Search

This paper presents an accurate approach to predict the probability of line-of-sight for unmanned aerial vehicle (UAV) communications. We present a new numerical approach to calculate the probability of line-of-sight as a function of UAV height and distance to a ground based receiver, using a two- dimensional model of building and UAV locations. We use this numerical approach to calculate the probability of line-of-sight as a function of elevation angle. We also provide closed-form formulas for the probability of LoS as a function of elevation angle. We show that our approaches predict the probability of line-of-sight more accurately than the existing approaches.

This paper presents a new clustering algorithm inspired by Newtonian gravity that iteratively groups data and eliminates outliers. In particular, we impose a grid over the region of interest and define a particle with data-dependent mass for each grid square. We then calculate a Newtonian inspired force on each of the particles and move them in the direction of the force. We repeat the process until there is no further movement. We compare performance with existing algorithms and show that in cases of medium to high clutter, our algorithm has an order of magnitude lower estimation error.

This paper considers localization using a noncoherent bi-static radar with multiple receivers. In particular, we consider the multiple target scenario, where the targets are located behind a wall. We present an algorithm that eliminates ambiguities that arise in this case by iteratively enforcing a set of constraints. We also present a clustering algorithm to further improve performance by making use of multiple sets of measurements. The results are compared with the CRLB. Location accuracy is shown to be in the order of 10cm.

With huge amounts of data being generated from almost everywhere, our universe has become data-driven. Decision making, risk prevention and mitigation, and systems assessment will not be as effective as desired without having the right data. The projected impacts and benefits of Vehicular Ad Hoc Networks (VANETs) are the driving forces for researchers to develop and further enhance VANET technology. One of the challenging and imperative issues in VANETs research is the unavailability of data. To the best of our knowledge, in this research, we are the first to create a VANET traffic dataset by using real-life traffic data. We massage the data by applying VANET human behavioral model. We experiment and validate our dataset by focusing on traffic congestion prediction. Traffic congestion can be determined by traffic density and average speed at any given point. Highly dense roads are the basic definition of congestion resulting in lower speeds of moving vehicles. We develop three time-series models ARIMA, BATS, TBATS, and a neural network model and apply them to our created VANET data to analyze and predict the total number of nodes in a cluster (density) and the average speed of the nodes. We have validated these time series prediction models by comparing the four developed models in terms of MSE, MAE, MAPE, and MASE. The created dataset and developed models can assist in predicting cluster density and average node speed to detect congestion, which will enhance route navigation.

The issue of cybersecurity is becoming more and more serious every day at all levels and in all domains. Cyber-attacks threaten the national security of every country and nation. Furthermore, cyber-attacks can significantly harm the economy. With the rapid and continuous growth of the cyber-universe, more software is being created, more data is being generated, and cybersecurity breaches and defense strategies are getting more complex. For such a problem, considering the size and complexity of the cyber-universe, big data mining techniques and advanced machine learning solutions will be most suitable to use for predicting brand-new attacks. This is because traditional machine learning methods would not help combat today’s cybersecurity issues. Anomaly-based intrusion detection systems are receiving tremendous attention nowadays. This is because of the vast improvement and development in big data solutions. This paper utilizes highly imbalanced real-life benchmark network traffic datasets of multiple types of attacks. After resolving the class imbalance issue in our datasets by applying oversampling approach, our study becomes twofold. First, we are building prediction models for each type of attacks separately and optimizing the model with the highest accuracy. Then, we build a prediction model for all attacks together using deep learning with the smallest number of features and we optimize the model to achieve the highest accuracy. Our developed model can accurately predict the threat and the type of attack.