TUM Logo

Anomaly Detection Challenges

Anomaly Detection Challenges  

Praktika 6sws / 10ects
Veranstalter:
Zeit und Ort:

Mo, 16:00 – 17:30 Uhr, 01.06.011

Beginn: 2015-10-19

The lecture is given in english
The slides are available in english

Updates

  • Challenge 5 is published, as well as homework 7.
  • Challenge 4 is published. This time we have a spam detection task, however in the fifth challenge we will be back to analyzing malware.
  • As the deadline for challenge 3 is postponed, next class (07.12.) contains a tutorial in Bayesian Networks and Belief Propagation instead of student presentations.
  • Challenge 3 is published
  • Challenge 2 is published
  • Challenge 1 is published and the participants have received invitations.
  • Schedule for the course is published below. There will be 5 challenges and 7 short homeworks. Students have two weeks for each homework and challenge. For challenges there are presentation days, where students present and discuss their solutions. In addition to the presentation, students need to submit a short report about their solutions for each challenge, explaining what methods they used to solve their tasks. Deadline for each report is midnight before the presentation day.
  • Kick-off meeting: 03.07.2015. at 14:30, room 01.08.033; Presentation from the meeting can be found here.
  • How to apply?

 

  1. Select our course in the matching system
  2. Send us an e-mail with a very short CV, indicating your previous knowledge and experience relevant to the topic of the course (e.g. Machine Learning or IT Security course, internships, course projects...). This should be sent by 10.07. 12.07.2015. (update) to Bojan Kolosnjaji.

 

Schedule

Date Topic Assignments
19.10.2015.

Intro to numpy

Machine Learning in Python

PDF, IPython Notebook, CSV

 

 

Homework 1

 

 

 

26.10.2015.

From linear regression to SVM

PDF, HW2, dataset1, dataset2

Homework 2 & Challenge 1
02.11.2015.

Approaches in anomaly detection

PDF, HW3

Homework 3
09.11.2015. Presentation day for challenge 1 Challenge 2
16.11.2015.

Semi-supervised Learning

PDF, HW4

Homework 4
23.11.2015.

Introduction to malware analysis systems

Ensemble Learning

PDF, HW5

Challenge 3
30.11.2015. Presentation day for challenge 2 Homework 5
07.12.2015.

Bayesian Networks, Belief Propagation

PDF HW6

Homework 6

14.12.2015.

Presentation day for challenge 3 Challenge 4
21.12.2015.

Belief Propagation 2

HW7 (deadline 17.01.)

 

Challenge 5
11.01.2016.

Deep Learning for anomaly detection

Presentation day for challenge 4

None
18.01.2016. Presentation day for challenge 5 None

 

Description

In information security domain, anomaly detection gains its own importance from researchers day by day. It is discovered that occasionally emerged frauds or intrusions in modern information systems have incurred significant loss when the suspicious activities were not detected or inefficiently processed. Therefore, numerous efforts have been devoted to devise effective detection methods, and then proceed to automation of anomaly identification in real time. In this practical course, we challenge students with a set of well designated learning tasks, which require the participants to understand the task scenario entirely and to design their own algorithms to identify corresponding anomalies. A competition on discovering the anomalies will be conducted based on the performance of their algorithms.

Requirement

  • Basic knowlege of Machine learning and data mining
  • Good programming skills
  • Basic knowlege in Python
  • Registered Master or Diplom students in Informatics or a similar area

Instruction

ResizedImage350500-instruction.png

 

 

 

 

Contact

Bojan Kolosnjaji

Email: kolosnjaji@sec.in.tum.de

Room: 01.08.061

 

Huang Xiao

Email: xiaohu@in.tum.de

Room: 01.08.057

 

George Webster

Email: webstergd@sec.in.tum.de

Room: 01.08.057