Software-Defined Networking (SDN) is an emerging network paradigm that decouples the control plane from the data plane, which allows network administrators to consolidate common network services into a centralized module named SDN controller. Applications’ policies are transformed into standardized network rules in the data plane via SDN controller. Even though this centralization brings a great flexibility and programmability to the network, network rules generated by SDN applications cannot be trusted because there may exist malicious SDN applications, and insecure network flows can be made due to complex relations across network rules. In this dissertation, I investigate how to identify and resolve these security violations in SDN caused by the combination of network rules and applications’ policies. To this end, I propose a systematic policy management framework that better protects SDN itself and hardens existing network defense mechanisms using SDN.

More specifically, I discuss the following four security challenges in this dissertation: (1) In SDN, generating reliable network rules is challenging because SDN applications cannot be trusted and have complicated dependencies each other. To address this problem, I analyze applications’ policies and remove those dependencies by applying grid-based policy decomposition mechanism; (2) One network rule could accidentally affect others (or by malicious users), which lead to creating of indirect security violations. I build systematic and automated tools that analyze network rules in the data plane to detect a wide range of security violations and resolve them in an automated fashion; (3) A fundamental limitation of current SDN protocol (OpenFlow) is a lack of statefulness, which is extremely important to several security applications such as stateful firewall. To bring statelessness to SDN-based environment, I come up with an innovative stateful monitoring scheme by extending existing OpenFlow specifications; (4) Existing honeynet architecture is suffering from its limited functionalities of ’data control’ and ’data capture’. To address this challenge, I design and implement an innovative next generation SDN-based honeynet architecture.

Access control has been historically recognized as an effective technique for ensuring that computer systems preserve important security properties. Recently, attribute-based

access control (ABAC) has emerged as a new paradigm to provide access mediation

by leveraging the concept of attributes: observable properties that become relevant under a certain security context and are exhibited by the entities normally involved in the mediation process, namely, end-users and protected resources. Also recently, independently-run organizations from the private and public sectors have recognized the benefits of engaging in multi-disciplinary research collaborations that involve sharing sensitive proprietary resources such as scientific data, networking capabilities and computation time and have recognized ABAC as the paradigm that suits their needs for restricting the way such resources are to be shared with each other. In such a setting, a robust yet flexible access mediation scheme is crucial to guarantee participants are granted access to such resources in a safe and secure manner.

However, no consensus exists either in the literature with respect to a formal model that clearly defines the way the components depicted in ABAC should interact with each other, so that the rigorous study of security properties to be effectively pursued. This dissertation proposes an approach tailored to provide a well-defined and formal definition of ABAC, including a description on how attributes exhibited by different independent organizations are to be leveraged for mediating access to shared resources, by allowing for collaborating parties to engage in federations for the specification, discovery, evaluation and communication of attributes, policies, and access mediation decisions. In addition, a software assurance framework is introduced to support the correct construction of enforcement mechanisms implementing our approach by leveraging validation and verification techniques based on software assertions, namely, design by contract (DBC) and behavioral interface specification languages (BISL). Finally, this dissertation also proposes a distributed trust framework that allows for exchanging recommendations on the perceived reputations of members of our proposed federations, in such a way that the level of trust of previously-unknown participants can be properly assessed for the purposes of access mediation.

Mobile Cloud computing has shown its capability to support mobile devices for

provisioning computing, storage and communication resources. A distributed mobile

cloud service system called "POEM" is presented to manage the mobile cloud resource

and compose mobile cloud applications. POEM considers resource management not

only between mobile devices and clouds, but also among mobile devices. It implements

both computation offloading and service composition features. The proposed POEM

solution is demonstrated by using OSGi and XMPP techniques.

Offloading is one major type of collaborations between mobile device and cloud

to achieve less execution time and less energy consumption. Offloading decisions for

mobile cloud collaboration involve many decision factors. One of important decision

factors is the network unavailability. This report presents an offloading decision model

that takes network unavailability into consideration. The application execution time

and energy consumption in both ideal network and network with some unavailability

are analyzed. Based on the presented theoretical model, an application partition

algorithm and a decision module are presented to produce an offloading decision that

is resistant to network unavailability.

Existing offloading models mainly focus on the one-to-one offloading relation. To

address the multi-factor and multi-site offloading mobile cloud application scenarios,

a multi-factor multi-site risk-based offloading model is presented, which abstracts the

offloading impact factors as for offloading benefit and offloading risk. The offloading

decision is made based on a comprehensive offloading risk evaluation. This presented

model is generic and expendable. Four offloading impact factors are presented to show

the construction and operation of the presented offloading model, which can be easily

extended to incorporate more factors to make offloading decision more comprehensive.

The overall offloading benefits and risks are aggregated based on the mobile cloud

users' preference.

