This Master’s thesis documents and interrogates networks of regional interaction in southwest Asia and Anatolia during the Neolithic and Chalcolithic periods (12,000 - 5700 B.P.) by comparing the variable use of obsidian raw material variants at 151 sites. This represents an effort to bring together all of the obsidian sourcing data produced for this broad archaeological setting, and evaluate it from a heterarchical approach that highlights the distributed nature of regional interaction. Heterarchical perspectives are applied here through the use of network analysis in order to highlight clusters of sites that are more connected to each other than they are to others in the system, and to determine the roles of each site in the system’s overall structure. As such, order is highlighted as a result of the organization of data-driven ties among sites, which are unrestricted by presumptions relating to geographical position or of pre-defined rank. The results are compared with more established models of regional interaction in the settings of interest, and heterarchical perspectives through network analysis are shown to complement common understandings of broad-scale connectivity at various points in time. / Thesis / Master of Arts (MA)

No description available.

Reinforcement learning (RL) has been at the core of recent advances in fulfilling the AI promise towards general intelligence. Unlike other machine learning (ML) paradigms, such as supervised learning (SL) that learn to mimic how humans act, RL tries to mimic how humans learn, and in many tasks, managed to discover new strategies and achieved super-human performance. This is possible mainly because RL algorithms are allowed to interact with the world to collect the data they need for training by themselves. This is not possible in SL, where the ML model is limited to a dataset collected by humans which can be biased towards sub-optimal solutions. The downside of RL is its high cost when trained on real systems. This high cost stems from the fact that the actions taken by an RL model during the initial phase of training are merely random. To overcome this issue, it is common to train RL models using simulators before deploying them in production. However, designing a realistic simulator that faithfully resembles the real environment is not easy at all. Furthermore, simulator-based approaches don’t utilize the sheer amount of field-data available at their disposal. This work investigates new ways to bridge the gap between SL and RL through an offline pre-training phase. The idea is to utilize the field-data to pre-train RL models in an offline setting (similar to SL), and then allow them to safely explore and improve their performance beyond human-level. The proposed training pipeline includes: (i) a process to convert static datasets into RL-environment, (ii) an MDP-aware data augmentation process of offline-dataset, and (iii) a pre-training step that improves RL exploration phase. We show how to apply this approach to design an action recommendation engine (ARE) that automates network operation centers (NOC); a task that is still tackled by teams of network professionals using hand-crafted rules. Our RL algorithm learns to maximize the Quality of Experience (QoE) of NOC users and minimize the operational costs (OPEX) compared to traditional algorithms. Furthermore, our algorithm is scalable, and can be used to control large-scale networks of arbitrary size.

In recent years, the Internet of Things (IoT) has been proliferating in various fields, such as health care, smart city, and connected autonomous vehicles. Accompanying the popularity of IoT are security attacks that exploit the vulnerabilities of many IoT devices. To build IoT anomaly detection systems, we need to collect and label a large amount data from diverse IoT scenarios, which is a time-consuming and prohibitive task if without the support of an IoT testbed. This thesis fills this urgent need by developing a scalable, flexible, safe, and secure IoT testbed. To make the testbed scalable, we need to reduce the hardware cost and make its architecture easily extendable. For this, we build a hybrid testbed consisting of real IoT devices, such as motion sensors and smart cameras, and emulated devices with Raspberry Pi. The emulated devices can replace real IoT devices with the same operational behaviour as real IoT devices but at a much lower price. Flexibility means the testbed can easily simulate different application scenarios. To make the testbed flexible, we build a dedicated data management module to facilitate the complex tasks in generating diverse traffic patterns, reproducing security attacks, collecting, visualizing, and analyzing network traffic. Testbed safety means the testbed will not cause any adverse impact to the Internet, and testbed security means protecting it from outside attacks. For safety, we carefully analyze the source code and the behaviour of launched attacks and configure a traffic filter to strictly contain the attack traffic within the testbed. For security, we scan and analyze the security of all IoT devices within the testbed to ensure no exposed vulnerability in the used devices. / Graduate

