20656 - METHODS AND DATA ANALYTICS FOR RISK ASSESSMENT
Course taught in English
Go to class group/s: 25
Class-group lessons delivered on campus
- Probability. - Statistics. - Computer science. - Programming.
Data analytical methods are important for cyber security in two contrasting ways: On the one hand, they can be used for assessing and predicting risk in a large number of areas, and with the trend towards big data and the introduction of new machine learning paradigms this number is growing even further. In particular, Bayesian methods have become more and more important, since they are especially suited for dealing with uncertainty and also work when only a limited number of data is available. The purpose of this course is consequently two-fold: First, to teach the students the aspects of probabilistic thinking and Bayesian methods for risk prediction based on real-life applications from different fields. In the the course, we will also give an introduction to the method of Monte Carlo, which is widely used in the field in industry for simulations of processes involving risk. Time permitting we may also cover some advanced topics relating to the theory of modern cryptography.
- Bayesian approach to Probability
- Bayesian Methods in Risk Assessment
- Bayesian Networks and Causal Graphs
- Unsupervised learning in the Bayesian context
- Supervised learning in the Bayesian context
- Case studies: Risk prediction in business, software, hardware, law, medicine, insurance, climate
- Introduction to Monte Carlo Methods
- Monte Carlo Methods for Risk Assessment
- Cryptography: concepts and tools
- The Bayesian (and frequentist) view of the world
- Understand the principles and methods of supervised and unsupervised learning in Bayesian terms
- Gain an intuition for the building of a Bayesian data analysis pipeline from data collection to final predictions
- Understand the concept and purpose of Monte Carlo simulations
- Understand basic concepts in modern Cryptography
- Formulate a problem in terms of Bayesian methods
- Design and apply a Bayesian data analysis pipeline using large or small datasets
- Read current literature on the application of data analytics to risk assessment in different domains
- Implement Monte Carlo simulations and use them as tools for solving real-world problems
- Define, construct and analyze basic cryptographic primitives and assess their security
- Face-to-face lectures
- Group assignments
Group assignments: Devloping a risk model in AgenaRisk given a problem description
The exam consists of a theory part and a group project.
The theory part features open and closed answer questions. It will test the students' understanding of the Bayesian approach to modelling uncertainty, explain concepts and mathematical relations connected to it and use it for estimating numerical quantities in specified problem settings. There will be additional questions intended to test the students' familiarity with basic Monte Carlo methods. The exam is not open-book: any material outside of what are provided by the instructors is forbidden.
The group project will consist in developing a risk model based on provided literature. The students will be provided with the description of a specific scenario and will develop a model to estimate risks involved. The project will test the students ability to understand texts describing risk scenarios, design and apply a Bayesian analysis pipeline based on the text and identify sensible values for the different variables involved. This will also involve the analysis of the reliability of the model and test if the students can spot the weak points of the models they develop.
The group project will give a maximum of 16 points and the theoretical exam will give a maximum of 15 points.
For the Bayesian part: Fenton, Norman, and Martin Neil. Risk assessment and decision analysis with Bayesian networks. Crc Press, 2012. In this part, we will also use the software "AgenaRisk". Licenses for this software will be provided during the course.
Materials for the part on Monte Carlo methods will be provided during the course.