Smart communications devices are giving users instant access to applications that consume large amounts of data. These applications have different requirements on the network for delivery of data. In order to support these different applications, operators are required to support multiple service classes. Given the regulatory and technology constraints and the relatively high cost associated with wireless spectrum licensing and utilization, demand will exceed supply leading to congestion and overload conditions. In addition to new broadband radio technologies offering higher data rates, operators are looking at deploying alternate heterogeneous technologies, such as WLAN, to provide additional bandwidth for serving customers. It is expected that this will still fall short of providing enough network resources to meet the ITU requirement for 1% new call blocking probability. An economic mechanism that offers incentives to individuals for rational behavior is required in order in order to reduce the demand for network resources and resolve the congestion problem. The research in this dissertation demonstrates that the integration of a dynamic pricing with connection admission control mechanism for an operator deploying cooperative heterogeneous networks (e.g., LTE and WLAN) offering multiple QoS service classes reduces the new call blocking probability to the required 1% level. The experimental design consisted, first, of an analytical model of the CAC algorithm with dynamic pricing in a heterogeneous environment. The analytical model was subsequently validated through discrete-event simulation using Matlab.

The iOS 7 Core Bluetooth Framework (CB) is finally at a level where it can be used in projects to create valuable iOS applications. Due to its maximum broadcast radius of 30ft it lends itself to nearby communication. This thesis explores the Bluetooth space generally before delving into location-based data transfer using Bluetooth. The CB provided by Apple is powerful, but somewhat cumbersome. It forces the developer to deal with details related to device discovery and connections which can be tangential to the goal of the developer: sending data between devices. I built the Lightweight Bluetooth Framework (LBF) which makes the features of the CB more accessible by abstracting away from the implementation details of CB. LBF supports any number of data types being transferred as well as any number of total pieces of data regardless of data type. The system accomplishes this by assigning specific Characteristics to a particular data type and having pieces of data be uniquely identified with an ID when they are broadcasted. This unique ID is then used to associated the proper object with the received data. This will enable developers to focus on the implementation of their App without getting hung up on the details of the CB. Additionally, benchmark tests are done on the Lightweight Bluetooth Framework to determine what data transfer speed the framework supports. These tests reveal that transfer speed depends on hardware, but independent of hardware are too slow to transfer images, video, or sound.

Web servers often become overloaded with irregular surges in web traffic. Several techniques have been explored to cope with these overloads such as distributing load throughout different servers. This thesis presents Pipelined Apache HTTP Server, a modified version of the Apache Software Foundation’s HTTP Server that utilizes a pipelined execution of Apache’s request cycle. We discuss Apache’s original architecture, the modifications necessary for implementation of pipelined execution, and analyze its run time. Ultimately, we hoped to increase throughput of Apache but fall short because of unbalanced request phases and pipelining overhead.

The Internet provides a powerful medium for communication and collaboration. The ability one has to connect and interact with web-based tools from anywhere in the world makes the Internet ideal for such tasks. However, the lack of native tools can be a hindrance when deploying collaborative initiatives, as many current projects require specialized software in order to operate. This thesis demonstrates, with the comparably recent advances in browser technology and Document Object Model (DOM) implementation, a web-based collaborative annotation system can be developed that can be accessed by a user through a standards-compliant web browser. Such a system, demonstrated to work on the commonly-used web browsers constituting the vast majority of web traffic, was implemented using open-source tools and industry-recognized standards. Additionally, it accepts static copies of most standard document formats for both handwritten and typed annotations, while maintaining an archived copy of the original. The system developed for this thesis lends itself to use in a number of different process domains, as most collaborative annotation approaches can be described by a single process model. While a number of possible usage scenarios are discussed, this thesis approaches system usage only in an academic setting, focusing on applicability of the system to electronic grading and document exchange. From here, additional system usage can be easily extrapolated.

Although the nearing end of Moore’s Law has been predicted numerous times in the past, it will eventually come to pass. In forethought of this, many modern computing systems have become increasingly complex, distributed, and parallel. As software is developed on and for these complex systems, a common API is necessary for gathering vital performance related metrics while remaining transparent to the user, both in terms of system impact and ease of use. Several distributed performance monitoring and testing systems have been proposed and implemented by both research and commercial institutions. However, most of these systems do not meet several fundamental criterion for a truly useful distributed performance monitoring system: 1) variable data delivery models, 2) security, 3) scalability, 4) transparency, 5) completeness, 6) validity, and 7) portability. This work presents dCAMP: Distributed Common API for Measuring Performance, a distributed performance framework built on top of Mark Gabel and Michael Haungs’ work with CAMP. This work also presents an updated and extended set of criterion for evaluating distributed performance frameworks and uses these to evaluate dCAMP and several related works.

