Reconfigurable intelligent surface (RIS) is envisioned to be an essential component of the paradigm for beyond 5G networks as it can potentially provide similar or higher array gains with much lower hardware cost and energy consumption compared massive multiple-input multiple-output (MIMO) technology. In this paper, we focus on one fundamental challenges, namely channel acquisition, in RIS-assisted multiuser MIMO system. The state-of-the-art acquisition approach such a system fully passive...

To exploit massive amounts of data generated at mobile edge networks, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">federated learning</i> (FL) has been proposed as an attractive substitute for centralized machine learning (ML). By collaboratively training a shared model devices, FL avoids direct transmission and thus overcomes high communication latency privacy issues compared to ML. improve the efficiency in aggregation,...

Federated learning (FL) has recently emerged as a promising technology to enable artificial intelligence (AI) at the network edge, where distributed mobile devices collaboratively train shared AI model under coordination of an edge server. To significantly improve communication efficiency FL, over-the-air computation allows large number concurrently upload their local models by exploiting superposition property wireless multi-access channels. Due channel fading, aggregation error server is...

The large model size, high computational operations, and vulnerability against membership inference attack (MIA) have impeded deep learning or neural networks (DNNs) popularity, especially on mobile devices. To address the challenge, we envision that weight pruning technique will help DNNs MIA while reducing storage operation. In this work, propose a algorithm, show proposed algorithm can find subnetwork prevent privacy leakage from achieves competitive accuracy with original DNNs. We also...

Matrix inverse computation is one of the most fundamental mathematical problems in large-scale data analytics and computing. It often too expensive to be solved resource-constrained devices such as sensors. Outsourcing task a cloud server or fog potential approach able perform scientific computations on behalf users with special software. However, outsourcing brings new security concerns challenges privacy violations result invalidation. In this paper, we propose secure verifiable scheme...

Although distributed learning has increasingly gained attention in terms of effectively utilizing local devices for data privacy enhancement, recent studies show that publicly shared gradients the training process can reveal private (gradient leakage) to a third-party. We have, however, no systematic understanding gradient leakage mechanism on Transformer based language models. In this paper, as first attempt, we formulate attack problem Transformer-based models and propose algorithm, TAG,...

In this paper, we study blind channel-and-signal estimation by exploiting the burst-sparse structure of angular-domain propagation channels in massive MIMO systems. The state-of-the-art approach utilizes structured channel sparsity sampling representation with a uniform angle-sampling grid, a.k.a. virtual representation. However, is only applicable to linear arrays and may cause substantial performance loss due mismatch between true angle information. To tackle these challenges, propose...

In this paper, we study the beamforming design problem in frequency-division duplexing (FDD) downlink massive MIMO systems, where instantaneous channel state information (CSI) is assumed to be unavailable at base station (BS). We propose extract of angle-of-departures (AoDs) and corresponding large-scale fading coefficients (a.k.a. spatial information) from uplink estimation procedure, based on which a novel presented. By separating subpaths for different users hidden sparsity physical...

Recent studies on attacks exploiting processor hardware vulnerabilities have raised significant concern for information security. Particularly, transient execution such as Spectre augment microarchitectural side channels with speculative executions that lead to exfiltration of secretive data not intended be accessed. Many prior works demonstrated the manipulation branch predictors triggering executions, and thereafter leaking sensitive through components. In this paper, we present a new...

Future wireless networks are expected to support diverse mobile services, including artificial intelligence (AI) services and ubiquitous data transmissions. Federated learning (FL), as a revolutionary approach, enables collaborative AI model training across distributed edge devices. By exploiting the superposition property of multiple-access channels, over-the-air computation allows concurrent uploading from massive devices over same radio resources, thus significantly reduces communication...

Federated learning (FL) enables edge devices to collaboratively train machine models, with model communication replacing direct data uploading. While over-the-air aggregation improves efficiency, uploading models an server over wireless networks can pose privacy risks. Differential (DP) is a widely used quantitative technique measure statistical in FL. Previous research has focused on FL single-antenna server, leveraging noise enhance user-level DP. This approach achieves the so-called "free...

