A5061196481- Internet Traffic Analysis and Secure E-voting
- Network Security and Intrusion Detection
- Advanced Image and Video Retrieval Techniques
- Image Retrieval and Classification Techniques
- Advanced Steganography and Watermarking Techniques
- Image and Video Quality Assessment
- Advanced Malware Detection Techniques
- Machine Learning and Algorithms
- Machine Learning and Data Classification
- Domain Adaptation and Few-Shot Learning
- Robotics and Sensor-Based Localization
- Advanced Vision and Imaging
- Face and Expression Recognition
- Remote-Sensing Image Classification
- Video Analysis and Summarization
- Diabetes Management and Research
- Advanced Data Compression Techniques
- Gene expression and cancer classification
- Video Surveillance and Tracking Methods
- Computability, Logic, AI Algorithms
- Video Coding and Compression Technologies
- Color perception and design
- Hate Speech and Cyberbullying Detection
- Graph Theory and Algorithms
- Image Processing Techniques and Applications
Ariel University
2016-2021
University of Pennsylvania
2013
Hebrew University of Jerusalem
2007-2011
We present a new algorithm for robust family of Earth Mover's Distances - EMDs with thresholded ground distances. The transforms the flow-network EMD so that number edges is reduced by an order magnitude. As result, we compute magnitude faster than original algorithm, which makes it possible to on large histograms and databases. In addition, show distances have many desirable properties. First, they correspond way humans perceive Second, are outlier noise quantization effects. Third,...
Desktops and laptops can be maliciously exploited to violate privacy. There are two main types of attack scenarios: active passive. In this paper, we consider the passive scenario where adversary does not interact actively with device, but he is able eavesdrop on network traffic device from side. Most Internet encrypted thus attacks challenging. Previous research has shown that information extracted multimedia streams. This includes video title classification non HTTP adaptive streams...
This paper describes a method for robust real time pattern matching. We first introduce family of image distance measures, the "Image Hamming Distance Family". Members this are to occlusion, small geometrical transforms, light changes and non-rigid deformations. then present novel Bayesian framework sequential hypothesis testing on finite populations. Based framework, we design an optimal rejection/acceptance sampling algorithm. algorithm quickly determines whether two images similar with...
Desktops and laptops can be maliciously exploited to violate privacy. There are two main types of attack scenarios: active passive. In this paper, we consider the passive scenario where adversary does not interact actively with device, but he is able eavesdrop on network traffic device from side. Most internet encrypted thus attacks challenging. show that an external attacker identify operating system, browser application HTTP (HTTPS). To best our knowledge, first work shows this. We provide...
Abstract This study was designed to improve blood glucose level predictability and future hypoglycemic hyperglycemic event alerts through a novel patient‐specific supervised‐machine‐learning (SML) analysis of based on continuous‐glucose‐monitoring system (CGM) that needs no human intervention, minimises false‐positive alerts. The CGM data over 7 50 non‐consecutive days from 11 type‐1 diabetic patients aged 18 39 with mean HbA1C 7.5% ± 1.2% were analysed using four SML models. algorithm...
The increasing popularity of HTTP adaptive video streaming services has dramatically increased bandwidth requirements on operator networks, which attempt to shape their traffic through Deep Packet Inspection (DPI). However, Google and certain content providers have started encrypt services. As a result, operators often encounter difficulties in shaping encrypted via DPI. This highlights the need for new classification methods enable smart shaping. These will effectively estimate quality...
The increasing popularity of HTTP adaptive video streaming services has dramatically increased bandwidth requirements on operator networks, which attempt to shape their traffic through Deep Packet Inspection (DPI). However, Google and certain content providers have started encrypt services. As a result, operators often encounter difficulties in shaping encrypted via DPI. This highlights the need for new classification methods enable smart shaping. These will effectively estimate quality...
We suggest a new color distance based on two observations. First, perceptual differences were designed to be used compare very similar colors. They do not capture human perception for medium and large well. Thresholding was proposed solve the problem differences, i.e. totally different colors are always same apart. show that thresholding alone cannot improve differences. alleviate this using basic terms. Second, when is edge detection, many small distances around just noticeable difference...
This study explored the relevancy of internet discourse among football fans and real-life stadium violence. Our hypothesis implies that there are positive correlations between violent utterances in fans’ social networks prior to games The sample included 6 clubs who participated Israeli Premier League 2015–2016 season. traffic was automatically collected into MySQL database. content includes various forms discursive interactions within susceptible predicting, provoking or expressing tension,...
The increasing popularity of HTTP adaptive video streaming services has dramatically increased bandwidth requirements on operator networks, which attempt to shape their traffic through Deep Packet inspection (DPI).However, Google and certain content providers have started encrypt services.As a result, operators often encounter difficulties in shaping encrypted via DPI.This highlights the need for new classification methods enable smart shaping.These will effectively estimate quality...
Desktops and laptops can be maliciously exploited to violate privacy. In this paper, we consider the daily battle between passive attacker who is targeting a specific user against that may adversarial opponent. scenario, while tries choose best vector attack by surreptitiously monitoring victims encrypted network traffic in order identify users parameters such as Operating System (OS), browser apps. The use tools Virtual Private Network (VPN) or even change protocols protect his/her We...
We present a spectral-based method for the 2D/3D rigid registration of X-ray images to CT scan. The uses Fourier-based representation decompose six transformation parameters problem into twoparameter out-of-plane rotation and four-parameter in-plane problems. Preoperatively, set Digitally Reconstructed Radiographs (DRRs) are generated offline from in expected location ranges fluoroscopic imaging devices. Each DRR is transformed device invariant features space. Intraoperatively, few 2D...
We propose a new embedding method for single vector and pair of vectors. This enables: a) efficient classification regression functions vectors; b) approximation distance functions; c) non-Euclidean, semimetric learning. To the best our knowledge, this is first work that enables learning any general, semimetrics. That is, universal can approximate function with as high accuracy needed or without constraints. The project homepage including code at: http://www.ariel.ac.il/sites/ofirpele/ID
We propose a new embedding method which is particularly well-suited for settings where the sample size greatly exceeds ambient dimension. Our technique consists of partitioning space into simplices and then data points features corresponding to simplices' barycentric coordinates. train linear classifier in rich feature obtained from simplices. The decision boundary may be highly non-linear, though it within each simplex (and hence piecewise-linear overall). Further, our can approximate any...