Computer networks were not designed with security in mind, making research into the subject of network security vital. Virtual Networks are similar to computer networks, except the components of a Virtual Network are in software rather than hardware. With the constant threat of attacks on networks, security is always a big concern, and Virtual Networks are no different. Virtual Networks have many potential attack vectors similar to physical networks, making research into Virtual Network security of great importance. Virtual Networks, since they are composed of virtualized network components, have the ability to dynamically change topologies. In this paper, we explore Virtual Networks and their ability to quickly shift their network topology. We investigate the potential use of this flexibility to protect network resources and defend against malicious activities. To show the ability of reactively shifting a Virtual Network’s topology to se- cure a network, we create a set of four experiments, each with a different dynamic topology shift, or “dynamic defense”. These four groups of experiments are called the Server Protection, Isolated Subnet, Distributed Port Group, and Standard Port Group experiments. The Server Protection experiments involve detecting an attack against a server and shifting the server behind a protected subnet. The other three sets of experiments, called Attacker Prevention experiments, involve detecting a malicious node in the internal network and initiating a dynamic de- fense to move the attacker behind a protected subnet. Each Attacker Prevention experiment utilizes a different dynamic defense to prevent the malicious node from attacking the rest of the Virtual Network. For each experiment, we run 6 different network attacks to validate the effectiveness of the dynamic defenses. The network attacks utilized for each experiment are ICMP Flooding, TCP Syn Flooding, Smurf attack, ARP Spoofing, DNS Spoofing, and NMAP Scanning. Our validation shows that our dynamic defenses, outside of the standard port group, are very effective in stopping each attack, consistently lowering the at- tacks’ success rate significantly. The Standard Port Group was the one dynamic defense that is ineffective, though there are also a couple of experiments that could benefit from being run with more attackers and with different situations to fully understand the effectiveness of the defenses. We believe that, as Virtual Networks become more common and utilized outside of data centers, the ability to dynamically shift topology can be used for network security purposes.

Satellites are playing an increasingly important role in collecting scientific information, providing communication services, and revolutionizing navigation. Until recently satellites were large and very expensive, creating a high barrier to entry that only large corporations and government agencies could overcome. In the past few years the CubeSat project at California Polytechnic University in San Luis Obispo (Cal Poly) has worked to refine the design and launching of small, lightweight, and less expensive satellites called pico-satellites, opening space up to a wider audience. Now that Cal Poly has the launch logistics and hardware under control, a new problem has arisen. These pico-satellites are within communication range of a ground station only 40 minutes a day. This, combined with their 1200 bps communication speed, limits the usefulness of the satellite missions to those only transmitting small amounts of data back to Earth. This thesis proposes a novel protocol that allows a sparse network of pico-satellites to communicate among one another and to larger satellites called data mules, which relay the information back to the ground station at much higher speeds. The data mules are able to provide higher speeds because they are larger satellites with less power constraints. This protocol makes it possible for a pico-satellite to send more data over a given amount of time with less end-to-end delay. When every satellite has large amounts of data almost three times as much aggregate data can be sent through the network, and almost five times more data can be sent if only a single satellite has large amounts of data to send. The end-to-end delay is cut almost in half when sending 1 MB of data per day per satellite and is decreased by a factor of at least three when sending large amounts of data from only one satellite.