The offloading topology may change during the whole application life. A set of

algorithms are presented to address the service topology reconfiguration problem in

several mobile cloud representative application scenarios, i.e., they are modeled as

finite horizon scenarios, infinite horizon scenarios, and large state space scenarios to

represent ad hoc, long-term, and large-scale mobile cloud service composition scenarios,

respectively.

Wireless communication technologies have been playing an important role in modern society. Due to its inherent mobility property, wireless networks are more vulnerable to passive attacks than traditional wired networks. Anonymity, as an important issue in mobile network environment, serves as the first topic that leads to all the research work presented in this manuscript. Specifically, anonymity issue in Mobile Ad hoc Networks (MANETs) is discussed with details as the first section of research.



To thoroughly study on this topic, the presented work approaches it from an attacker's perspective. Under a perfect scenario, all the traffic in a targeted MANET exhibits the communication relations to a passive attacker. However, localization errors pose a significant influence on the accuracy of the derived communication patterns. To handle such issue, a new scheme is proposed to generate super nodes, which represent the activities of user groups in the target MANET. This scheme also helps reduce the scale of monitoring work by grouping users based on their behaviors.



The first part of work on anonymity in MANET leads to the thought on its major cause. The link-based communication pattern is a key contributor to the success of the traffic analysis attack. A natural way to circumvent such issue is to use link-less approaches. Information Centric Networking (ICN) is a typical instance of such kind. Its communication pattern is able to overcome the anonymity issue with MANET. However, it also comes with its own shortcomings. One of them is access control enforcement. To tackle this issue, a new naming scheme for contents transmitted in ICN networks is presented. This scheme is based on a new Attribute-Based Encryption (ABE) algorithm. It enforces access control in ICN with minimum requirements on additional network components.



Following the research work on ABE, an important function, delegation, exhibits a potential security issue. In traditional ABE schemes, Ciphertext-Policy ABE (CP-ABE), a user is able to generate a subset of authentic attribute key components for other users using delegation function. This capability is not monitored or controlled by the trusted third party (TTP) in the cryptosystem. A direct threat caused from this issue is that any user may intentionally or unintentionally lower the standards for attribute assignments. Unauthorized users/attackers may be able to obtain their desired attributes through a delegation party instead of directly from the TTP. As the third part of work presented in this manuscript, a three-level delegation restriction architecture is proposed. Furthermore, a delegation restriction scheme following this architecture is also presented. This scheme allows the TTP to have full control on the delegation function of all its direct users.

Detecting cyber-attacks in cyber systems is essential for protecting cyber infrastructures from cyber-attacks. It is very difficult to detect cyber-attacks in cyber systems due to their high complexity. The accuracy of the attack detection in the cyber systems depends heavily on the completeness of the collected sensor information. In this thesis, two approaches are presented: one to detecting attacks in completely observable cyber systems, and the other to estimating types of states in partially observable cyber systems for attack detection in cyber systems. These two approaches are illustrated using three large data sets of network traffic because the packet-level information of the network traffic data provides details about the cyber systems.

The approach to attack detection in cyber systems is based on a multimodal artificial neural network (MANN) using the collected network traffic data from completely observable cyber systems for training and testing. Since the training of MANN is computationally intensive, to reduce the computational overhead, an efficient feature selection algorithm using the genetic algorithm is developed and incorporated in this approach.

In order to detect attacks in cyber systems in partially observable environments, an approach to estimating the types of states in partially observable cyber systems, which is the first phase of attack detection in cyber systems in partially observable environments, is presented. The types of states of such cyber systems are useful to detecting cyber-attacks in such cyber systems. This approach involves the use of a convolutional neural network (CNN), and unsupervised learning with elbow method and k-means clustering algorithm.

In traditional networks the control and data plane are highly coupled, hindering development. With Software Defined Networking (SDN), the two planes are separated, allowing innovations on either one independently of the other. Here, the control plane is formed by the applications that specify an organization's policy and the data plane contains the forwarding logic. The application sends all commands to an SDN controller which then performs the requested action on behalf of the application. Generally, the requested action is a modification to the flow tables, present in the switches, to reflect a change in the organization's policy. There are a number of ways to control the network using the SDN principles, but the most widely used approach is OpenFlow.

With the applications now having direct access to the flow table entries, it is easy to have inconsistencies arise in the flow table rules. Since the flow rules are structured similar to firewall rules, the research done in analyzing and identifying firewall rule conflicts can be adapted to work with OpenFlow rules.

The main work of this thesis is to implement flow conflict detection logic in OpenDaylight and inspect the applicability of techniques in visualizing the conflicts. A hierarchical edge-bundling technique coupled with a Reingold-Tilford tree is employed to present the relationship between the conflicting rules. Additionally, a table-driven approach is also implemented to display the details of each flow.

Both types of visualization are then tested for correctness by providing them with flows which are known to have conflicts. The conflicts were identified properly and displayed by the views.

