Middleboxes, such as firewalls, Network Address Translators (NATs), Wide Area Network (WAN) optimizers, or Deep Packet Inspector (DPIs), are widely deployed in modern networks to improve network security and performance. Traditional middleboxes are typically hardware based, which are expensive and closed systems with little extensibility. Furthermore, they are developed by different vendors and deployed as standalone devices with little scalability. As the development of networks in scale, the limitations of traditional middleboxes bring great challenges in middlebox deployments. Network Function Virtualization (NFV) technology provides a promising alternative, which enables flexible deployment of middleboxes, as virtual machines (VMs) running on standard servers. However, the flexibility also creates a challenge for efficiently placing such middleboxes, due to the availability of multiple hosting servers, capabilities of middleboxes to change traffic volumes, and dependency between middleboxes. In our first two work, we addressed the optimal placement challenge of NFV middleboxes by considering middlebox traffic changing effects and dependency relations. Since each VM has only a limited processing capacity restricted by its available resources, multiple instances of the same function are necessary in an NFV network. Thus, routing in an NFV network is also a challenge to determine not only via a path from the source to destination but also the service (middlebox) locations. Furthermore, the challenge is complicated by the traffic changing effects of NFV services and dependency relations between them. In our third work, we studied how to efficiently route a flow to receive services in an NFV network. We conducted large-scale simulations to evaluate our proposed solutions, and also implemented a Software-Defined Networking (SDN) based prototype to validate the solutions in realistic environments. Extensive simulation and experiment results have been fully demonstrated the effectiveness of our design.

<p>This thesis presents research that created an application protocol for wireless networks and functional brain spectroscopy. The proposed protocol was tested through the integration of several types of networks, devices, and sensors to facilitate functional brain spectroscopy. The need for reliability and speed to transmit medical data in near real time can make medical application uniquely challenging. This research addresses one of the main challenges faced when building medical solutions that monitor the human brain. The findings proposed a protocol that was implemented using an architectural model for a solution that provides full mobility in an everyday environment using a near-infrared light sensor designed to monitor brain function in humans. Moreover the study showed it is possible to use heterogenic networks and heterogenic devices to provide useful data that can be used for medical purposes [103]. A system implemented the proposed protocol was built to allow the possibility of testing subjects to be monitored in their real environment [104]. To test this hypothesis, heterogenic communication software was developed to allow for the collection of physiological data from a mobile near-infrared sensor via a mobile telephone that had Bluetooth support and Global Standard for Mobile Communications (GSM) support. The result of this work introduced Medical Data Transfer Protocols (MDTP) [106], an algorithm [103], and an architectural model (Four-node Model) [105].</p> / Doctor of Philosophy (PhD)

With the present need for on the move networking, innovative technologies strive to establish a technological basis for managing secure and reliable systems in a highly interconnected information enabled world, and prevent reliance on a fixed networking infrastructure, hence the implementation of ad-hoc networks. There are numerous applications where ad-hoc networks are deployed including military, tele-health and mobile education. As such the need for security is imperative. Not much research work has been invested in the area of trust in ad hoc networks which proves to be a challenging subject relative to the characteristics of these types of networks. The objective of this thesis was to develop a model for trust computation between the nodes of the network. Eventually, the confidence level for each node was quantified, which lead to a better constancy among the nodes. Therefore, communication was trust worthy, and malicious nodes were punished and secluded from the network.

The recent advancement in computational and communication systems has led to the introduction of high-performing neural networks and high-speed wireless vehicular communication networks. As a result, new technologies such as cooperative perception and cognition have emerged, addressing the inherent limitations of sensory devices by providing solutions for the detection of partially occluded targets and expanding the sensing range. However, designing a reliable cooperative cognition or perception system requires addressing the challenges caused by limited network resources and discrepancies between the data shared by different sources. We examine the requirements, limitations, and performance of different cooperative perception techniques, and present an in-depth analysis of the notion of Deep Feature Sharing (DFS). We explore different cooperative object detection designs and evaluate their performance in terms of average precision. We use the Volony dataset for our experimental study. The results confirm that the DFS methods are significantly less sensitive to the localization error caused by GPS noise. Furthermore, the results attest that detection gain of DFS methods caused by adding more cooperative participants in the scenes is comparable to raw information sharing technique while DFS enables flexibility in design toward satisfying communication requirements. Furthermore, in the environments where there is noise in GPS positioning estimates, cooperative perception performance will decrease. To alleviate the performance decrease we introduce a method to estimate the relative positioning of cooperative vehicles by comparing feature maps extracted from LIDAR observations of the cooperative vehicles. The results show that GPS positioning estimates of all participating vehicles will be improved as the number of cooperative vehicles increases in the scene.

