simha Simha Sethumadhavan

Professor of Computer Science
Alfred P. Sloan Research Fellow
simha@columbia.edu

Research Areas

Computer Architecture, Computer Security

Research Overview

My research is in computer architecture and computer security, and how computer architecture can be used to improve computer security.

I have been professing the hardware-up method for securing computer systems for awhile now. While security is a full-system property, and both software and hardware have to be secure for a system to be secure, I believe that the most practical and principled way to achieve full-system security is to secure the hardware first against attacks, and then systematically harden each layer in the systems stack as we move up the stack through the microarchitecture, ISA interface, system software and application software. This method allows us to avoid security blind spots in the lower levels when designing security at higher levels in the systems stack.

Following this approach, we have invented methods for detecting and mitigating backdoors/trojans embedded in the hardware during its design or manufacturing, measuring and mitigating security vulnerabilities due to microarchitectural side channels, and developed architecture support for mitigating common software problems such as memory safety and malware attacks. We have built several end-to-end FPGA based demonstrations including our Polyglot system for preventing code injection and code reuse attacks, and the Silicon AV malware detection system that uses on-chip performance counters. Most recently we have been studying mechanism design for improving security.

A common denominator among computer architecture and computer security is the need to understand program behavior. What began as a quest to automatically discover special-purpose functional units (accelerators) has resulted in large number of works from methods to precisely measure execution times, to techniques for identifying similar code sequences in large code bases, to techniques for detecting malware at runtime, and the discovery of power law behaviors in programs!

We have also worked on improving the energy-efficiency and performance of post-Moore's-Law computers. We devised an analog accelerator that can provide performance benefits beyond digital accelerators today by side-stepping the 'digital tax' in terms of clock and value discretization, showed how they can be integrated into systems. We built some ASIC chips, and ran robotics and scientific applications, and showed that certain types of workloads run faster and consume less energy compared to state-of-the-art GPUs.

Finally, a long time ago, when I was a graduate student at UT-Austin, I contributed to the design, implementation, validation and bring-up of the distributed TRIPS processor, and also devised a follow-on composable microarchitecture.

Recognition

Ongoing Projects

Mechanism Design for Security (1,2,3 4), Improving Memory Safety, Co-design of Reliability and Security features, and Cloud Security.

Publications

Hardware Security

Microarchitectural Side Channels

Architectural Support for Memory Safety

Exploit Detection and Mitigation using Machine Learning

Architecture Support for Isolation

Software Only Approaches for Security

Next Gen Accelerators and Architectures

Program Characterization

Traditional Architecture and Microarchitecture

Miscellaneous (Concept papers, Wild and Crazy)

Recruiting

If you are an exceptional student and interested in doing cutting-edge research in Computer Security, Computer Architecture and Digital VLSI Design and Security, I invite you to apply to join our group. We welcome people of any gender identity or expression, race, ethnicity, nationality, sexual orientation, religion, culture, subculture, and political opinion. (Partially reproduced from Dreamwidth) .

PhD Students

Research Alumni

    PhD:
  1. Dr. John Demme
    Thesis: Overcoming the Intuition Wall: Measurement and Analysis in Computer Architecture
  2. Dr. Adam Waksman
    Thesis: Producing Trustworthy Hardware Using Untrusted Components, Personnel and Resources
  3. Dr. Mike Su
    Thesis: Uncovering Features in Behaviorally Similar Programs
  4. Dr. Adrian Tang
    Thesis: Security Engineering of Hardware-Software Interfaces
  5. Dr. Yipeng Huang
    Thesis: Hybrid Analog-Digital Co-processing for Scientific Computation
  6. Dr. Kanad Sinha
    Thesis: Repurposing Software Defenses with Specialized Hardware
  7. Dr. Miguel Arroyo
    Thesis: Bespoke Security for Resource Constrained Cyber-Physical Systems
  8. Dr. Mohamed Tarek Ibn Ziad
    Thesis: Hardware-Software Co-design for Practical Memory Safety

Teaching

Select Recent Professional Activities

Contact Information

Office: 465 CSB Directions
Phone:1 212 939 7062
Fax:1 212 666 0140
Snail mail:
Prof. Sethumadhavan
Department of Computer Science
Columbia University
1214 Amsterdam Ave.
450 CSB
Mail Code 0401
New York NY 10027.