Recent advances in computer vision tasks are mainly due to the success of large deep neural networks. The current state-of-the-art models have high computational costs during inference and suffer from a high memory footprint. Therefore, deploying these large networks on edge devices remains a serious concern. Furthermore, training these over-parameterized networks is computationally expensive and requires a longer training time. Thus, there is a demand to develop techniques that can efficiently reduce training costs and also be able to deploy neural networks on mobile and embedded devices. This dissertation presents practices like designing a lightweight network architecture and increasing network resource utilization. These solutions improve the efficiency of large networks during training and inference. We first propose an efficient micro-architecture (slim modules) to construct a light-weight Slim-CNN to predicting face attributes. Slim modules uses depthwise separable convolutions with pointwise convolutions, making them computationally efficient for embedded applications. Next, we investigate the problem of obtaining a compact pruned model from an untrained original network in a single-stage process. We introduce our RAPID framework that distills knowledge to a pruned student model from a teacher model under online settings. Next, we analyze the phenomena of inactive channels in a trained neural network. We take a deep dive into the gradient updates of these channels and discover that these channels have no weight update after a few early epochs. Thus, we present our channel regeneration technique that reinitializes batch normalization gamma values of all inactive channels. The gradient updates of these channels improve after the regeneration step, resulting in an increase in the contribution of these channels to the network performance. Finally, we introduce a method to improve computational efficiency in pre-trained vision transformers by reducing redundancy in visual data. Our method selects image windows or regions with high objectness measures, as these regions may contain an object of any class. Across all works in this dissertation, we extensively evaluate our proposed methods and demonstrate that our techniques improve the computational efficiency of deep neural networks during training and inference.

Vehicular ad hoc Networks (VANETs) allow vehicles to form a self-organized network. VANETs are likely to be widely deployed in the future, given the interest shown by industry in self-driving cars and satisfying their customers various interests. Problems related to Mobile ad hoc Networks (MANETs) such as routing, security, etc.have been extensively studied. Even though VANETs are special type of MANETs, solutions proposed for MANETs cannot be directly applied to VANETs because all problems related to MANETs have been studied for small networks. Moreover, in MANETs, nodes can move randomly. On the other hand, movement of nodes in VANETs are constrained to roads and the number of nodes in VANETs is large and covers typically large area. The following are the contributions of the thesis. Secure, authenticated, privacy preserving message dissemination in VANETs: When vehicles in VANET observe phenomena such as accidents, icy road condition, etc., they need to disseminate this information to vehicles in appropriate areas so the drivers of those vehicles can take appropriate action. When such messages are disseminated, the authenticity of the vehicles disseminating such messages should be verified while at the same time the anonymity of the vehicles should be preserved. Moreover, to punish the vehicles spreading malicious messages, authorities should be able to trace such messages to their senders when necessary. For this, we present an efficient protocol for the dissemination of authenticated messages. Incentive-based architecture for vehicular cloud: Due to the advantages such as exibility and availability, interest in cloud computing has gained lot of attention in recent years. Allowing vehicles in VANETs to store the collected information in the cloud would facilitate other vehicles to retrieve this information when they need. In this thesis, we present a secure incentive-based architecture for vehicular cloud. Our architecture allows vehicles to collect and store information in the cloud; it also provides a mechanism for rewarding vehicles that contributing to the cloud. Privacy preserving message dissemination in VANETs: Sometimes, it is sufficient to ensure the anonymity of the vehicles disseminating messages in VANETs. We present a privacy preserving message dissemination protocol for VANETs.

This thesis describes and develops two planarization algorithms for geographic routing and a geographic routing protocol for mobile ad hoc and sensor networks. As all nodes are mobile and there is no fixed infrastructure, the design of routing protocols is one of the most challenging issues in mobile ad hoc and sensor networks. In recent years, greedyface- greedy (GFG) geographic routing protocols have been widely used, which need nodes to construct planar graphs as the underlying graphs for face routing. Two kinds of planarization algorithms have been developed, idealized and realistic planarization algorithms, respectively. The idealized planarization algorithms make the ideal assumption that the original network graph is a unit-disk graph (UDG). On the other hand, the realistic planarization algorithms do not need the original network to be a UDG. We propose an idealized planarization algorithm, which constructs an Edge Constrained Localized Delaunay graph (ECLDel). Compared to the existing planarized localized Delaunay graph [42], the construction of an ECLDel graph is far simpler, which reduces the communication cost and saves the network bandwidth. We propose a Pre-Processed Cross Link Detection Protocol (PPCLDP), which generates a planar spanning subgraph of the original network graph in realistic environments with obstacles. The proposed PPCLDP outperforms the existing Cross Link Detection Protocol [32] with much lower communication cost and better convergence time. In GFG routing protocols, greedy routing may fail at concave nodes, in which case, face routing is applied to recover from the greedy routing failure. This may cause extra hops in routing in networks containing voids. We propose a Hill-Area-Restricted (HAR) routing protocol, which avoids the extra hops taken in the original GFG routing. Compared to the existing Node Elevation Ad hoc Routing [4], the proposed HAR guarantees the packet delivery and decreases the communication cost greatly.