Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Consumer Science

First Advisor

Elisa Bertino

Committee Chair

Elisa Bertino

Committee Member 1

Sonia Fahmy

Committee Member 2

Cristina Nita-Rotaru

Committee Member 3

Mathias Payer

Committee Member 4

Dongyan Xu


Sensor systems are becoming pervasive in many domains, and are recently being generalized by the Internet of Things (IoT). This wide deployment, however, presents significant security issues.

We develop security techniques for sensor systems and IoT, addressing all security management phases. Prior to deployment, the nodes need to be hardened. We develop nesCheck, a novel approach that combines static analysis and dynamic checking to efficiently enforce memory safety on TinyOS applications. As security guarantees come at a cost, determining which resources to protect becomes important. Our solution, OptAll, leverages game-theoretic techniques to determine the optimal allocation of security resources in IoT networks, taking into account fixed and variable costs, criticality of different portions of the network, and risk metrics related to a specified security goal.

Monitoring IoT devices and sensors during operation is necessary to detect incidents. We design Kalis, a knowledge-driven intrusion detection technique for IoT that does not target a single protocol or application, and adapts the detection strategy to the network features. As the scale of IoT makes the devices good targets for botnets, we design Heimdall, a whitelist-based anomaly detection technique for detecting and protecting against IoT-based denial of service attacks.

Once our monitoring tools detect an attack, determining its actual cause is crucial to an effective reaction. We design a fine-grained analysis tool for sensor networks that leverages resident packet parameters to determine whether a packet loss attack is node- or link-related and, in the second case, locate the attack source. Moreover, we design a statistical model for determining optimal system thresholds by exploiting packet parameters variances.

With our techniques' diagnosis information, we develop Kinesis, a security incident response system for sensor networks designed to recover from attacks without significant interruption, dynamically selecting response actions while being lightweight in communication and energy overhead.