters (log g, T eff and [Fe/H]) and heliocentric radial velocity (RV) for 173 971 spectra of 126 172 stars. These parameters were obtained with the most recent version of the LAMOST Stellar Parameter Pipeline v 2.9.7. Nearly one half, namely 76 283 targets, are observed both by LAMOST and Kepler telescopes. These spectra, establishing a large spectroscopy library, will be useful for the entire astronomical community, particularly for planetary science and stellar variability on Kepler targets.

astro-ph.SR↗

arXiv2016

LAMOST observations in the Kepler field. Analysis of the stellar parameters measured with the LASP based on the low-resolution spectra

All of the 14 subfields of the Kepler field have been observed at least once with the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST, Xinglong Observatory, China) during the 2012-2014 observation seasons. There are 88,628 reduced spectra with SNR$_g$ (signal-to-noise ratio in g band) $\geq$ 6 after the first round (2012-2014) of observations for the LAMOST-Kepler project (LK-project). By adopting the upgraded version of the LAMOST Stellar Parameter pipeline (LASP), we have determined the atmospheric parameters ($T_{\rm eff}$ , $\log g$, and $\rm [Fe/H]$) and heliocentric radial velocity $v_{\rm rad}$ for 51,406 stars with 61,226 spectra. Compared with atmospheric parameters derived from both high-resolution spectroscopy and asteroseismology method for common stars in Huber et al. (2014), an external calibration of LASP atmospheric parameters was made, leading to the determination of external errors for the giants and dwarfs, respectively. Multiple spectroscopic observations for the same objects of the LK-project were used to estimate the internal uncertainties of the atmospheric parameters as a function of SNR$_g$ with the unbiased estimation method. The LASP atmospheric parameters were calibrated based on both the external and internal uncertainties for the giants and dwarfs, respectively. A general statistical analysis of the stellar parameters leads to discovery of 106 candidate metal-poor stars, 9 candidate very metal-poor stars, and 18 candidate high-velocity stars. Fitting formulae were obtained segmentally for both the calibrated atmospheric parameters of the LK-project and the KIC parameters with the common stars. The calibrated atmospheric parameters and radial velocities of the LK-project will be useful for studying stars in the Kepler field.

astro-ph.SR↗