Boyou Zhou

Security Researcher

Education

PhD in Computer Engineering

Boston University
2013 - 2019

BSc in Communication Engineering

Southeast University
2009 - 2013

Languages

  • Chinese (Native)
  • English (Professional)

Interests

  • Hiking
  • Swimming
  • Cooking

Career Profile

Summary I am a Ph.D Student in Integrated Circuits and System Group of Electrical and Computer Engineering in Boston University. My research focus is Computer Security. During my graduate school, I have worked on 3 projects, and currently on the 4th project. I have 3 internships with 2 companies. I am looking for a full-time job in the security area and my target graduation time is Jan 2019.

Experiences

Research Assistant

2013 - Present
Boston University

Current research work Control Flow Integrity (CFI) prevents the computer system from diverting control ow. Current commercially avail- able CFI enforcement are software-based techniques. All the hardware-assisted CFIs under research have low target precisions. I am working on developing the fine-grained CFI assisted by hardware, evaluated on RISCV evaluation platform.

Internship

2016
Qualcomm

Internship work I implemented a Digital Signal Processing module in the touch-screen chip design. I modeled my signal processing in Python with the builder design pattern, to validate my digital design. The builder pattern in the digital design models helped the team to model other blocks, or re-model my design in the future. To facilitate the usability of my models and validations, I also provided the websites for documenting the tools during the developments. During the integration, I designed a tool for Verilog automatic connection using Python and other scripting languages. The tool can detect the connection between dierent blocks, and generate a higher level structure that connects each individual blocks. With this tool, the high-level block generation is automated without the risk of introducing manual errors.

Internship

2014,2015
Analog Devices Inc.

Internship work Real Number Modeling (RNM) is the technique to model analog blocks on the modular level, instead of continuous signal simulation. The benefits of RNM is to provide fast simulation, and direct integration with digital blocks. I mainly explored the Real Number Modeling in modeling ICs and also peripheral structures. I used Verilog, Verilog-ams, and SystemVerilog to model the current and voltage behaviors.

Projects

My past projects and related publications During my 5 1⁄4 year graduate school years, I have worked on a variety of projects related to computer system security design. Below is the listed work and publications.

Hardware Performance Counter Malware Detection - BEST PAPER in AsiaCCS 2018 By literature survey, I found prevalent unrealistic assumptions and misconducts in Hardware Performance Counters (HPCs) malware detection using machine learning. To re-evaluate the feasibility of malware detection using Hardware Performance Counters (HPC), I measured the HPC traces on various programs, including 1,000 malware and 1,000 benignware. To perform the measurements, I used Rabbitmq to orchestra the communications in a cluster of machines. During the experiments, I modified the GRUB boot loader to reset the disk image after every malware experiment. I performed data analysis on these HPC values using various machine learning algorithms, such as K Nearest Neighbors (KNN), Decision Tree (DT), Multilayer Perceptrons (MLP), Naive Bayes, AdaBoost and Random Forest (RF). By applying these machine learning algorithms, I evaluated the robustness of malware detection using cross-validations.
Hardware Trojan Detection Through Near-IR Imaging - Hardware Trojans (HT) are maliciously-inserted hardware blocks in Integrated Circuits (IC). Traditional testing/validation techniques, for example, power and timing validation methods, are not able to detect HTs. Our group collaborated with optical imaging group to develop optical-backside imaging. I worked as the support for the IC design, data-processing in this project. I engineered optical structures in standard cells, and designed hardware watermarks in the ICs to detect any modifications made by the intruders. After designing the optical structures, I used various data processing, such as correlations, noise analysis, and machine learning, to elevate the HT detection accuracies. With these engineered structures, we can perform classications between dierent kinds of standard cells, such that any modifications or replacement to these cells can be easily detected through optical imaging.
Hardware Assisted Encryption Acceleration - To decrease the power consumption and accelerate computations on the Internet of Things (IoT) devices, I worked on the FPGA-assisted cryptographic accelerations. In this project, I used Zedboard as the reconfigurable substrate evaluation platform. The implemented cryptographic algorithms included symmetric cryptography, asymmetric cryptography, and secure hash functions. I also integrated our cryptographic engines with the OpenSSL library to inherit the library’s support for block cipher modes. At the same time, the FPGA implementation of cryptographic operations was more exible compared to custom hardware implementations of cryptographic components.
On-going Project: Contro Flow Integrity - Control Flow Integrity (CFI) prevents the computer system from diverting control ow. Current commercially avail- able CFI enforcement are software-based techniques. All the hardware-assisted CFIs under research have low target precisions. I am working on developing the fine-grained CFI assisted by hardware, evaluated on RISCV evaluation platform.
Class Project: Detection of Ever-cookies - In the cyber-security class, we developed a technique to detect ever-cookies, the cookies that cannot be deleted by default cookie-deletion in the browser. To achieve the ever-cookie detection, we checked the current website under our blacklist and whitelist. We then forwarded the website address that was not on the lists to our evaluation server. The evaluation server browsed the website, deleted the cookies and visited the website for the second time. If our information resided in the website information, we recorded the website domain in our blacklist. I was in charge of developing the extension to forward website, maintained the blacklist and whitelist.

Skills & Proficiency

Python

C/C++

Verilog/Verilog-ams/System-Verilog

HTML & JAVASCRIPT & CSS

Bash & Perl & Any Other Scripting