arXiv subjects

Jiahui Hu

Publications and source records attributed to Jiahui Hu.

23 records · Page 2Linked to original sources

arXiv2024

Breaking Secure Aggregation: Label Leakage from Aggregated Gradients in Federated Learning

Federated Learning (FL) exhibits privacy vulnerabilities under gradient inversion attacks (GIAs), which can extract private information from individual gradients. To enhance privacy, FL incorporates Secure Aggregation (SA) to prevent the server from obtaining individual gradients, thus effectively resisting GIAs. In this paper, we propose a stealthy label inference attack to bypass SA and recover individual clients' private labels. Specifically, we conduct a theoretical analysis of label inference from the aggregated gradients that are exclusively obtained after implementing SA. The analysis results reveal that the inputs (embeddings) and outputs (logits) of the final fully connected layer (FCL) contribute to gradient disaggregation and label restoration. To preset the embeddings and logits of FCL, we craft a fishing model by solely modifying the parameters of a single batch normalization (BN) layer in the original model. Distributing client-specific fishing models, the server can derive the individual gradients regarding the bias of FCL by resolving a linear system with expected embeddings and the aggregated gradients as coefficients. Then the labels of each client can be precisely computed based on preset logits and gradients of FCL's bias. Extensive experiments show that our attack achieves large-scale label recovery with 100\% accuracy on various datasets and model architectures.

cs.CR↗

arXiv2023

Privacy-preserving Adversarial Facial Features

Face recognition service providers protect face privacy by extracting compact and discriminative facial features (representations) from images, and storing the facial features for real-time recognition. However, such features can still be exploited to recover the appearance of the original face by building a reconstruction network. Although several privacy-preserving methods have been proposed, the enhancement of face privacy protection is at the expense of accuracy degradation. In this paper, we propose an adversarial features-based face privacy protection (AdvFace) approach to generate privacy-preserving adversarial features, which can disrupt the mapping from adversarial features to facial images to defend against reconstruction attacks. To this end, we design a shadow model which simulates the attackers' behavior to capture the mapping function from facial features to images and generate adversarial latent noise to disrupt the mapping. The adversarial features rather than the original features are stored in the server's database to prevent leaked features from exposing facial information. Moreover, the AdvFace requires no changes to the face recognition network and can be implemented as a privacy-enhancing plugin in deployed face recognition systems. Extensive experimental results demonstrate that AdvFace outperforms the state-of-the-art face privacy-preserving methods in defending against reconstruction attacks while maintaining face recognition accuracy.

cs.CV↗

arXiv2020

Novel Porous Aluminum Nitride Monolayer Structures A First-principles Study

By using ab inito calculations within Density Functional Theory, we have explored the possible structures and various properties of porous AlN monolayer materials. Two kinds of porous AlN monolayer are identified, and the phonon dispersion spectrum together with the ab inito molecular dynamics simulations demonstrate that their structures are stable. We further show that these AlN porous monolayers have well-defined porous nanostructures and even higher specific surface areas, namely, which can be comparable with graphene, and can also be maintained evenly at high temperatures. Furthermore, both porous monolayers exhibit semiconductor properties with 2.89 eV and 2.86 eV indirect band gap, respectively. In addition, the electronic structures of such porous monolayers can be modulated by strains. The band gap of porous AlN monolayer experiences an indirect-direct transition when biaxial strain is applied. A moderate 9% compression can trigger this gap transition. These results indicate that the porous AlN monolayer may potentially be used for optoelectronic applications, as well as for underlying catalysts in the future.

cond-mat.mtrl-sci↗

arXiv2017

Numerical algorithm for two-dimensional time-fractional wave equation of distributed-order with a nonlinear source term

In this paper, an alternating direction implicit (ADI) difference scheme for two-dimensional time-fractional wave equation of distributed-order with a nonlinear source term is presented. The unique solvability of the difference solution is discussed, and the unconditional stability and convergence order of the numerical scheme are analysed. Finally, numerical experiments are carried out to verify the effectiveness and accuracy of the algorithm.

math.NA↗

arXiv2016

Maximum-norm error analysis of compact difference schemes for the backward fractional Feynman-Kac equation

The fractional Feynman-Kac equations describe the distribution of functionals of non-Brownian motion, or anomalous diffusion, including two types called the forward and backward fractional Feynman-Kac equations, where the fractional substantial derivative is involved. This paper focuses on the more widely used backward version. Based on the discretized schemes for fractional substantial derivatives proposed recently, we construct compact finite difference schemes for the backward fractional Feynman-Kac equation, which has q-th (q=1, 2, 3, 4) order accuracy in temporal direction and fourth order accuracy in spatial direction, respectively. In the case q=1, the numerical stability and convergence of the difference scheme in the discrete L-infinity norm are proved strictly, where a new inner product is defined for the theoretical analysis. Finally, numerical examples are provided to verify the effectiveness and accuracy of the algorithms.

math.NA↗