Bachelor and Master Theses

Smart environments today are made up of many different connected devices (e.g., IoT sensors, smart appliances, robotics kits, and home automation tools), which together create complex systems. The interconnectedness and diversity of these devices, such as Raspberry Pi models, Arduino kits, Zigbee and Z-Wave controllers, environmental sensors, and smart home gadgets, introduce numerous security challenges and weaknesses. Finding these vulnerabilities in such dynamic and heterogeneous systems demands an analytical framework beyond traditional security methods.

This thesis proposes using graph theory to map out these smart environments. By treating devices, their connections, data flows, and points where users interact as parts of a graph (with nodes and edges), it becomes easier to analyze how everything is linked and identify potential security risks. This graph-based approach captures the physical setup, logical relationships, and security aspects, giving a fuller picture of how the smart system is built and where it might be vulnerable.

Research Questions:

1-How can graph theory and its associated algorithms be applied to model, analyze, and identify critical vulnerabilities in complex smart environments?
2-Which graph-theoretic metrics and techniques are most effective in detecting and prioritizing security threats within smart device networks?

Methodology:

The smart environment will be modeled as a graph, where devices, sensors, actuators, user points, and subsystems are nodes, and their interactions—such as communication and control permissions—are edges. Nodes and edges will include security attributes like access rights and encryption status.

Graph theory techniques such as centrality measures (Degree, Betweenness, Closeness), community detection algorithms (Louvain, Girvan-Newman), shortest path analysis, and PageRank will be applied to analyze the network. Centrality measures will identify key or vulnerable components; community detection will reveal clusters of risk; shortest path analysis will map possible attack routes; and PageRank will highlight critical targets.

Advanced graph embedding methods (e.g., node2vec, DeepWalk) will transform the graph into a vector space for machine learning-based threat prediction.

Implementation will use Python libraries such as NetworkX, igraph, and PyTorch Geometric. The proposed approach will be demonstrated using theoretical case studies or representative device configurations, and its effectiveness will be discussed based on established criteria from existing literature.

Expected Outcomes:

• A structured framework for representing smart environments using graph theory, incorporating both physical and logical device relationships.

• Application of key graph algorithms (e.g., centrality measures, community detection, shortest path, PageRank) to identify critical nodes, attack paths, and security risks.

• Exploration of advanced techniques like graph embeddings (e.g., node2vec, DeepWalk) to support threat prediction through machine learning.

• Demonstration of the approach using theoretical case studies or representative device configurations, providing a practical context for analysis.

• Evaluation and discussion of the method’s effectiveness based on existing security models and literature benchmarks.

• Proficiency in a programming language such as Python, with the ability to implement and work with graph algorithms.

• Familiarity with graph analysis libraries (e.g., NetworkX, igraph, PyTorch Geometric) used for modeling and evaluating network structures.

• A basic understanding of smart environments or IoT systems, including components, communication flows, and interaction patterns.

• Foundational knowledge of cybersecurity concepts, particularly in threat modeling, network vulnerabilities, and system-level security analysis.

•  

• Strong motivation and commitment to completing the thesis within the assigned semester.