Graduate Research Assistant

Alireza Bahramali

413-***-**** **********@**.*****.*** LinkedIn

Education

University of Massachusetts Amherst Amherst, MA

MS/PhD in Computer Science; GPA: 3.94 Sep. 2017 – June 2022 (Expected)

• Coursework: Machine Learning, Deep Learning, Advanced Algorithms, Advanced Computer Networks, Database Design & Implementation, Advanced Information Assurance. University of Tehran Tehran, Iran

BS in Electrical and Computer Engineering; GAP: 3.68 Sep. 2012 – June 2017

• Coursework: Advanced Programming, Computer Networks, Pattern Recognition. Experience

Graduate Research Assistant Sep. 2017 – Present

The Secure, Private Internet Research Group, UMass Amherst Amherst, MA

• Implemented a framework to perform traffic analysis on Tor connections using Deep Learning (DL), Python, and PyTorch. Improved the flow correlation accuracy on Tor connections by 92%.

• Developed a framework to perform traffic analysis on messaging applications using REST API and Python.

• Designed Universal Adversarial Examples to defeat DL-based traffic analysis attacks such as website fingerprinting and flow correlation using Python and PyTorch. Reduced the accuracy of such attacks by 90% by only adding 10% bandwidth overhead.

• Worked with different DL models such as Convolutional Neural Networks (CNN) and RESNET and integrated them into network traffic domain.

• Integrated Adversarial Example defense techniques such as Adversarial Training, Randomized Smoothing, Input-Gradient Regularization, and Region-based Classification into network traffic domain. Technical Skills

Languages: Python, PyTorch, C/C++, TensorFlow, MySQL, MATLAB. Developer Tools: Linux, Git, REST API, Docker, Selenium, Tor, pandas DataFrame, Jupyter Notebooks. Technology: Traffic Analysis, Data Structures, Deep Learning, Machine Learning, Adversarial Examples, Network Security, Network Programming, Algorithms.

Projects

Simple LinkedIn C++, Object Oriented Programming, QT, Git

• Developed a social network similar to LinkedIn using C++ and Qt as the user interface. Packet Scheduling C++, Object Oriented Programming, Graph Theory, Git

• Implemented and compared packet scheduling algorithms in network switches using C++ and graph theory. English Premier League Prediction Python, Pytorch, Git

• Designed a DL classifier to predict English Premier League soccer matches using PyTorch. Top-K Insights From Multi-Dimensional Data Python, pandas DataFrames, MySQL

• Automated the process of extracting useful insights from multi-dimensional data. Messaging Application Bots Python, REST API, Selenium, Docker

• Automated message sending and receiving in Telegram, Signal, and Wickr using Python and REST API. Publications

• Nasr M., Bahramali A., Houmansadr A. Defeating DNN-Based Traffic Analysis Systems in Real-Time With Blind Adversarial Perturbations. The USENIX Security Symposium. 2021.

• Bahramali A., Soltani R., Houmansadr A., Goeckel D., Towesly D. Practical Traffic Analysis Attacks on Secure Messaging Applications. The Network and Distributed System Security Symposium. 2020.

• Nasr M., Bahramali A., Houmansadr A. DeepCorr: Strong Flow Correlation Attacks on Tor Using Deep Learning. The ACM Conference on Computer and Communications Security. 2018. Honours

• Ranked 70th among 260000 participants in Iran’s National Universities Entrance Exam (Konkur), 2012.

Contact this candidate