This work addresses Channel Assignment in a multi-radio multi-channel (MRMC) Wireless Mesh Network (WMN) using both Wi-Fi and Dynamic Spectrum Access (DSA) spectrum bands and standards. This scenario poses new challenges because nodes are spread out geographically so may have differing allowed channels and experience different levels of external interference in different channels. A solution must meet two conflicting requirements simultaneously: 1) avoid or minimise interference within the network and from external interference sources, and 2) maintain connectivity within the network. These two requirements must be met while staying within the link constraints and the radio interface constraints, such as only assigning as many channels to a node as it has radios. This work's original contribution to the field is a unified framework for channel optimisation and assignment in a WMN that uses both DSA and traditional Wi-Fi channels for interconnectivity. This contribution is realised by providing and analysing the performance of near-optimal Channel Assignment (CA) solutions using metaheuristic algorithms for the MRMC WMNs using DSA bands. We have created a simulation framework for evaluating the algorithms. The performance of Simulated Annealing, Genetic Algorithm, Differential Evolution, and Particle Swarm Optimisation algorithms have been analysed and compared for the CA optimisation problem. We introduce a novel algorithm, used alongside the metaheuristic optimisation algorithms, to generate feasible candidate CA solutions. Unlike previous studies, this sensing and CA work takes into account the requirement to use a Geolocation Spectrum Database (GLSD) to get the allowed channels, in addition to using spectrum sensing to identify and estimate the cumulative severity of both internal and external interference sources. External interference may be caused by other secondary users (SUs) in the vicinity or by primary transmitters of the DSA band whose emissions leak into adjacent channels, next-toadjacent, or even into further channels. We use signal-to-interference-plus-noise ratio (SINR) as the optimisation objective. This incorporates any possible source or type of interference and makes our method agnostic to the protocol or technology of the interfering devices while ensuring that the received signal level is high enough for connectivity to be maintained on as many links as possible. To support our assertion that SINR is a reasonable criterion on which to base the optimisation, we have carried out extensive outdoor measurements in both line-of-sight and wooded conditions in the television white space (TVWS) DSA band and the 5 GHz Wi-Fi band. These measurements show that SINR is useful as a performance measure, especially when the interference experienced on a link is high. Our statistical analysis shows that SINR effectively differentiates the performance of different channels and that SINR is well correlated with throughput and is thus a good predictor of end-user experience, despite varying conditions. We also identify and analyse the idle times created by Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) contention-based Medium Access Control (MAC) operations and propose the use of these idle times for spectrum sensing to measure the SINR on possible channels. This means we can perform spectrum sensing with zero spectrum sensing delay experienced by the end user. Unlike previous work, this spectrum sensing is transparent and can be performed without causing any disruption to the normal data transmission of the network. We conduct Markov chain analysis to find the expected length of time of a sensing window. We also derive an efficient minimum variance unbiased estimator of the interference plus noise and show how the SINR can be found using this estimate. Our estimation is more granular, accurate, and appropriate to the problem of Secondary User (SU)-SU coexistence than the binary hypothesis testing methods that are most common in the literature. Furthermore, we construct confidence intervals based on the probability density function derived for the observations. This leads to finding and showing the relationships between the number of sampling windows and sampling time, the interference power, and the achievable confidence interval width. While our results coincide with (and thus are confirmed by) some key previous recommendations, ours are more precise, granular, and accurate and allow for application to a wider range of operating conditions. Finally, we present alterations to the IEEE 802.11k protocol to enable the reporting of spectrum sensing results to the fusion or gateway node and algorithms for distributing the Channel Assignment once computed. We analyse the convergence rate of the proposed procedures and find that high network availability can be maintained despite the temporary loss of connectivity caused by the channel switching procedure. This dissertation consolidates the different activities required to improve the channel parameter settings of a multi-radio multi-channel DSA-WMN. The work facilitates the extension of Internet connectivity to the unconnected or unreliably connected in rural or peri-urban areas in a more cost-effective way, enabling more meaningful and affordable access technologies. It also empowers smaller players to construct better community networks for sharing local content. This technology can have knock-on effects of improved socio-economic conditions for the communities that use it.

This study deals with a) Understanding familiarity with transportation network companies (TNCs) and their use frequency b) Understanding travel choices in alcohol-related situations in Virginia. Ordered logistic regression models were used to identify factors associated with the respondents perceived familiarity with transportation network companies (TNCs) and use frequency. Based on the two models, the consistent factors were using a mobile wallet, a cell phone for entertainment, an app for taxi services, or an app for hotel booking/air transport arrangements, living in Northern Virginia, normally using multiple transportation modes for a single trip, higher education levels, and higher household income which were associated with increased TNC familiarity and use frequency. Self-identifying as White/Caucasian was also associated with increased TNC use frequency. Increased age was associated with decreasing TNC familiarity and use frequency. Subsequently, travel choices in alcohol related situations were studied with the objective of understanding the role of Transportation Network Companies (TNCs) in these situations and whether they have an impact on DUIs. For this objective, this study analyzes travel-choices associated with three scenarios alcohol related situations: (a) the last time the respondent consumed alcohol, (b) when avoiding driving after drinking, and (c) when avoiding riding with a driver who had been drinking. Multinomial Logistic Regression models were developed for all the three scenarios. For model (a), significant factors included use of a personal vehicle to arrive at the location where last consuming alcohol, being comfortable with having a credit card tied to a cell phone app, age, income, travelling alone when leaving the location where last consuming alcohol, having the highest educational attainment of high school graduate (GED), consumption of alcohol at bar/tavern/club, consumption of alcohol at home of friends/acquaintance place, and transportation network company (TNC – e.g., Uber, Lyft) weekly use frequency. For (b), use of a personal vehicle to arrive at the location where last consuming alcohol, consumption of alcohol at a bar/tavern/club, consumption of alcohol at the home of friends/acquaintance place, comfort with tying of credit card to apps, age, gender, income, multi-modal travel for a regular trip, TNC weekly use frequency, and use of an app for hotel reservations and/or air transportation arrangements are significant factors. For (c), use of a personal vehicle to arrive at the location where last consuming alcohol, walking to the location where last consuming alcohol, consumption of alcohol at a bar/tavern/club, comfort with tying a credit card to apps, age, income, TNC weekly use frequency, previously riding in a car with a driver who may have drunk too much to drive safely, and being employed full time are the significant factors. / Master of Science

