Thijs Laarhoven's Homepage

Welcome

Welcome to my homepage! My name is Thijs Laarhoven, and I am currently a research scientist at TNO, the Dutch Organization for Applied Scientific Research. Before this, I was a postdoctoral researcher at the Eindhoven University of Technology, during which I also held a temporary visiting scientist position at the University of California, Berkeley, for the Spring 2020 semester. Before that, I was a postdoctoral researcher at IBM Research in Zurich, Switzerland. And before that, I completed my PhD at the Eindhoven University of Technology, in the Netherlands.

Research

Quantum-Safe Cryptography

Shor's breakthrough work in the late 1990s demonstrated that if an attacker has a large-scale quantum computer, then many encryption schemes which are in use today (e.g. RSA, Diffie-Hellman) no longer provide security. Although such large-scale quantum computers (likely) do not exist yet, experts predict that within a few decades such computers may well exist, rendering classical forms of encryption insecure. The field of quantum-safe cryptography concerns the development of new cryptographic primitives, which offer security even against such powerful quantum adversaries.

Homomorphic Encryption

Cryptography is commonly used for data at rest (e.g. static data stored on server), and for data in transit (e.g. for securing communication). In an ideal world, however, data remains encrypted and invisible to the outside world even when it is in use. Homomorphic encryption is a special type of encryption, enabling computations to be performed on encrypted data, without having to decrypt data in between. Using this form of encryption can offer tremendous privacy benefits, but work still needs to be done to make this form of encryption practical in real-world applications.

Multiparty Computation

In a world where big data is the norm, many organizations may benefit from being able to use all this data for analytics and insights. However, this data often cannot be used for these purposes due to privacy regulations. Multiparty computation is a technology that lets multiple organizations, who each have partial data sets, to collaborate and perform analytics on their combined data sets, without ever having to disclose their data sets to the other parties. By doing joint computations and exchanging "secret-shared" data, the parties can thereby obtain improved insights.

Secure Machine Learning

With the rise of machine learning, and neural networks overtaking "classical" algorithms in terms of performance and efficiency for various tasks (from playing chess to generating images), we are heading to a world where machine learning will be the norm for arbitrary tasks. This new computing paradigm offers many opportunities, but also various challenges related to private data processing. Various privacy-enhancing technologies can be used to make machine learning more privacy-friendly, and allow the world to use these tools without having to give up their private data.

Statistics

Citations

2264

Publications

Talks

Chess title

Resume

Summary

Ever since my childhood I enjoyed solving puzzles and mathematical problems, especially those naturally appearing in real life. After graduating in applied mathematics I pursued a research career in lattice‑based cryptography, focusing on the practical hardness of solving hard lattice problems, contributing to constructing efficient lattice‑based cryptographic primitives, and studying better high‑dimensional nearest neighbor methods, which form a key ingredient for state‑of‑the‑art lattice‑based cryptanalysis.

Besides being able to work independently and setting my own agenda, one of my strengths is to find the right perspective for a complex task, and to see how techniques from other areas can be used to solve these problems efficiently. For instance, in my paper at Crypto 2015 I made a new connection between lattice algorithms and nearest neighbor searching, marking the beginning of a long line of new work on this topic.

Experience

Dec 2021 - present

Research Scientist

At TNO, the Dutch Organization for Applied Scientific Research, I contribute to bringing digital security innovations from the lab into the field. This involves going beyond academic research, seeing how technologies can land in society and make a positive impact in government or industry settings, and demonstrating their added value through proof-of-concept demonstrations of the technology. Focus areas at TNO include the migration to quantum-safe cryptography, and finding ways to collaborate on privacy-sensitive distributed data with tools such as multiparty computation and (fully) homomorphic encryption.

Nov 2017 - Nov 2021

Postdoctoral Researcher

For my postdoc at the TU/e I studied topics related to lattices and nearest neighbor searching, in the context of lattice-based (post-quantum) cryptography, and for various applications that require finding similar items in large databases. For the period 2019-2021, this research was funded by a three-year personal NWO Veni innovational research grant of EUR 250.000, which supports outstanding researchers to pursue their line of research for three years at any Dutch university.

