Research Experiences in Cybersecurity Algorithms

The 2023 Summer Research Experience for Undergraduates (REU) program in the Computer Science Department at Montana State University

sponsored by the National Science Foundation

Topics Include:

• Tools for Reducing the Technical Debt of Systems and quantifying quality (e.g., cyber-security)
• Tools for Cyber-security
• Static analysis of source code (CWEs, CVEs, ATT&CK)
• Attack pattern detection through ML techniques
• Network threat detection
• Blockchain technology and cybersecurity
• Binary source code analysis for malware detection
• Unsupervised approaches for attack detection

Highlights:

• 10 weeks, May 29 - August 4
• $6000 stipend
• Travel expenses will be reimbursed up to $600
• Paid housing at MSU dorms
• Meal allowance $120/week
• State-of-the-art research facilities
• Faculty-student mentoring
• Weekend outings, social activities (Yellowstone Park, hiking, outdoor music concerts,...)

Project Director:

• Dr. Clemente Izurieta, Computer Science, MSU
  Email: clemente.izurieta@montana.edu Phone: 406.994.3720

Project Co-director:

• Dr. Brendan Mumey, Computer Science, MSU
  Email: brendan.mumey@montana.edu Phone: 406.994.7811

About MSU:

• Located in Bozeman, Montana in the heart of the Rocky Mountains
• Approximately 16,000 students
• www.montana.edu, www.gsoc.montana.edu

Nearby attractions:

• Yellowstone National Park, www.nps.gov/yell
• Hiking and biking trails 10 minutes from campus

Program Objectives:

• Expose students to real-world, innovative research and development focused on the above-listed topics
• Encourage more undergraduates to continue their academic careers and seek graduate degrees in computer science, engineering, and mathematics
• Help participants develop research skills and improve communication and collaborative skills
• Have fun!! We will organize several group activities such as hikes, and a trip to Yellowstone. Montana State University is located in the heart of the Rocky Mountains in Bozeman, Montana

Applications for the Summer of 2023 are OPEN

To Apply (deadline March 2, 2023) you will need to send the project director the following documents:

• A personal statement (You can email your statement as an attached pdf file)
• A transcript from your undergraduate institution (you may email to Dr. Clemente Izurieta) It is ok to send an unofficial transcript for the application.
• Two letters of recommendation. Please have your letters emailed directly to Dr. Izurieta with a subject line indicating REU MSU <your name>

Requirements:

Applicant must be a citizen or a permanent resident of the United States

Applicant must be an undergraduate student with good standing

All applicants are expected to have completed a course in Algorithms and Programming (no exceptions)

Sample Past Projects:

Quality Assurance through source code and binary disassembly

This project will explore recognizing threats in source code as well as binaries to help develop hierarchical models for quality assurance that properly reflect the observed threats.

In collaboration with Idaho National Laboratories, we can use state-of-the-art tools to recognize and visualize threats to help develop quality models.

Auditing End-to-End Security Within Wireless Networks In The Wild

The goal of this project is to contribute to ongoing work to analyze a new dataset of network traces with the aim of uncovering insecure network connections between mobile applications and popular smart home and smart-health devices. This work may reveal potential security and privacy leaks within such networks. Students will gain practical research experience by working to implement and apply various algorithms to analyze network traces and automate the process of identifying secure/insecure data channels. Students should have strong analytical skills, familiarity with mobile platforms, and an interest in learning about wireless networks and protocols.

Securing the Electric Power Grid

Securing the electric power grid remains a major focus of grid operators and government entities. The move to smart grid technologies
allows unprecedented control over the generation and consumption of electric power, but at the cost of enabling more complex attacks. We
aim to bring predictability to smart grid operations by scheduling jobs in such a way as to enable optimal grid functioning (peak demand
minimization). Of particular interest are jobs with explicit dependencies, a class of power jobs that appear often in industrial settings. We will develop novel approximation algorithms and test their viability in real-world scenarios.

Random Forest-Based Anomaly Detection For Improved Cybersecurity

As cyber attacks increase in both frequency and complexity, intrusion detection is becoming an increasingly critical task in the cybersecurity domain. Broadly, the goal of intrusion detection is to analyze network traffic and distinguish "good," authorized connections, from unauthorized, "bad" connections that could indicate attacks on a computer network. Intrusion detection often relies on anomaly detection methods. Anomaly detection methods, in general, seek to find anomalous instances in a data set. Applied to the intrusion detection task, anomaly deletion methods can be used to find anomalous instances of network traffic that could signal a network attack has occurred. In this project, students will explore the use of random forest-based approaches to anomaly detection. Random forests are generally used in supervised machine learning tasks, such as classification. However, extensions to random forests allow for their use in unsupervised anomaly detection as well. Students will explore the use of Isolation forests as well as the more recent variant Extended Isolation Forests in detecting anomalies in network traffic data, starting with the well-known KDD-Cup99 data set. Students will compare the anomaly detection accuracy and computational efficiency of random forest-based anomaly detection to other popular methods, such as Local Outlier Factor and One-Class-SVM.

Learning Traffic Control Policies for Assured Autonomy

With the increasing development of technologies for self-driving vehicles and the machine learning methods being developed and applied in these technologies, it is becoming increasingly important to examine potential security and ethical impacts on using these technologies, especially when such vehicles are mixed with traditional, manually controlled vehicles. To what extent can safe control be assured that is also resistant to external interference (either intentional or unintentional)? Currently, the Numerical Intelligent Systems Laboratory is exploring these questions by considering distributed learning and optimization methods based on "Factored Evolutionary Algorithms" to develop robust strategies for traffic control, for example when dealing with merging traffic or traffic interacting at intersections. This project will extend the methods being developed to assess and respond to potential safety-related issues that could arise in traffic systems involving a mix of autonomous and manually-controlled vehicles.