Driven by the ever-increasing data volume through the Internet, the per-port speed of network devices reached 400 Gbps, and high-end switches are capable of processing 25.6 Tbps of network traffic. To improve the efficiency and security of the network, network traffic measurement becomes more important than ever. For fast and accurate traffic measurement, managing an accurate working set of active flows (WSAF) at line rates is a key challenge. WSAF is usually located in high-speed but expensive memories, such as TCAM or SRAM, and thus their capacity is quite limited. To scale up the per-flow measurement, we pursue three thrusts. In the first thrust, we propose to use In-DRAM WSAF and put a compact data structure (i.e., sketch) called FlowRegulator before WSAF to compensate for DRAM's slow access time. Per our results, FlowRegulator can substantially reduce massive influxes to WSAF without compromising measurement accuracy. In the second thrust, we integrate our sketch into a network system and propose an SDN-based WLAN monitoring and management framework called RFlow+, which can overcome the limitations of existing traffic measurement solutions (e.g., OpenFlow and sFlow), such as a limited view, incomplete flow statistics, and poor trade-off between measurement accuracy and CPU/network overheads. In the third thrust, we introduce a novel sampling scheme to deal with the poor trade-off that is provided by the standard simple random sampling (SRS). Even though SRS has been widely used in practice because of its simplicity, it provides non-uniform sampling rates for different flows, because it samples packets over an aggregated data flow. Starting with a simple idea that "independent per-flow packet sampling provides the most accurate estimation of each flow," we introduce a new concept of per-flow systematic sampling, aiming to provide the same sampling rate across all flows. In addition, we provide a concrete sampling method called SketchFlow, which approximates the idea of the per-flow systematic sampling using a sketch saturation event.

In this Dissertation, we have investigated the underlying theories of probabilistic models for application in large scale machine learning tasks. First, we introduce the maximum probability theorem and its consequences. We present a theoretical framework of probabilistic learning derived from the Maximum Probability (MP) Theorem. In this probabilistic framework, a model is defined as an event in the probability space, and a model or the associated event - either the true underlying model or the parameterized model - have a quantified probability measure. This quantification of a model's probability measure is derived from the MP Theorem, where we have shown that an event's probability measure has an upper-bound given its conditional distribution on an arbitrary random variable. Through this alternative framework, the notion of model parameters is encompassed in the definition of the model or the associated event. Therefore, this framework deviates from the conventional approach of assuming a prior on the model parameters. Instead, the regularizing effects of assuming prior over parameters are imposed through maximizing probabilities of models or according to information theory, minimizing the information content of a model. The probability of a model in MP framework is invariant to reparameterization and is solely dependent on the model's likelihood function. Also, rather than maximizing the posterior in a conventional Bayesian setting, the objective function in our alternative framework is defined as the probability of set operations (e.g. intersection) on the event of the true underlying model and the event of the model at hand. The MP framework adds clarity to probabilistic learning through solidifying the definition of probabilistic models, quantifying their probabilities, and providing a visual understanding of objective functions. Furthermore, we discuss Finite "K"onvolutional Neural Networks (FKNN) as a step towards constructing a discrete counterpart to Convolutional Neural Networks (CNN). In FKNNs, the linear and non-linear components of the network are naturally derived and justified in terms of Bayes' Theorem. The building blocks of our network are classifiers operating on the domain of categorical distributions. This property enables the composition of Bayesian classifiers to construct more expressive models. The resulting composite model consists of linear and non-linear components, which are remarkably similar to modern CNNs and their variations, yet the roles of parameters, variables, and layers are less ambiguous from a statistical perspective. Parameters and variables represent categorical distributions in FKNNs, providing the potential for usage of statistical and information-theoretical methods. We further introduce two methods of parameter initialization, inspired by the natural parameterization of categorical distribution and the Jeffreys priors. Finally, we transform some well-known CNN architectures for image classification task into their FKNN counterparts and compare their performance. Experimental results show that the FKNNs and their corresponding CNN architecture exhibit comparable performances. The functional similarity of CNNs and FKNNs, the empirical results, and the explicit connection of FKNNs and Bayes' rule encourage the investigation of finite-state probabilistic models.