With the software-defined networking trend growing, several network virtualization controllers have been developed in recent years. These controllers, also called network hypervisors, attempt to manage physical SDN based networks so that multiple tenants can safely share the same forwarding plane hardware without risk of being affected by or affecting other tenants. However, many areas remain unexplored by current network hypervisor implementations. This thesis presents and evaluates some of the features offered by network hypervisors, such as full header space availability, isolation, and transparent traffic forwarding capabilities for tenants. Flow setup time and throughput are also measured and compared among different network hypervisors. Three different network hypervisors are evaluated: FlowVisor, VeRTIGO and OpenVirteX. These virtualization tools are assessed with experiments conducted on three different testbeds: an emulated Mininet scenario, a physical single-switch testbed, and also a remote GENI testbed. The results indicate that network hypervisors bring SDN flexibility to network virtualization, making it easier for network administrators to define with precision how the network is sliced and divided among tenants. This increased flexibility, however, may come with the cost of decreased performance, and also brings additional risks of interoperability due to a lack of standardization of virtualization methods.

A load balancer is an essential part of many network systems. A load balancer is capable of dividing and redistributing incoming network traffic to different back end servers, thus improving reliability and performance. Existing load balancing solutions can be classified into two categories: hardware-based or software-based. Hardware-based load balancing systems are hard to manage and force network administrators to scale up (replacing with more powerful but expensive hardware) when their system can not handle the growing traffic. Software-based solutions have a limitation when dealing with a single large TCP flow. In recent years, with the fast developments of virtualization technology, a new trend of network function virtualization (NFV) is being adopted. Instead of using proprietary hardware, an NFV network infrastructure uses virtual machines running to implement network functions such as load balancers, firewalls, etc. In this thesis, a new load balancing system is designed and evaluated. This system is high performance and flexible. It can fully utilize the bandwidth between a load balancer and back end servers compared to traditional load balancers such as HAProxy. The experimental results show that using this NFV load balancer could have $n$ ($n$ is the number of back end servers) times better performance than HAProxy. Also, an extract, transform and load (ETL) application was implemented to demonstrate that this load balancer can shorten data load time. The experiment shows that when loading a large data set (18.3GB), our load balancer needs only 28\% less time than traditional load balancer.

Smartphones are pervasive nowadays. They are supported by mobile platforms that allow users to download and run feature-rich mobile applications (apps). While mobile apps help users conveniently process personal data on mobile devices, they also pose security and privacy threats and put user's data at risk. Even though modern mobile platforms such as Android have integrated security mechanisms to protect users, most mechanisms do not easily adapt to user's security requirements and rapidly evolving threats. They either fail to provide sufficient intelligence for a user to make informed security decisions, or require great sophistication to configure the mechanisms for enforcing security decisions. These limitations lead to a situation where users are disadvantageous against emerging malware on modern mobile platforms. To remedy this situation, I propose automated and systematic approaches to address three security management tasks: monitoring, assessment, and confinement of mobile apps. In particular, monitoring apps helps a user observe and record apps' runtime behaviors as controlled under security mechanisms. Automated assessment distills intelligence from the observed behaviors and the security configurations of security mechanisms. The distilled intelligence further fuels enhanced confinement mechanisms that flexibly and accurately shape apps' behaviors. To demonstrate the feasibility of my approaches, I design and implement a suite of proof-of-concept prototypes that support the three tasks respectively.

Lighting systems and air-conditioning systems are two of the largest energy consuming end-uses in buildings. Lighting control in smart buildings and homes can be automated by having computer controlled lights and window blinds along with illumination sensors that are distributed in the building, while temperature control can be automated by having computer controlled air-conditioning systems. However, programming actuators in a large-scale environment for buildings and homes can be time consuming and expensive. This dissertation presents an approach that algorithmically sets up the control system that can automate any building without requiring custom programming. This is achieved by imbibing the system self calibrating and self learning abilities.

For lighting control, the dissertation describes how the problem is non-deterministic polynomial-time hard(NP-Hard) but can be resolved by heuristics. The resulting system controls blinds to ensure uniform lighting and also adds artificial illumination to ensure light coverage remains adequate at all times of the day, while adjusting for weather and seasons. In the absence of daylight, the system resorts to artificial lighting.

For temperature control, the dissertation describes how the temperature control problem is modeled using convex quadratic programming. The impact of every air conditioner on each sensor at a particular time is learnt using a linear regression model. The resulting system controls air-conditioning equipments to ensure the maintenance of user comfort and low cost of energy consumptions. The system can be deployed in large scale environments. It can accept multiple target setpoints at a time, which improves the flexibility and efficiency of cooling systems requiring temperature control.

The methods proposed work as generic control algorithms and are not preprogrammed for a particular place or building. The feasibility, adaptivity and scalability features of the system have been validated through various actual and simulated experiments.