Secure authentication from ambient audio: Difference between revisions
| (6 intermediate revisions by the same user not shown) | |||
| Line 8: | Line 8: | ||
== Project Description == | == Project Description == | ||
In | In preliminary studies, fingerprints from ambient audio have been utilised to generate a secure key among devices without the necessitiy of a trusted third party | ||
[[ | ([http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6112755 1], [http://link.springer.com/chapter/10.1007%2F978-3-642-30973-1_31 2], [http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6246148 3]). | ||
In this implementation, similar recordings of ambient audio from mobile devices in proximity are utilised for the generation of a secure key. | |||
An attacker that is farther away is not able to guess the same key with a recording of her own since the features in the ambient audio differ with location. | |||
However, the robustness of the generated key is dependent on properties of ambient audio. | |||
In previous studies, we experienced best pairing characteristics when devices are in an environment with a single loud audio source. | |||
When environmental noise is, however, more ubiquitous and noisy, it is easier for an attacker to generate a key that is similar to that of the legitimate communication partners. | |||
In this project we will develop a mechanism to adapt the parameter of the underlying approach to generate secure to changing environmental parameters. | |||
[[File:AudioSynch.png]] | |||
== Required Skills == | == Required Skills == | ||
* Natural curiosity, high motivation and a good sense of creativity :) | * Natural curiosity, high motivation and a good sense of creativity :) | ||
* | * Some basic understanding of machine learning techniques, audio-fingerprinting, error correction or fuzzy cryptograpy might be helpful but will also be acquired throughout the project (Since there are implementations already, none of these have to be done from scratch) | ||
* | * Experience with audio processing might be helpful but is not mandatory | ||
Latest revision as of 13:35, 31 July 2014
Details
| Supervisor: | Stephan Sigg |
| Duration: | 3-6 months |
| Type: | Master Thesis or Student Project |
| Status: | open |
Project Description
In preliminary studies, fingerprints from ambient audio have been utilised to generate a secure key among devices without the necessitiy of a trusted third party (1, 2, 3).
In this implementation, similar recordings of ambient audio from mobile devices in proximity are utilised for the generation of a secure key. An attacker that is farther away is not able to guess the same key with a recording of her own since the features in the ambient audio differ with location. However, the robustness of the generated key is dependent on properties of ambient audio. In previous studies, we experienced best pairing characteristics when devices are in an environment with a single loud audio source. When environmental noise is, however, more ubiquitous and noisy, it is easier for an attacker to generate a key that is similar to that of the legitimate communication partners.
In this project we will develop a mechanism to adapt the parameter of the underlying approach to generate secure to changing environmental parameters.
Required Skills
- Natural curiosity, high motivation and a good sense of creativity :)
- Some basic understanding of machine learning techniques, audio-fingerprinting, error correction or fuzzy cryptograpy might be helpful but will also be acquired throughout the project (Since there are implementations already, none of these have to be done from scratch)
- Experience with audio processing might be helpful but is not mandatory