No description available.

No description available.

An analysis of snow removal injury data from the National Electronic Injury Surveillance System (NEISS) revealed a persistent gender gap in injuries and deaths during snow clearing activities. In general, men, those who identified as White and those aged 60-79 represented the vast majority of injuries and deaths sustained during automated snow removal. Injuries and deaths from manual snow clearing had greater representation across gender lines, as well as across various age groups and race categories. This indicates that a greater cross-section of society relies on the standard shovel in comparison to the snow blower for snow removal. The most likely injuries sustained during shoveling were to the neck and back, while hand and finger injuries were far more common during the use of a snow blower. Similar percentages of cardiac (30%) and non-cardiac chest injuries (70%) were found for both manual and automated modes of snow removal. While the majority of cardiac chest injuries were in those aged 40-59 for shoveling and 60-79 for snow blowing, the majority of cardiac fatalities were in those aged 60-79 for both methods of snow removal. Daily all-cause mortality and daily deaths from acute heart attacks showed a weak but inverse relationship to daily maximum, minimum and average temperatures. Mortality related to temperatures had significant lag effects for two days. Daily all-cause and heart attack mortality were also significantly related to the depth of the existing snowpack. Snow to liquid ratios indicating differences between heavy, wet snow and dry, powdery snow were not significant. However, the water equivalent of the existing snowpack was significantly related to daily mortality. Comparisons between all age and elderly mortality showed weaker and opposite relationships for the elderly group suggesting the use of protective behaviors such as cold and snow avoidance.

Academic institutes' networks, like commercial networks, have confidential and valuable information that attracts hackers. From 6 October 2000 to 29 March 2001, the authors collected data on possible attacks and probes against East Tennessee State University's campus network. The number of suspicious activities detected daily varied from 200,000 to more than 2,000,000, with ICMP-based attacks accounting for more than 81% of all attacks. While ICMP-based attacks were reasonably harmless, these activities as a whole depleted network bandwidth significantly. Severe attacks were detected daily. Port scans and host scans that involving 2 or more /24 subnets were detected every week. Attacks and probes were distributed throughout a typical day and week. Our research results suggested policy makers in academic institutions like ETSU should adopt standard measures to secure campus networks, including firewalls, intrusion detection systems, server management, and risk assessment.

Graph theoretic modeling has served as an invaluable tool for solving a variety of problems since its introduction in Euler's paper on the Bridges of Königsberg in 1736 . Two amongst them of contemporary interest are the modeling of Redundant Arrays of Inexpensive Disks (RAID), and the identification of network attacks. While the former is vital to the protection and uninterrupted availability of data, the latter is crucial to the integrity of systems comprising networks. Both are of practical importance due to the continuing growth of data and its demand at increasing numbers of geographically distributed locations through the use of networks such as the Internet. The popularity of RAID has soared because of the enhanced I/O bandwidths and large capacities they offer at low cost. However, the demand for bigger capacities has led to the use of larger arrays with increased probability of random disk failures. This has motivated the need for RAID systems to tolerate two or more disk failures, without sacrificing performance or storage space. To this end, we shall first perform a comparative study of the existing techniques that achieve this objective. Next, we shall devise novel graph-theoretic algorithms for placing data and parity in arrays of n disks (n ≥ 3) that can recover from two random disk failures, for n = p - 1, n = p and n = 2p - 2, where p is a prime number. Each shall be shown to utilize an optimal ratio of space for storing parity. We shall also show how to extend the algorithms to arrays with an arbitrary number of disks, albeit with non-optimal values for the aforementioned ratio. The growth of the Internet has led to the increased proliferation of malignant applications seeking to breach the security of networked systems. Hence, considerable effort has been focused on detecting and predicting the attacks they perpetrate. However, the enormity of the Internet poses a challenge to representing and analyzing them by using scalable models. Furthermore, forecasting the systems that they are likely to exploit in the future is difficult due to the unavailability of complete information on network vulnerabilities. We shall present a technique that identifies attacks on large networks using a scalable model, while filtering for false positives and negatives. Furthermore, it also forecasts the propagation of security failures proliferated by attacks over time and their likely targets in the future.

Deep Neural Networks (DNNs) have become ubiquitous due to their performance on prediction and classification problems. However, they face a variety of threats as their usage spreads. Model extraction attacks, which steal DNN models, endanger intellectual property, data privacy, and security. Previous research has shown that system-level side channels can be used to leak the architecture of a victim DNN, exacerbating these risks. We propose a novel DNN architecture extraction attack, called EZClone, which uses aggregate rather than time-series GPU profiles as a side-channel to predict DNN architecture. This approach is not only simpler, but also requires less adversary capability than earlier works. We investigate the effectiveness of EZClone under various scenarios including reduction of attack complexity, against pruned models, and across GPUs with varied resources. We find that EZClone correctly predicts DNN architectures for the entire set of PyTorch vision architectures with 100\% accuracy. No other work has shown this degree of architecture prediction accuracy with the same adversarial constraints or using aggregate side-channel information. Prior work has shown that, once a DNN has been successfully cloned, further attacks such as model evasion or model inversion can be accelerated significantly. Then, we evaluate several mitigation techniques against EZClone, showing that carefully inserted dummy computation reduces the success rate of the attack.