Jan 2020 - May 2020

Visiting Scientist

During Spring 2020 I was an invited visiting scientist to the research program titled "Lattices: Algorithms, Complexity, and Cryptography" at the Simons Institute for the Theory of Computing. During this program, researchers visited from all around the world to collaborate on open questions in lattice algorithms and lattice-based cryptography, and to discuss ideas and potential new approaches, until in March 2020 the world went into lockdown, and the program came to an early end.

Mar 2016 - Sep 2017

Postdoctoral Researcher

During my time at the IBM Research lab in Zurich, Switzerland, I studied methods to further improve lattice-based cryptographic solutions, as well as methods that attempt to break these schemes, to get a better understanding of the true security of these lattice-based schemes. I further continued my research on efficient nearest neighbor techniques, which resulted in a nice joint paper on practical and optimal time-space trade-offs.

Oct 2011 - Feb 2016

Doctoral Candidate

In my PhD research, I focused on obtaining a better understanding of, and finding improvements for, various lattice algorithms that lie at the foundation of lattice-based cryptanalysis. Specifically, I focused on lattice sieving methods for the shortest vector problem, and I helped transform sieving from a mostly irrelevant theoretical novelty (2011) to the main algorithm to consider when choosing parameters (2016). The time complexity 2^{0.292n + o(n)} from our SODA 2016 paper currently still stands as the asymptotically fastest method for solving the shortest vector problem in high dimensions, while the 2^{0.265n + o(n)} from my PhD thesis was long the best quantum complexity for the shortest vector problem, until it was finally improved by Chailloux-Loyer at AsiaCrypt 2021.

Oct 2010 - Jun 2011

Research Intern

To protect copyrighted content against piracy, fingerprints or watermarks are commonly embedded in the content, allowing the distributor of the content to trace a pirate copy to the responsible user. To combat this solution, several pirates may collude, and mix their watermarked copies into a new copy, with a fingerprint which is a mix of the individual pirates' fingerprints. With fingerprinting codes it is possible to find the pirates even if they collude. Various previous solutions were aimed at non-adaptive settings, and during my internship I studied adaptive solutions, leading to the invention of the (patented) dynamic Tardos scheme.

Volunteering

Oct 2020 - Dec 2021

Lichess.org Team Member

Lichess is the second largest online chess site, and each day millions of games are played between users from all around the world. Lichess is free, open-source, ad-free, and exists as a non-profit organization funded solely through donations from users. As a team member I assisted with moderating the chat, finding cheaters, handling ban appeals, developing improved cheat detection tools, scouting and guiding new team members, and outlining the path towards the future for Lichess in terms of policies, governance, goals, vision, and more.

Education

Oct 2011 - Feb 2016

PhD in Cryptography (cum laude)

The research in my PhD at the Eindhoven University of Technology (TU/e) focused on two topics: improving collusion-resistant fingerprinting schemes, and a new direction in lattice sieving algorithms by combining them with nearest neighbor search techniques. Besides the results in these two main topics, I showed how to improve upon the state-of-the-art for group testing and nearest neighbor searching. The resulting PhD thesis titled Search problems in cryptography: From fingerprinting to lattice sieving from December 2015 can be downloaded here.

Sep 2009 - Aug 2011

MSc in Applied Mathematics (cum laude)

I graduated in the group Coding Theory and Cryptology, which is part of the section Discrete Mathematics. In the second year of my Masters I did a final project combined with an internship at Irdeto BV. This internship concluded with writing a Masters thesis, titled Collusion-resistant traitor tracing schemes, which can be downloaded here.

Sep 2006 - Aug 2009

BSc in Applied Mathematics (cum laude)

During the first three years of my studies I took various courses in both Applied Mathematics and in Computer Science. In the first year I obtained propaedeutic diplomas in both Mathematics and Computer Science, and later I also completed a so-called "minor" in Advanced Computer Science. The last part of these three years focused on discrete mathematics, which concluded with a final project and thesis about the Collatz conjecture. This thesis titled The 3n+1 conjecture can be downloaded here.