Ir. S.E. (Sicco) Verwer

Ir. S.E. (Sicco) Verwer

Profiel

Bachelor Theses

Master Theses until 2018

MSc Theses 2018-2019

News

My team (Azqa Nadeem, Laurens Bliek, Chris Hammerschmidt, and me) recently participated and won the first robust malware detection challenge! We developed a new algorithm for adversarial machine learning in discrete spaces, which won both the attack and defence tracks! More information at: sites.google.com/view/advml/advml19-challenge
I got awarded a prestigious VIDI grant from TTW to continue my work on learning state machine models from software! In particular, the goal of this grant is to use properties of data coming from software such as the near absence of noise to improve learning algorithms when applied such data. I have openings for two PhD students and one scientific programmer on this project. 

Expertise

I am an assistant professor in machine learning with applications in cyber security, software engineering, and mathematical optimization. My broad interest is in the development of new machine learning technology, in particular, state machine learning. 
I focus on using machine learning for tasks other than prediction, such as analysis, optimization, control, and verification. 
My main research line is to learn interpretable models from software logs such as network traces. To this aim, we have developed the open source  flexfringe tool ( https://automatonlearning.net), an implementation of flexible learning of state machines from trace data. We have used flexfringe for learning behavioral profiles of  malware, learn models for  intrusion detection in industrial control systems, and  discover bugs in payment systems.

Key Publications

Biografie

I am assistant professor in machine learning with applications in cyber security and software engineering at TU Delft since 2014. Before this, I have been a postdoctoral researcher for several years at RU Nijmegen, KU Leuven, and TU Eindhoven. I have worked on several topics in machine learning and am best known for my work in grammatical inference, i.e., learning state machines from trace data. I have researched and implemented several algorithms for learning such models including RTI, which is one of the first that is able to learn timed automata. In 20130, I received a VENI grant from STW to extend this work and apply it in cyber security. Other recent work includes several methods for declarative modelling of machine learning problems using mathematical solvers, and making classifiers discrimination-aware.
I teach two courses in the cyber security master at TU Delft: Cyber Data Analytics and Automated Software Testing and Reverse Engineering. If you are interested in the research performed by my lab, or joining as PhD or MSc student, please have a look at my publications and past publicly available MSc and BSc theses.

Lees meer

Nevenwerkzaamheden