The quality of the flow over a 75 degree-sweep delta wing was documented for steady angles of attack and during dynamic maneuvers with and without the use of two control surfaces. The three-dimensional velocity field over a delta wing at a steady angle of attack of 38 degrees and Reynolds number of 72,000 was mapped out using laser-Doppler velocimetry over one side of the wing. The three-dimensional streamline and vortex line distributions were visualized. Isosurfaces of vorticity, planar distributions of helicity and all three vorticity components, and the indicator of the stability of the core were studied and compared to see which indicated breakdown first. Visualization of the streamlines and vortex lines near the core of the vortex indicate that the core has a strong inviscid character, and hence Reynolds number independence, upstream of breakdown, with viscous effects becoming more important downstream of the breakdown location. The effect of cavity flaps on the flow over a delta wing was documented for steady angles of attack in the range 28 degrees to 42 degrees by flow visualization and surface pressure measurements at a Reynolds number of 470,000 and 1,000,000, respectfully. It was found that the cavity flaps postpone the occurrence of vortex breakdown to higher angles of attack than can be realized by the basic delta wing. The effect of continuously deployed cavity flaps during a dynamic pitch-up maneuver of a delta wing on the surface pressure distribution were recorded for a reduced frequency of 0.0089 and a Reynolds number of 1,300,000. The effect of deploying a set of cavity flaps <u>during</u> a dynamic pitch-up maneuver on the surface pressure distribution was recorded for a reduced frequency of 0.0089 and a Reynolds number of 1,300,000 and 187,000. The active deployment of the cavity flaps was shown to have a short-lived beneficial effect on the surface pressure distribution. The effect on the surface pressure distribution of the varying the reduced frequency at constant Reynolds number for a plain delta wing was documented in the reduced frequency range of 0.0089 to 0.0267. The effect of the active deployment of an apex flap <u>during</u> a pitch-up maneuver on the surface pressure distribution at Reynolds numbers of 532,000, 1,000,000, and 1,390,000 were documented with reduced frequencies of 0.0053 to 0.0114 with flap deployment locations in the range of 21° to 36° . The apex flap deployment was found to have a beneficial effect on the surface pressure distribution during the maneuver and in the post-stall regime after the maneuver is completed. / Ph. D.

There has been tremendous progress in machine learning and specifically deep learning in the last few decades. However, due to some inherent nature of deep neural networks, many questions regarding explainability and robustness still remain open. More specifically, as deep learning models are shown to be brittle against malicious changes, when do the models fail and how can we construct a more robust model against these types of attacks are of high interest. This work tries to answer some of the questions regarding explainability and robustness of deep learning by tackling the problem at four different topics. First, real world datasets often contain noise which can badly impact classification model performance. Furthermore, adversarial noise can be crafted to alter classification results. Geometric multi-resolution analysis (GMRA) is capable of capturing and recovering manifolds while preserving geomtric features. We showed that GMRA can be applied to retrieve low dimension representation, which is more robust to noise and simplify classification models. Secondly, I showed that adversarial defense in the image domain can be partially achieved without knowing the specific attacking method by employing preprocessing model trained with the task of denoising. Next, I tackle the problem of adversarial generation in the text domain within the context of real world applications. I devised a new method of crafting adversarial text by using filtered unlabeled data, which is usually more abundant compared to labeled data. Experimental results showed that the new method created more natural and relevant adversarial texts compared with current state of the art methods. Lastly, I presented my work in referring expression generation aiming at creating a more explainable natural language model. The proposed method decomposes the referring expression generation task into two subtasks and experimental results showed that generated expressions are more comprehensive to human readers. I hope that all the approaches proposed here can help further our understanding of the explainability and robustness deep learning models.

Connection setup in software-defined networks (SDN) requires considerable amounts of processing, communication, and memory resources. Attackers can target SDN controllers defense mechanism based on a proof-of-work protocol. This thesis proposes a new protocol to protect controllers against such attacks, shows implementation of the system and analyze the its performance. The key characteristics of this protocol, namely its one-way operation, its requirement for freshness in proofs of work, its adjustable difficulty, its ability to work withmultiple network providers, and its use of existing TCP/IP header fields, ensure that this approach can be used in practice.

On July 18, 1994, the Asociación Mutual Israelita Argentina (AMIA), the Jewish Community Center in Buenos Aires, was bombed in what has been called the worst terrorist attack on Argentina in history. The bombing killed an estimated 86 people and wounded over 200. The Argentine government began a judicial investigation and vowed to bring the perpetrators to justice. Twenty years later, the case remains unresolved. The investigation has been marked by inefficiency and allegations of corruption. The recent, suspicious death of the lead prosecutor of the case has further called into question the legitimacy of Argentina’s government and justice system. This thesis examines the Argentine government’s response to the AMIA bombing in the context of Argentine politics. This thesis discusses Argentina’s democratic stability, political performance, history of corruption, and economic situation as key factors in understanding the government’s response to the AMIA attack and investigation. Argentina’s response does not fit within the established models of a democracy’s response to terrorism. Argentina’s response, or lack thereof, to this terrorist attack is perplexing, but this surprising outcome may be explained by the country’s political problems. The findings of this thesis support the conclusion that the Argentine government’s response to the AMIA case can be attributed to its functioning but flawed democracy and faulty political performance. Argentina has a history of political and police corruption and a weak judicial branch, which has hindered the effectiveness of the justice system and complicated the AMIA investigation. The extent to which these factors have allowed corruption and economic interests to derail the investigation is still debated and offers an area for future research.