Using Near-Realtime Voice Data for Communication Authentication Andreas Brauchli, and Depeng Li

Using Near-Realtime Voice Data for
Communication Authentication
Andreas Brauchli, and Depeng Li
University of Hawai‘i at Mānoa
1. Establish communication
Problem
(insecure)
...
2. Estimate link quality
...
(latency, jitter)
Channel latency
Impostor
Public common history
Decision-error implications
Creating an authentic channel
without an authenticated Public Key
Tools
Challenges
3. Recognize voice
...
(anomalies, tension)
4. Challenge - Response
... !
Insecure channel
Known voice
known behavior
Common history
Human decider
Public key cryptography
(until convinced)
... ?
5. Describe Public Key
6. Get, verify and use PK
(fingerprint)
Human Decider
...
Realtime Aspect
Important to assess
authenticity:
Reaction to challenge
questions
Realistic impersonation
much harder
Multi-Hop link latency
...
Generally harder to fool
Intuition
Tolerance
Decision is not binary
Implementation
AllNet (future)
Fig. 1 One-sided protocol illustration
Conclusions
Humans are inherrently good at recognizing voices, behavior and detecting anomalies.
This valuable resource should always be considered when developing authentication systems.
https://www.alnt.org
Abstract:
https://www2.hawaii.edu/~andreasb/
pub/poster_realtime-voice-auth.pdf
Contact:
[email protected]
[email protected]