Network virtualization promises to play a dominant role in shaping the future Internet by overcoming the Internet ossification problem. Since a single protocol stack cannot accommodate the requirements of diverse application scenarios and network paradigms, it is evident that multiple networks should co-exist on the same network infrastructure. Network virtualization supports this feature by hosting multiple, diverse protocol suites on a shared network infrastructure. Each hosted virtual network instance can dynamically instantiate custom set of protocols and functionalities on the allocated resources (e.g., link bandwidth, CPU, memory) from the network substrate. As this technology matures, it is important to consider the security issues and develop efficient defense mechanisms against potential vulnerabilities in the network architecture. The architectural separation of network entities (i.e., network infrastructures, hosted virtual networks, and end-users) introduce set of attacks that are to some extent different from what can be observed in the current Internet. Each entity is driven by different objectives and hence it cannot be assumed that they always cooperate to ensure all aspects of the network operate correctly and securely. Instead, the network entities may behave in a non-cooperative or malicious way to gain benefits. This work proposes set of defense mechanisms that addresses the following challenges: 1) How can the network virtualization architecture ensure anonymity and user privacy (i.e., confidential packet forwarding functionality) when virtual networks are hosted on third-party network infrastructures?, and 2) With the introduction of flexibility in customizing the virtual network and the need for intrinsic security guarantees, can there be a virtual network instance that effectively prevents unauthorized network access by curbing the attack traffic close to the source and ensure only authorized traffic is transmitted?. To address the above challenges, this dissertation proposes multiple defense mechanisms. In a typical virtualized network, the network infrastructure and the virtual network are managed by different administrative entities that may not trust each other, raising the concern that any honest-but-curious network infrastructure provider may snoop on traffic sent by the hosted virtual networks. In such a scenario, the virtual network might hesitate to disclose operational information (e.g., source and destination addresses of network traffic, routing information, etc.) to the infrastructure provider. However, the network infrastructure does need sufficient information to perform packet forwarding. We present Encrypted IP (EncrIP), a protocol for encrypting IP addresses that hides information about the virtual network while still allowing packet forwarding with longest-prefix matching techniques that are implemented in commodity routers. Using probabilistic encryption, EncrIP can avoid that an observer can identify what traffic belongs to the same source-destination pairs. Our evaluation results show that EncrIP requires only a few MB of memory on the gateways where traffic enters and leaves the network infrastructure. In our prototype implementation of EncrIP on GENI, which uses standard IP header, the success probability of a statistical inference attack to identify packets belonging to the same session is less than 0.001%. Therefore, we believe EncrIP presents a practical solution for protecting privacy in virtualized networks. While virtualizing the infrastructure components introduces flexibility by reprogramming the protocol stack, it doesn't directly solve the security issues that are encountered in the current Internet. On the contrary, the architecture increases the chances of additive vulnerabilities, thereby increasing the attack space to exploit and launch several attacks. Therefore it is important to consider a virtual network instance that ensures only authorized traffic is transmitted and attack traffic is squelched as close to their source as possible. Network virtualization provides an opportunity to host a network that can guarantee such high-levels of security features thereby protecting both the end systems and the network infrastructure components (i.e., routers, switches, etc.). In this work, we introduce a virtual network instance using capabilities-based network which present a fundamental shift in the security design of network architectures. Instead of permitting the transmission of packets from any source to any destination, routers deny forwarding by default. For a successful transmission, packets need to positively identify themselves and their permissions to each router in the forwarding path. The proposed capabilities-based system uses packet credentials based on Bloom filters. This high-performance design of capabilities makes it feasible that traffic is verified on every router in the network and most attack traffic can be contained within a single hop. Our experimental evaluation confirm that less than one percent of attack traffic passes the first hop and the performance overhead can be as low as 6% for large file transfers. Next, to identify packet forwarding misbehaviors in network virtualization, a controller-based misbehavior detection system is discussed as part of the future work. Overall, this dissertation introduces novel security mechanisms that can be instantiated as inherent security features in the network architecture for the future Internet. The technical challenges in this dissertation involves solving problems from computer networking, network security, principles of protocol design, probability and random processes, and algorithms.

