Petros Maniatis
Intel Research Berkeley
2150 Shattuck Ave., Ste. 1300
Berkeley, CA 94704-1347 USA
T +1-510-495-3087
F +1-510-495-3049
firstname dot lastname at company dot com

Biography

Welcome! This is a corporation and my bio can only be in the third person for maximum effect so now that I've welcomed you, here it is:

Petros Maniatis is a Research Scientist at Intel Research, Berkeley. He joined Intel Research in October of 2003, immediately after getting his MSc and Ph.D. from the Computer Science Department at Stanford University. Prior to Stanford, he obtained his BSc with honors at the Department of Informatics of the University of Athens in Greece. His research interests lie primarily in the confluence of distributed systems, security, and fault tolerance.

Research

My research interests lie in distributed systems, dependability, and security. I am particularly interested in making it easier to build large-scale distributed systems in a sustainable way that reduces unnecessary work for the designers and maintainers and improves their ability to obtain efficient, dependable results.

My current main project is Customized Secure Networked Systems. Its focus is to build MOMMIE (which we have barely managed to make stand for Middleware for Optimized Messaging in Insecure Environments), a messaging middleware that makes it easier for developers to build secure distributed applications while getting high performance from the resources they have available (computers, networks, software). The main idea is to abstract the security properties of an application above implementational details (for example, building blocks such as the particular crypto algorithm or auditing protocol) by using annotations for data in a typical language such as C++. A compiler then implements the abstractions with available building blocks.

This approach has some neat features:

• If you move your application to a different computer with different available building blocks, you don't have to reimplement it; the compiler (in the future, an on-line rewriter) will adapt to what is available. For example, if you move your application to a computer with an asymmetric cryptography accelerator, the compiler will use the accelerator without bothering you, the developer.
• Most security properties can be implemented in many simple or complex ways. For instance, authentication (that is, "Mary said 'I approve'") can be implemented with digital signatures based on RSA, with HMAC based on a shared key and a hash function, or even without any cryptography if you hear the statement directly from Mary. MOMMIE can pick any of the possible ways to implement a security property in your application to effect your strategy: optimize for minimum cost, greatest uniformity, highest diversity, etc.
• Many distributed applications rely on the security properties of individual protocol steps to prove some high-level guarantees such as consistency under faults or preservation of private information. With MOMMIE one can prove the high-level guarantees of an application in terms of the high-level security properties. Even though the application is implemented in different ways as it is optimized for each set of resources, the high-level guarantees are maintained without the application designer having to re-prove everything from scratch (which has been the only choice typically).

I am also working on a number of projects that seek to provide new, fault-tolerant solutions to existing distributed problems. I have worked both at the algorithmic/protocol layer, for instance to describe systems with weak consistency that can tolerate Byzantine faults, and at the hardware layer, defining trusted abstractions that can make existing fault-tolerant protocols less complex and more powerful. In fact, my work on customized secure networked systems and MOMMIE comes precisely in response to my experiences building large-scale fault-tolerant and secure distributed systems, and the problems I faced in the process.

Service

I have been involved in the organization of the following conferences and workshops.

• NSDI 2010
• NetDB 2009
• HotOS 2009
• EuroSys 2009
• NSDI 2009
• HotDep 2008 (co-chair)
• DSN/PDS 2008
• EuroSys 2008
• FAST 2008
• HotDep 2007
• NSDI 2007
• HotDep 2006
• DISC 2006

Collaborators

Over the years, I have had the pleasure of supervising a number of talented students, some of whom I still actively collaborate with, and some of whom have moved on to bigger and better things. Here is a partial list:

• Tyron Stading, undergrad senior thesis at Stanford in 2002. Now in startupland (Innography).
• Aydan Yumerefendi, summer intern in 2004 at Intel Research Berkeley. Now doing wonders while finishing his PhD at Duke.
• Tyson Condie, summer intern in 2005 and 2006, and year-round current collaborator at Intel Research Berkeley.
• Boon Thau Loo, summer intern in 2005 at Intel Research Berkeley. Now a professor at UPenn.
• Atul Singh, intern for much of 2005 to 2007 at Intel Research Berkeley. Now finishing his PhD at Rice and the Max Planck Institute for Software Systems.
• Varun Kacholia, several class projects at Intel Research Berkeley in 2005. Now at Google.
• Byung-Gon Chun, summer intern in 2006 at Intel Research Berkeley and postdoc at ICSI in 2007-2008.
• Eric Yu-En Lu, intern in 2006 at Intel Research Berkeley. Finishing his PhD at Cambridge University.
• Prince Mahajan, summer intern in 2007 at Intel Research Berkeley. Continuing his Ph.D. studies at UT Austin.
• Rusty Sears, intern in 2007 at Intel Research Berkeley. Continuing his Ph.D. studies at UC Berkeley.
• Charalambos (Babi) Papamanthou, summer intern in 2008 at Intel Research Berkeley. Continuing his Ph.D. studies at Brown University.

I have also had the great opportunity to be in the thesis committees of several brilliant students:

• Varun Kacholia (Master's at UC Berkeley)
• Aydan Yumerefendi (Ph.D. at Duke)
• Atul Singh (Ph.D. at Rice)
• Nalini Belaramani (Ph.D. at UT Austin)
• Harry Li (Ph.D. at UT Austin)
• Robert Soulé (Ph.D. at NYU)

Peer-Reviewed Publications

Technical Reports

Talks and Posters

• "Building Diagnosable Distributed Systems." A talk given at the Security Crystal Ball workshop at ICSI, Berkeley, CA. January 19, 2006. [PDF]
• "Declarative Overlays." A talk given at the Information Systems Seminar at the Department of Informatics, University of Athens, Greece (in Greek). December 20, 2005. [PDF]
• "A Guided Research Tour." An overview of my current (and recent) research, given at the Intel Research Cambridge seminar. November 5, 2005. [PDF]
• "Declarative Overlays." A talk given at the PDOS group at MIT and at the Syrah group at Harvard, September 17, 2005. [PDF]
• "The LOCKSS P2P Digital Preservation System." A talk given at the Colloquium of the CS Department at Duke University, November 15, 2004. [PDF]
• "Providing Packet Obituaries." A talk given to the UC Berkeley Systems Lunch, October 22, 2004. [PDF]
• "Providing Packet Obituaries." A poster presented at the Intel Research Open House, October 15, 2004. [PDF]
• "Attrition Defenses in P2P Systems." A poster presented at the Intel Research Open House, October 15, 2004. [PDF]
• "Preserving Peer Replicas By Rate-Limited Sampled Voting." A talk given to present the paper by the same name at the 19th ACM Symposium on Operating Systems Principles, October 20, 2003. [PowerPoint]
• "Historic Integrity in Distributed Systems." My Ph.D. Thesis Defense, given on September 9th, 2002. [PDF]
• "Secure History Preservation Through Timeline Entanglement." A talk given to present the paper by the same name at USENIX Security 2002, August 8, 2002. [PowerPoint]
• "Secure History Preservation Through Timeline Entanglement." A poster presented during the 2002 Stanford-Berkeley Day, February 23rd, 2002. [PowerPoint][PS][PDF]
• "Historic Integrity in Peer-to-Peer Systems." A talk given at the Stanford Networking Seminar on February 21st, 2002. [PowerPoint]
• "Enabling the Archival Storage of Signed Documents." A talk given at the 2002 USENIX Conference on File and Storage Technologies (FAST). [PowerPoint]