As the Internet continues to expand, the global default-free zone (DFZ) forwarding table has begun to grow faster than hardware can economically keep pace with. Various policies are in place to mitigate this growth rate, but current projections indicate policy alone is inadequate. As such, a number of technical solutions have been proposed. This work builds on many of these proposed solutions, and furthers the debate surrounding the resolution to this problem. It discusses several design decisions necessary to any proposed solution, and based on these tradeoffs it proposes a Hierarchical Inter-Domain Routing Architecture - HIDRA, a comprehensive architecture with a plausible deployment scenario. The architecture uses a locator/identifier split encapsulation scheme to attenuate both the immediate size of the DFZ forwarding table, and the projected growth rate. This solution is based off the usage of an already existing number allocation policy - Autonomous System Numbers (ASNs). HIDRA has been deployed to a sandbox network in a proof-of-concept test, yielding promising results.

One challenge associated with developing location-based social applications for iOS devices is building a framework on top of Apple’s Core Bluetooth Framework to drive user discovery. Many applications on Apple’s App Store use Bluetooth to enable location-based user discovery. These Social Location-Based Frameworks are private and often are lacking. An ideal Social Location-Based Bluetooth Framework would be public, would be responsive while the application is minimized, have a light battery footprint, and securely transfer the necessary data to enable social interaction. Samantha, a Social Location-Based Framework, meets all of these characteristics. In a Test Application, Samantha took no more than 5 seconds to start up and discover all nearby users. The average discovery time was 3.5 seconds. In addition the battery draw, measured using Apple’s Battery Monitoring Instrument, during testing never exceed 1/20. This means an application running Samantha in the background for a 12-hour period would not significantly drain the battery. In terms of security, Samantha transfers a Unique User Identifier (UUID) across Bluetooth than contains no sensitive information. This UUID, a string of random characters, contains no personal information and it is only useful because it allows specific identification of a nearby user in a database holding additional information. This two-step process ensures that confidential information is never exposed. An example application, Ripple, uses Samantha to create a location-based social application and highlights the framework’s intended use.

This thesis project presents a MATLAB based application which is designed to synthesize any arbitrary stable transfer function. Our application is based on the Cauer synthesis procedure. It has an interactive front which allows inputs either in the form of residues and poles of a transfer function, in the form of coefficients of the numerator and denominator of the transfer impedance or in the form of samples of an impulse response. The program synthesizes either a single or double resistively terminated LC ladder network. Our application displays a chart showing the variation of stability of an impulse response with the addition of delay. An attempt is made to synthesize usually unstable impulse responses by calculating the delay that would make them stable.

Smartphones now account for the majority of all cell phones in use today [23]. Ubiquitous Internet access is a valuable feature offered by these devices and the vast majority of smartphone applications make use of the Internet in one way or another. However, the bandwidth offered by these cellular networks is often much lower than we typically experience on our standard home networks, leading to a less-than-optimal user experience. This makes it very challenging and frustrating to access certain types of web content such as video streaming, large file downloads, loading large webpages, etc. Given that most modern smartphones are multi-homed and are capable of ac- cessing multiple networks simultaneously, this thesis attempts to utilize all available network interfaces in order to achieve the aggregated bandwidth of each to improve the overall network performance of the phone. To do so, I implement a bandwidth aggregation system for iOS that combines the bandwidths of multiple devices located within close proximity of each other. Deployed on up to three devices, speedups of up to 1.82x were achieved for downloading a single, 10mb file. Webpage loading saw speedups of up to 1.55x.