Browsing by Subject "Cyber security"

Now showing 1 - 2 of 2

Mexico’s national security framework in the context of an interdependent world : a comparative architecture approach
(2013-08) Martinez Espinosa, Cesar Alfredo; Ward, Peter M., 1951-
In a more complex and interdependent world, nations face new challenges that threaten their national security. National security should not be understood exclusively in the way of military threats by adversarial states but in a broader way: how old and new sectoral threats affect not only a state and its institutions but a nation as a whole, physically and economically. This dissertation looks into how the nature of security threats and risks has evolved in recent years. This dissertation then explores how different nations have decided to publish national security strategy documents and analyzes the way in which they include this broadened understanding of security: it finds that there is evidence of international policy diffusion related to the publication of such security strategies and that nations are evolving towards a broader understanding of security that includes models like whole-of-government, and whole-of-society. In the second half, this dissertation analyzes the route through which Mexico has reformed its national security framework since the year 2000 through a policy streams approach. After looking at the path that led to the creation of Mexico’s modern national security institutions, it analyzes the way in which Mexico national interests can be determined and how these interests inform the way in which Mexico understands national security threats and risks in the 21st Century.
Using machine learning to detect web application attacks
(2020-12-02) Franklin, Robin Maria; Caramanis, Constantine
With the increased ease in cloud deployment platforms, web applications have become an easy target for cyber-criminals and state-sponsored hackers. In this paper, I propose a detection solution to help identify network traffic generated by web application attacks. My experimental results reveal that the gradient boosting models, like LightGBM and XGBoost, all yielded extremely high ROC-AUC scores above .98. When compared to traditional anomaly detection models, like K-nearest neighbors, the ROC-AUC scores are higher and training times are much faster. Finally, I also identified several network data features like the window length in bytes, packet length, and flow duration that are critical when identifying web application attacks using network traffic data.