Hasso-Plattner-Institut20 Jahre HPI
Hasso-Plattner-Institut20 Jahre HPI
  
Login
 

Behavior-Based Authentication and Machine Learning (Wintersemester 2019/2020)

Lecturer: Prof. Dr. Christoph Meinel (Internet-Technologien und -Systeme) , M.Sc. Eric Klieme (Internet-Technologien und -Systeme) , M.Sc. Christian Tietz (Internet-Technologien und -Systeme)

General Information

  • Weekly Hours: 4
  • Credits: 6
  • Graded: yes
  • Enrolment Deadline: 01.10.-30.10.2019
  • Teaching Form: Seminar
  • Enrolment Type: Compulsory Elective Module
  • Course Language: English
  • Maximum number of participants: 12

Programs & Modules

Cybersecurity MA
  • IDMG-Konzepte und Methoden
  • IDMG-Techniken und Werkzeuge
  • IDMG-Spezialisierung
Digital Health MA
IT-Systems Engineering MA

Description

Background

Passwords are used for securing computer systems for a long time. Nevertheless, people use short and weak passwords for their own accounts. The recent identity leaks (MasterCard, Collection#1 - #5) showed how easy these accounts may be compromised. To solve this problem, the Secure Identity Lab of HPI is researching approaches to let the devices (Computer, Smartphones, etc.) verify their owner's identities. This is done by detecting and evaluating different kinds of behavior.

One approach we are following is using the gait of the user to determine a trust level, a probability that the user is also the owner. This trust level is later sent to service providers for authentication. Afterwards, entering a password may not be required anymore.

Problem

There exist several requirements for ‘good’ biometrics and for ‘good’ authentication systems. In the case of biometrics these are e.g. “availability” of the biometric data in many situations or “acceptance” of the user that the data is collected.

In terms of authentication systems there are requirements for low error rates or resistances against several attacks, for example. Referring to our approach described above and the current level of research, a lot of questions are not answered yet. Our goal of deploying a real world system adds even more questions in the field of deployability.

Goal

In this seminar we want to look at several topics:

  • Advanced smartphone activity recognition (positions and interactions while walking)
  • Integrate user's into the labelling process during data collection for big scale studies
  • Evaluation of smart door handles in a real world setting

Please note that all of our topics are basically general proposals. Please feel free to join our seminar and share your own very special ideas in this field of research.

Topic Presentation
Door Handle Short

Requirements

In general, there are no specific requirements.

Nevertheless it helps if you are already familiar with topics like (depends on your final topic):

  • Python for Data Science / Engineering
  • Java/Kotlin for Android Development
  • Experiment Design
  • Machine Learning Algorithms
  • Deep Learning Architectures
  • Signal Processing

Learning

All participating students will form groups (usually 2-3 people) to work each on a specific topic.

Examination

The following disciplines will be considered for the overall grade:

  • intermediate presentation
  • final presentation
  • source code
  • project documentation
  • student activity

Dates

First Meeting

Tuesday, 15.10.2019, 9.15am in room H.E-52

Topic submission until 20th of October 2019 via email to Christian or Eric

Zurück