Secure human action recognition by encrypted neural network inference

Kim, Miran; Jiang, Xiaoqian; Lauter, Kristin; Ismayilzada, Elkhan; Shams, Soheil

doi:10.1038/s41467-022-32168-5

Cited by 24 publications

(14 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Most of the existing works (i.e., [21], [33]- [38]) investigate the training of NNs on encrypted data rely on homomorphic encryption (HE). An example of HE implementation is the CKKS scheme [39], which can support approximate computation with floating-point numbers.…”

Section: B Neural Network (Nn) For Training Encrypted Datamentioning

confidence: 99%

“…To make the machine learning model compatible with the HE scheme, the authors in [35] redesign the conventional NNs to be HE-compatible by replacing nonlinear activation functions with low-degree polynomials that only include addition and multiplication calculations. However, all the works in [21], [33]- [38] considered outsourcing the training task to a single central server (e.g., the CS), which limits the capacity to handle a massive amount of data in largescale FL systems.…”

Section: B Neural Network (Nn) For Training Encrypted Datamentioning

confidence: 99%

See 1 more Smart Citation

Encrypted Data Caching and Learning Framework for Robust Federated Learning-Based Mobile Edge Computing

Nguyen,

Saputra,

Hoang

et al. 2024

IEEE/ACM Trans. Networking

View full text Add to dashboard Cite

Federated Learning (FL) plays a pivotal role in enabling artificial intelligence (AI)-based mobile applications in mobile edge computing (MEC). However, due to the resource heterogeneity among participating mobile users (MUs), delayed updates from slow MUs may deteriorate the learning speed of the MEC-based FL system, commonly referred to as the straggling problem. To tackle the problem, this work proposes a novel privacy-preserving FL framework that utilizes homomorphic encryption (HE) based solutions to enable MUs, particularly resource-constrained MUs, to securely offload part of their training tasks to the cloud server (CS) and mobile edge nodes (MENs). Our framework first develops an efficient method for packing batches of training data into HE ciphertexts to reduce the complexity of HE-encrypted training at the MENs/CS. On that basis, the mobile service provider (MSP) can incentivize straggling MUs to encrypt part of their local datasets that are uploaded to certain MENs or the CS for caching and remote training. However, caching a large amount of encrypted data at the MENs and CS for FL may not only overburden those nodes but also incur a prohibitive cost of remote training, which ultimately reduces the MSP's overall profit. To optimize the portion of MUs' data to be encrypted, cached, and trained at the MENs/CS, we formulate an MSP's profit maximization problem, considering all MUs' and MENs' resource capabilities and data handling costs (including encryption, caching, and training) as well as the MSP's incentive budget. We then show that the problem is convex and can be efficiently solved using an interior point method. Extensive simulations on a real-world human activity recognition dataset show that our proposed framework can achieve much higher model accuracy (improving up to 24.29%) and faster convergence rate (by 2.86 times) than those of the conventional FedAvg approach when the straggling probability varies between 20% and 80%. Moreover, the proposed framework can improve the MSP's profit up to 2.84 times compared with other baseline FL approaches without MEN-assisted training. I. INTRODUCTIONAlong with the ever-growing number of connected Internet of Things (IoT) devices, machine learning (ML) has been

show abstract

Section: B Neural Network (Nn) For Training Encrypted Datamentioning

confidence: 99%

Section: B Neural Network (Nn) For Training Encrypted Datamentioning

confidence: 99%

Encrypted Data Caching and Learning Framework for Robust Federated Learning-Based Mobile Edge Computing

Nguyen,

Saputra,

Hoang

et al. 2024

IEEE/ACM Trans. Networking

View full text Add to dashboard Cite

show abstract

“…Taking the United States as an example, it already has a complete security infrastructure, security assessment criteria, certified encryption standards and related regulations, and has formed a sound information security industry chain [5]. Literature [6] pointed out that "Cloud Security Alliance" is a non-profit, non-profit organization, its task is to solve the security problems existing in the cloud computing network, propose solutions, and improve the security index of cloud computing. After the Cloud Security Alliance was established in the United States, cloud computing providers such as Microsoft and Google have joined the alliance.…”