Assessing a network's vulnerability to attack and random failure is a difficult and important problem that changes with network application and representation. We furnish algorithms that bound the robustness of a network under attack. We utilize both static graph-based and dynamic trace-driven representations to construct solutions appropriate for different scenarios. For static graphs we first introduce a spectral technique for developing a lower bound on the number of connected pairs of vertices in a graph after edge removal, which we apply to random graphs and the power grid of the Philippines. To address the problem of resource availability in networks we develop a second technique for bounding the number of nominally designated client vertices that can be disconnected from all server vertices after either edge or vertex removal (or both). This algorithm is also tested on the power grid and a wireless mesh network, the Internet AS level graph, and the highway systems of Iowa and Michigan. Dynamic networks are modeled as disruption tolerant networks (DTNs). DTNs are composed of mobile nodes that are intermittently connected via short-range wireless radios. In the context of both human and vehicular mobility networks we study both the effect of targeted node removal and the effect of augmentation with stationary relays.

Graphs are a great tool for representing complex physical and social systems, where the interactions among many units, from tens of animal species in a food-web, to millions of users in a social network, give rise to emergent, complex system behaviours. In the field of network science this representation, which is usually called a complex network, can be complicated at will to better represent the real system under study. For instance, interactions may be directed or may differ in their strength or cost, leading to directed weighted networks, but they may also depend on time, like in temporal networks, or nodes (i.e. the units of the system) may interact in different ways, in which case edge-coloured multi-graphs and multi-layer networks represent better the system. Besides this rich repertoire of network structures, we cannot forgot that edges represent interactions and that this interactions are not static, but are, instead, purposely established to reach some function of the system, as for instance, routing people and goods through a transportation network or cognition, through the exchange of neuro-physiological signals in the brain. Building on the foundations of spectral graph theory, of non-linear dimensionality reduction and diffusion maps, and of the recently introduced diffusion distance [Phys. Rev. Lett. 118, 168301 (2017)] we use the simple yet powerful tool of continuous-time Markov chains on networks to model their process-driven geometry and characterise their functional shape. The main results are: (i) the generalisation of the diffusion geometry framework to different types of interconnected systems (from edge-coloured multigraphs to multi-layer networks) and of random walk dynamics [Phys. Rev. E 103, 042301 (2021)] and (ii) the introduction of new descriptors based on the diffusion geometry to quantify and describe the micro- (through the network depth [J. Complex Netw. 8, 4 (2020)]), meso- (functional rich-club) and macro-scale (using statistics of the pairwise distances between the network's nodes [Comm. Phys. 4, 125 (2021)]) of complex networks.

This thesis concerns the optimization of the routing path in packet-switched computer-communication networks. Computer-communication networks over the past decade are outlined. A glossary of some of the terms used throughout this thesis are introduced. A brief description follows of the advantages of packet switching over the more conventional circuit-switched scheme for information transfer. The important design variables that a network planner is faced with in the design of these networks are discussed. A general design problem is stated and then decomposed into simpler subproblems one of which is the link-capacity assignment problem, which is briefly discussed. The route-assignment problem is identified as being of particular importance and is specified. A network model is introduced and relationships between performance measures, input parameters and constraints that appear in the general design problem are discussed. The routing problem is the formulated and a heuristic routing procedure is suggested as a sub-optimum solution to the problem. Basic routing methods are discussed. The principles of datagram and virtual circuit techniques are explained with reference to the routing of packets throughout the network. The directory routing technique with alternate routing is identified as being a specific requirement and the operation of this technique is explained in more detail. Two basic algorithms are introduced. The first which determines the shortest, second shortest, third shortest, etc., paths between all pairs of nodes in a network. The second which determines from all the paths in the first algorithm, the best alternative paths between all pairs of nodes in a network. A heuristic routing algorithm for establishing routing tables at each of the individual nodes in a packet switched data network is presented. Among the properties of a desirable routing algorithm is that the paths established between all node pairs are such that the average packet delay from source to destination node is minimal. The heuristic-routing algorithm proposed is to-be implemented on a newly proposed SAPONET packet-switching network, with special emphasis on the minimization of the average packet delay of the network. Results are presented and discussed for different combinations of the primary, secondary, tertiary and fourth alternative paths obtained. Finally, results are summarized and areas for further work identified.