Decision trees are widely used and often assembled as a forest to boost prediction accuracy. However, using decision for inference on GPU is challenging, because of irregular memory access patterns imbalance workloads across threads. This paper proposes Tahoe, tree structure-aware high performance engine ensemble. Tahoe rearranges nodes enable efficient coalesced accesses; also trees, such that with similar structures grouped together in assigned threads balanced way. Besides efficiency, we...

The performance and efficiency of distributed training Deep Neural Networks (DNN) highly depend on the gradient averaging among participating processes, a step bound by communication costs. There are two major approaches to reduce overhead: overlap communications with computations (lossless), or (lossy). lossless solution works well for linear neural architectures, e.g. VGG, AlexNet, but more recent networks such as ResNet Inception limit opportunity overlapping. Therefore, that amount data...

In this paper, we study joint antenna activity detection, channel estimation, and multiuser detection for massive multiple-input multiple-output (MIMO) systems with general spatial modulation (GSM). We first establish a double-sparsity MIMO model by considering the sparsity of signal GSM. Based on model, formulate blind problem. To solve problem, develop message-passing based channel-and-signal estimation (BCSE) algorithm. The BCSE algorithm basically follows affine sparse matrix...

We study over-the-air model aggregation in federated edge learning (FEEL) systems, where channel state information at the transmitters (CSIT) is assumed to be unavailable. leverage <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">reconfigurable intelligent surface</i> (RIS) technology align cascaded coefficients for xmlns:xlink="http://www.w3.org/1999/xlink">CSIT-free</i> aggregation. To this end, we jointly optimize RIS and receiver by...

Classical graph matching aims to find a node correspondence between two unlabeled graphs of known topologies.This problem has wide range applications, from identities in social networks identifying similar biological network functions across species.However, when the underlying are unknown, use conventional methods requires inferring topologies first, process that is highly sensitive observation errors.In this paper, we tackle blind with unknown directly using observations signals, which...

In this paper, we propose a novel gender bias detection method by utilizing attention map for transformer-based models. We 1) give an intuitive judgement comparing the different relation degree between genders and occupation according to scores, 2) design detector modifying module, 3) insert into positions of model present internal flow, 4) draw consistent conclusion scanning entire Wikipedia, BERT pretraining dataset. observe that matrices, Wq Wk introduce much more than other modules...

Transformer-based deep learning models have become a ubiquitous vehicle to drive variety of Natural Language Processing (NLP) related tasks beyond their accuracy ceiling. However, these also suffer from two pronounced challenges, that is, gigantic model size and prolonged turnaround time. To this end, we introduce ET. r<u>E</u>-thinks self-attention computation for <u>T</u>ransformer on GPUs with the following contributions: First, novel architecture, which encompasses tailored operators...

The problem of graph matching involves finding a node correspondence between two unlabeled graphs with known topologies, which has applications in various fields such as social network analysis and species identification. In this paper, we tackle without prior knowledge the underlying using only observations signals. We assume that these signals are generated by applying filters to signal excitations. construct sample covariance matrices from match nodes based on selected eigenvectors...

Distributed learning such as federated or collaborative enables model training on decentralized data from users and only collects local gradients, where is processed close to its sources for privacy. The nature of not centralizing the addresses privacy issue privacy-sensitive data. Recent studies show that a third party can reconstruct true in distributed machine system through publicly-shared gradients. However, existing reconstruction attack frameworks lack generalizability different Deep...

Hardware acceleration of deep learning (DL) systems has been increasingly studied to achieve desirable performance and energy efficiency. The FPGA strikes a balance between high efficiency fast development cycle therefore is widely used as DNN accelerator. However, there exists an architecture-layout mismatch in the current designs, which introduces scalability flexibility issues, leading irregular routing resource imbalance problems. To address these limitations, this work, we propose FTDL,...