Tag: Mathematics

Attacker Behavior Modeling and Learning in Security Using Deep Neural Networks

Presenter(s): Alyssa Huque—Mathematics & Computer Science, Political Science

Faculty Mentor(s): Thanh Nguyen

Session: Prerecorded Poster Presentation

A common dilemma many security agencies face is how to effectively allocate limited resources to protect assets . This predicament is known as the Stackelberg Security Games . In order to optimize defense tactics, security agencies need to be able to anticipate adversarial behavior . Currently, there are computer modeling techniques that simulate the Stackelberg Security Games . However, these implementations are not completely optimized for human adversaries . The quantal response model (QR) operates under the assumption that humans act with perfect rationality, a flawed assumption that was improved upon in the SUQR model . The SUQR integrated a subjective utility function (SU) that could take learned parameters, but only from limited data (Nguyen, et . al 2013) . Deep neural networks have the potential to improve further than the SUQR by providing a better prediction of the attacker’s behavior . Deep neural networks can allow for a more robust set of input features that would be able to account for more nuances of human behavior . A model that could accurately predict adversarial actions has the potential to improve resource allocation and enhance the security of valuable assets .

Longitudinal Analysis of Major Video Streaming Services in the US

Presenter(s): Donna Hooshmand—Computer Science, Mathematics

Faculty Mentor(s): Reza Rejaie

Session: Prerecorded Poster Presentation

This study relies on several years of NETFLOW data for exchanged traffic between the University of Oregon network (UOnet) and the Internet to perform a longitudinal analysis on the characteristics of popular Internet Applications . We develop techniques to identify connections related to video streams from their NETFLOW records . We then investigate how the fraction of UOnet traffic associated with (i .e . popularity of) major video streaming applications (e .g . YoutTube, Netflix, Amazon Prime), the basic characteristics of their video (e .g . bandwidth and duration) and their delivery mechanism have evolved over the past few years . Our empirical findings will offer valuable insights into important practical aspects of video streams services and their evolution over time .

Time-SPIDER: Characterizing the Electric Field of Pulsed LASERs

Presenter(s): Jeremy Guenza-Marcus—Physics and Math

Faculty Mentor(s): Brian Smith

Session 1: It’s a Science Thing

Quantifying precise measurements is critical in any field . Our research focuses on advancing quantum optical methods in the study of metrology . SPIDER is an interferometric approach to characterizing (mathematically describing) ultrashort laser pulses in the frequency domain . Our research aims to develop a sister method to the accepted SPIDER approach, dubbed Time-SPIDER . Its purpose is to use the same approach as SPIDER, but rather in the temporal domain . The procedure is to first develop the theoretical framework, and then set up the experiment . At the moment, our work approaches the issue from a purely theoretical perspective . We find that the Time-SPIDER method is useful as a direct measurement technique for non-ultrashort pulses . Many industry-standard interferometers require an iterative approach to pulse characterization, which may not be well- calibrated if the pulse is not ultrashort . Time-SPIDER solves both of these issues . If we are able to move past the theory and create a working Time-SPIDER, it would be possible to continue with other projects in the lab that may require such set-up . In the grand scheme, Time-SPIDER is a step towards continuing the study of metrology, along with quantum optics itself .