High School Research

Pipes and drones were the subject of my relatively simple, independent research projects during my first two years of high school. However, the next two years, I preoccupied myself with problems in machine learning security and privacy, resulting in two major projects at the NYU Center for Cybersecurity where I was mentored by Prof. Ramesh Karri, Dr. Zahra Ghodsi, and Prof. Kanad Basu. I was also the Head TA (“Chief Trainer”) for QHSS’s math research program. While I fully intend to write more about my experiences, this post aims to organize information about my projects.

A New Method for the Exploitation of Speech Recognition Systems

Abstract

The rapid proliferation and adoption of speech recognition systems in our day-to-day lives result in greater consequences for possible vulnerabilities. Previous research has proven that host hardware and preprocessing can be leveraged to successfully deceive speech recognition systems. Additionally, neural networks, algorithms within modern systems, can be effectively fooled by generating adversarial noise. However, a method to exploit speech recognition systems by leveraging neural networks was notably absent. An algorithm was developed that crafts universal, transformable adversarial noise for the inputs of a speech recognition system that would result in deliberate misclassification. To evaluate this algorithm, adversarial noises for five randomly chosen target classes were produced using a substitute neural network. The noises were then added to the inputs of a victim system in a black-box setting. On average, the crafted adversarial noises led to deliberate misclassification 60.42% of the time. The universality of the generated noises increased the inconspicuousness, aided by limitations set on the noises. The feasibility and practicality of the attack was increased by the fact that the adversarial noises were transformable. Thus, the neural networks in speech recognition systems are a significant vulnerability. It is imperative that attacks such as these are mitigated for speech recognition systems to be considered safe. Future research can improve upon the proposed attack for the purpose of finding more vulnerabilities or focus upon building an optimal defense strategy.

Reference

S. Hussain, Z. Ghodsi, R. Karri, “A New Method for the Exploitation of Speech Recogniton Systems,” Computational Cybersecurity for Compromised Environments Workshop, 2018.

Honors and Awards

ISEF 2nd Award in Systems Software. Shanghai STEM Cloud Award (ISEF). NSA-RD 2nd “Science Security” Award (ISEF). GoDaddy Data Award (ISEF). ACM 4th Award (ISEF). NYCSEF 1st Award in Math and CS. Sarah and Morris Wiesenthal Award (NYCSEF). Naval Science Award (NYCSEF). NYU Tandon Press Release.

Publications and Presentations

Materials

Detecting Privacy Violations in Children’s Apps Using HPCs & COPPTCHA: COPPA Tracking by Checking Hardware-Level Activity

Abstract

The recent surge in children’s apps reflects the proliferation of new avenues for education and entertainment globally. However, more than half of all children’s apps on the Android platform violate the Children’s Online Privacy Protection Act (COPPA), indicating an invasion of user data privacy that threatens the safety of millions. Previous research on detecting COPPA violations rely upon an analysis of the system binary, which is neither scalable nor reliable. To overcome these challenges, hardware performance counters (HPCs) were utilized to detect COPPA violations. A novel dataset was established after the profiling of a number of COPPA-compliant and COPPA-violating Android apps. Based upon this dataset, two methods, a general COPPA violation detector and a series of specialized COPPA violation detectors, were formulated. The former detects the existence of any possible COPPA violation. Supervised learning algorithms were applied to the whole dataset and, to address HPC measurement constraints, to feature-reduced data. The latter detects the existence of a specific COPPA violation. Thus, several classifiers trained upon feature-reduced data were developed. In addition to yielding high accuracies and low misclassification rates, these classifiers are secure and efficient due to the nature of HPCs.

Reference

K. Basu, S. Hussain, U. Gupta, and R. Karri, “COPPTCHA: COPPA Tracking by Checking Hardware-Level Activity,” IEEE Transactions on Information Forensics and Security, vol. 15, pp. 3213–3226, 2020.

Honors and Awards

ACM 1st Award (ISEF). NSA-RD First “Science Security” Award (ISEF). NYCSEF 1st Award in Math and CS. NYC JSHS 3rd in CS (JSHS). The Cardinals Press Release

Publications and Presentations

Materials