A5049931304- Caching and Content Delivery
- Retinopathy of Prematurity Studies
- Privacy-Preserving Technologies in Data
- Optimization and Search Problems
- Wireless Signal Modulation Classification
- Advanced Bandit Algorithms Research
- Neonatal and fetal brain pathology
- Cooperative Communication and Network Coding
- Opportunistic and Delay-Tolerant Networks
- Machine Learning and Algorithms
- Recommender Systems and Techniques
- Retinal Imaging and Analysis
- Complex Network Analysis Techniques
- Complexity and Algorithms in Graphs
- Cryptography and Data Security
- Peer-to-Peer Network Technologies
- Face and Expression Recognition
- Domain Adaptation and Few-Shot Learning
- Neonatal Respiratory Health Research
- Mobile Crowdsensing and Crowdsourcing
- Stochastic Gradient Optimization Techniques
- Auction Theory and Applications
- Mobile Ad Hoc Networks
- Speech and Audio Processing
- Advanced Neural Network Applications
Northeastern University
2016-2025
Universidad del Noreste
2017-2024
Boston University
2020-2021
Dana (United States)
2019
Mater Dei Hospital
2018
University Hospitals Sussex NHS Foundation Trust
2018
Yahoo (United Kingdom)
2015
Yahoo (United States)
2011-2015
Technicolor (Germany)
2011-2014
Technicolor (United States)
2011-2014
Retinopathy of prematurity (ROP) is a leading cause childhood blindness worldwide. The decision to treat primarily based on the presence plus disease, defined as dilation and tortuosity retinal vessels. However, clinical diagnosis disease highly subjective variable.To implement validate an algorithm deep learning automatically diagnose from photographs.A convolutional neural network was trained using data set 5511 photographs. Each image previously assigned reference standard (RSD) consensus...
Ridge regression is an algorithm that takes as input a large number of data points and finds the best-fit linear curve through these points. The building block for many machine-learning operations. We present system privacy-preserving ridge regression. outputs in clear, but exposes no other information about data. Our approach combines both homomorphic encryption Yao garbled circuits, where each used different part to obtain best performance. implement complete experiment with it on real...
We study the dissemination of dynamic content, such as news or traffic information, over a mobile social network. In this application, users subscribe to dynamic-content distribution service, offered by their service provider. To improve coverage and increase capacity, we assume that share any content updates they receive with other meet. make two contributions. First, determine how provider can allocate its bandwidth optimally at "fresh" possible. More precisely, define global fairness...
Advances in software defined radio (SDR) technology allow unprecedented control on the entire processing chain, allowing modification of each functional block as well sampling changes input waveform. This article describes a method for uniquely identifying specific among nominally similar devices using combination SDR sensing capability and machine learning (ML) techniques. The key benefit this approach is that ML operates raw I/Q samples distinguishes only transmitter hardware-induced...
Recommender systems typically require users to reveal their ratings a recommender service, which subsequently uses them provide relevant recommendations. Revealing has been shown make susceptible broad set of inference attacks, allowing the learn private user attributes, such as gender, age, etc. In this work, we show that can profile items without ever learning provide, or even they have rated. We by designing system performs matrix factorization, popular method used in variety modern...
This paper describes the architecture and performance of ORACLE, an approach for detecting a unique radio from large pool bit-similar devices (same hardware, protocol, physical address, MAC ID) using only IQ samples at layer. ORACLE trains convolutional neural network (CNN) that balances computational time accuracy, showing 99% classification accuracy 16-node USRP X310 SDR testbed external database >100 COTS WiFi devices. Our work makes following contributions: (i) it studies...
Radio fingerprinting uniquely identifies wireless devices by leveraging tiny hardware-level imperfections inevitably present in off-the-shelf radio circuitry. This way, can be directly identified at the physical layer analyzing unprocessed received waveform - thus avoiding energy-expensive upper-layer cryptography that resource-challenged embedded may not able to afford. Recent advances have proven convolutional neural networks (CNNs) thanks their multidimensional mappings achieve accuracy...
RF fingerprinting is a key security mechanism that allows device identification by learning unchanging, hardware-based characteristics of the transmitter. In this article, we demonstrate how machine techniques impact analyzing dataset 400 GB in-phase (I) and quadrature (Q) signal data transmitted 10,000 radios. Our deep convolutional neural network architectures take raw processed IQ samples as input to identify devices under variety practical conditions, including changing channels, noise...
