VUB | COMO
You are here: Home \ Partners \ VUB | COMO

I. Description of UVB/COMO

The Artificial Intelligence lab was founded in 1982 and has a long-standing tradition in complex systems. The Computational Modeling group (COMO) of the AIlab headed by Prof. Dr. Ann Nowé and Prof. Dr. Bernard Manderick, that coordinates this project, focuses on Machine Learning and Learning in Multi-agent Systems. The group has experience in a wide range of learning techniques such as: Reinforcement Learning, Genetic Algorithms, Neural Networks, Support Vector Machines, Graphical models including Bayesian Networks, Genetic algorithms, Recommender Systems etc.  COMO was also the coordinator of the Machine Learning and Data Mining research network, funded by the National Science Foundation(FWO).

Besides projects oriented towards fundamental research, typically supported by FWO, the research group also participates in projects in collaboration with industry including EU and SBO projects funded by the Institute for the Promotion of Innovation by Science and Technology in Flanders (IWT). The focus of the contribution is to develop and evaluate algorithms and protocols for data mining, modeling and distributed control.  Application domains include: telecommunication, smart grids, mechatronics, planning and scheduling. The COMO group was also involved in the set up of a spin-off “Enlighten Bioscience” supported by the Spin-off in Brussels program funded through the INNOVIRIS as a result of a Bioinformatics project in collaboration with the IRIDIAlab of the ULB.

More information on this team member can be found at http://ai.vub.ac.be/

II. Key persons to the project

Prof Dr. Ann Nowe graduated from the University of Ghent in 1987, where she studied mathematics with optional courses in computer science. Then she became a research assistant at the University of Brussels where she finished her PhD in 1994 in collaboration with Queen Mary and Westfield College, University of London. The subject of her PhD is located in the intersection of Computer Science ( A.I.), Control Theory (Fuzzy Control) and Mathematics (Numerical Analysis, Stochastic Approximation). After a period of 3 years as senior research assistant at the VUB, she became a Postdoctoral Fellow of the Fund for Scientific Research-Flanders (F.W.O.). Nowadays, she is a professor both in the Computer Science Department of the faculty of Sciences as in the Computer Science group of the Engineering Faculty.
Dr. Kyriakos Efthymiadis is a post-doctoral researcher in the AI Lab at Vrije Universiteit Brussels. He obtained his PhD at the University of York, UK, working on knowledge revision and reward shaping in reinforcement learning. His expertise lies in deep learning and reinforcement learning. He is titular of the Multi-agent Learning Seminar a course taught to MSc level students. Within SeCloud he was working on the machine learning aspect of detecting potentially unsafe code in cloud based applications.

III. Contributions to the project

VUB/COMO has applied machine learning to predict interaction results of web and cloud applications with external services. This interaction model is consumed by JIPDA to effectively analyze web and cloud applications that typically orchestrate over various third-party external services. VUB/COMO also developed a technique using machine learning to predict the presence of sources and sinks in an application to assist developers in specifying security policies. Sources and sinks are the points where information enters and leaves an application, respectively, and are therefore crucial for securing the flow of sensitive information.

  • D2.3.1 specifies a meta-model for sharing vulnerability and interaction models of cloud services.
  • D2.3.2 presents a technique for obtaining an interaction model from run-time observations and making it predictive using machine learning.
  • D2.3.3 describes a technique and tool detecting vulnerabilities in the API of JavaScript codebases to assist developers in deploying security policies. A demo application that classifies sinks and sources in node.js applications is available, as well as a demonstrator of this tool.

As a joint effort we combined our techniques and tools in a single artefact in the form of a plugin for Visual Studio Code, a popular code editor. The plug-in works on JavaScript files that are part of a web or cloud applications and contain Guardia policies. It automatically detects possible sources and sinks in the code and points to access control violations that could be determined statically. It is also capable of instrumenting the program and running it to provide the results of dynamic access and information flow control.

  • The code for this plugin is available, together with a tool paper which we submitted. The tools and techniques that are combined in this artifact are covered by the deliverables, demonstrators, and publications mentioned above.

Institute: Vrije Universiteit Brussel

Research Unit: COMO

Projects: C-Cure, SeCloud