Section: Literature Reviewmentioning

confidence: 99%

Secure Storage of Computer Network Data based on Cloud Computing

Jiang

2023

SCPE

View full text Add to dashboard Cite

Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources. User’s data is stored in large database. The stored data can be accessed and modified by the clients over the Internet. The data is monitored by the Third Party Auditor (TPA) on behalf of the client. Therefore, integrity is lacked by the data stored on the servers. The data integrity is ensured by the cloud services that provide trust to the privacy of users. Aiming at the urgent problem of network data security in cloud computing operations, this paper proposed a security situation assessment system to grasp the security situation in real time. A set of cloud computing network data storage security is proposed. Through this model, the extraction of network data storage security situation elements, the design of network data storage security situation assessment scheme and the calculation method of network data storage security situation value are completed. The experimental results show the error of the predicted value obtained by the network data storage security situation assessment system. The effectiveness of the system model and the superiority of the improved algorithm is verified by the experimental results. This paper uses the cloud model to predict the cloud computing network security situation. On the other hand, the security situation value obtained by the situation assessment process can be used directly without training the original situation value. Performance improvement by the proposed technique over existing technique is seen and it is observe that the proposed technique is 23% and 34% better than the existing techniques.

show abstract

“…As a result, it has widespread applications in fields such as autonomous driving, remote sensing, and security surveillance systems. [10][11][12][13] However, despite the notable advantages of polarization imaging mentioned above, it still faces some limitations in further miniaturizing the sensor footprint and the implementation of image enhancement. For instance, achieving the recognition of band information of imaged objects and implementing image enhancement based on polarization imaging remains challenging, requiring the use of spectrometers or other components, [14] thereby increasing the complexity of the image processing system and hindering the trend of miniaturization.…”

Section: Introductionmentioning

confidence: 99%

Polarization Reversal of Group IV–VI Semiconductors with Pucker‐Like Structure: Mechanism Dissecting and Function Demonstration

Yu,

Xiong,

Liu

et al. 2023

Advanced Materials

View full text Add to dashboard Cite

Polarization imaging presents advantages in capturing spatial, spectral, and polarization information from target objects across various spectral bands. It can improve the perceptual ability of image sensors and has garnered more applications. Despite its potential, challenges persist in identifying band information from imaged objects and implementing image enhancement using polarization imaging. These challenges often necessitate integrating spectrometers or other components, resulting in increased complexities within image processing systems and hindering device miniaturization trends. Here, we systematically elucidate the characteristics of anisotropic absorption reversal in pucker‐like group IV‐VI semiconductors MX (M = Ge, Sn; X = S, Se) through theoretical predictions and experimental validations. Additionally, the fundamental mechanisms behind anisotropy reversal in different bands are also explored. The polarization‐sensitive photodetector is constructed by utilizing MX as light‐absorbing layer, harnessing polarization‐sensitive photoresponse for virtual imaging. The results indicate that the utilization of polarization reversal photodetectors, without excessively emphasizing anisotropy photocurrent ratios, holds advantages in achieving further multifunctional integration within the device structure while simplifying its configuration, including band information identification and image enhancement. This study provides a comprehensive analysis of polarization reversal mechanisms and presents a promising and reliable approach for achieving dual‐band image band identification and image enhancement without additional auxiliary components.This article is protected by copyright. All rights reserved

show abstract

Secure human action recognition by encrypted neural network inference

Cited by 24 publications

References 44 publications

Encrypted Data Caching and Learning Framework for Robust Federated Learning-Based Mobile Edge Computing

Encrypted Data Caching and Learning Framework for Robust Federated Learning-Based Mobile Edge Computing

Secure Storage of Computer Network Data based on Cloud Computing

Polarization Reversal of Group IV–VI Semiconductors with Pucker‐Like Structure: Mechanism Dissecting and Function Demonstration

Contact Info

Product

Resources

About