Due to the unprecedented scale of Internet Things, designing scalable, accurate, energy-efficient and tamper-proof authentication mechanisms has now become more important than ever. To this end, in paper we present ORACLE, a novel system based on convolutional neural networks (CNNs) “fingerprint” (i.e., identify) unique radio from large pool devices by deep-learning fine-grained hardware impairments imposed circuitry physical-layer I/Q samples. First, show how hardware-specific imperfections...
User demographics, such as age, gender and ethnicity, are routinely used for targeting content advertising products to users. Similarly, recommender systems utilize user demographics personalizing recommendations overcoming the cold-start problem. Often, privacy-concerned users do not provide these details in their online profiles. In this work, we show that a system can infer of with high accuracy, based solely on ratings provided by (without additional metadata), relatively small number...
Background Prior work has demonstrated the near-perfect accuracy of a deep learning retinal image analysis system for diagnosing plus disease in retinopathy prematurity (ROP). Here we assess screening potential this scoring by determining its ability to detect all components ROP diagnosis. Methods Clinical examination and fundus photography were performed at seven participating centres. A was trained disease, generating quantitative assessment vascular abnormality (the i-ROP score) on 1–9...
We propose introducing modern parallel programming paradigms to secure computation, enabling their execution on large datasets. To address this challenge, we present Graph SC, a framework that (i) provides paradigm allows non-cryptography experts write code, (ii) brings parallelism such implementations, and (iii) meets the need for obliviousness, thereby not leaking any private information. Using developers can efficiently implement an oblivious version of graph-based algorithms (including...
Radio fingerprinting provides a reliable and energy-efficient IoT authentication strategy by leveraging the unique hardware-level imperfections imposed on received wireless signal transmitter's radio circuitry. Most of existing approaches utilize hand-tailored protocol-specific feature extraction techniques, which can identify devices operating under pre-defined protocol only. Conversely, mapping inputs onto very large space, deep learning algorithms be trained to fingerprint populations any...
Using medical images to evaluate disease severity and change over time is a routine important task in clinical decision making. Grading systems are often used, but unreliable as domain experts disagree on category thresholds. These discrete categories also do not reflect the underlying continuous spectrum of severity. To address these issues, we developed convolutional Siamese neural network approach at single points between longitudinal patient visits spectrum. We demonstrate this two...
Unpredictable and potentially dangerous aggressive behavior by youth with Autism Spectrum Disorder (ASD) can isolate them from foundational educational, social, familial activities, thereby markedly exacerbating morbidity costs associated ASD. This study investigates whether preceding physiological motion data measured a wrist‐worn biosensor predict aggression to others We recorded peripheral (cardiovascular electrodermal activity) (accelerometry) signals worn 20 ASD (ages 6–17 years, 75%...
RF fingerprinting involves identifying characteristic transmitter-imposed variations within a wireless signal. Deep neural networks (DNNs) that do not rely on handcrafting features have proven to be remarkably effective in tasks, as long the channel remains invariant. However, DNNs trained at specific location and time perform poorly datasets collected under different conditions. This article proposes data augmentation step training pipeline exposes DNN many simulated noise are present...
The universal availability of unmanned aerial vehicles (UAVs) has resulted in many applications where the same make/model can be deployed by multiple parties. Thus, identifying a specific UAV given swarm, manner that cannot spoofed software methods, becomes important. We propose RF fingerprinting for this purpose, neural network learns subtle imperfections present transmitted waveform. For UAVs, constant hovering motion raises key challenge, which remains fundamental problem previous works...
With the recent surge in autonomous driving vehicles, need for accurate vehicle detection and tracking is critical now more than ever. Detecting vehicles from visual sensors fails non-line-of-sight (NLOS) settings. This can be compensated by inclusion of other modalities a multi-domain sensing environment. We propose several deep learning based frameworks fusing different (image, radar, acoustic, seismic) through exploitation complementary latent embeddings, incorporating multiple...
We study the problem of optimal content placement over a network caches, naturally arising in several networking applications, including ICNs, CDNs, and P2P systems. Given demand request rates paths followed, we wish to determine that maximizes expected caching gain, i.e., reduction routing costs due intermediate caching. The offline version this is NP-hard and, general, topology may be priori unknown. Hence, distributed, adaptive, constant approximation algorithm desired. show path...
Beam selection for millimeter-wave links in a vehicular scenario is challenging problem, as an exhaustive search among all candidate beam pairs cannot be assuredly completed within short contact times. We solve this problem via novel expediting by leveraging multimodal data collected from sensors like LiDAR, camera images, and GPS. propose individual modality distributed fusion-based deep learning (F-DL) architectures that can execute locally well at mobile edge computing center (MEC), with...
Sharing content over a mobile network through opportunistic contacts has recently received considerable attention.