A New Method for the Exploitation of Speech Recognition Systems
Paper Slides Code Sample
I determined how to hack a speech recognition system, specifically a neural network for phonetic classification, using adversarial machine learning. I developed an algorithm and tested it using 2 neural networks and the TIMIT data set. My research was conducted at NYU CCS under Professor Ramesh Karri and Zahra Ghodsi.
Areas: Cybersecurity, Deep Learning (Adversarial Machine Learning)
Skills: TensorFlow, SSH, NumPy, Jupyter Notebook, Python 2.7, Technical Writing, Engineering Research
Awards: Intel ISEF Second Award in Systems Software, Shanghai STEM Cloud Award, National Security Agency Research Directorate Second “Science Security” Award, GoDaddy Data Award, Association for Computing Machinery Fourth Award, Intel ISEF Finalist, Naval Science Award, NYCSEF First Award in Computer Science, Sarah & Morris Wiesenthal Award
I designed and developed a prototype conveyor belt system for a new product at Vengo Labs. The system consisted of an Arduino, a stepper motor driver chip, and a stepper motor in addition to other mechanical parts.
Areas: Mechanical Engineering, Hardware(Electrical) Engineering, Rapid Prototyping
Skills: EaglePCB, Arduino, CATIA, Hand Tools
Areas: Cybersecurity, Hardware Development
The Effect of Neural Network Architecture on Quadcopter Control Systems Performance
I tested the performance of several PID controllers that included neural networks (and one standard one), and then conducted the Chi-Square GOF test to ascertain dependence.
Skills: Matlab, Simulink, Statistical Analysis, Technical Writing, Engineering Research