Strategic attacks on trust models via bandit optimization
Strategic attacks on trust models via bandit optimization
Trust and reputation systems are designed to mitigate risks associated with decisions to rely upon systems over which there is no direct control. The effectiveness of trust models are typically evaluated against relatively shallow metrics regarding the sophistication of potential attacks. In reality, such systems may be open to strategic attacks, which need to be investigated in-depth if trust model resilience is to be more fully understood. Here, we devise an orchestrated attack strategy for a specific state-of-the-art statistical trust model (HABIT). We evaluate how these intelligent attack strategies can influence predictions of target trustworthiness by this model. Our conjecture is that this approach represents a stronger benchmark for the assessment of trust models in general.
87-95
Güneş, Taha D.
19edfddb-8c8e-49f1-83eb-f7fca56e737a
Tran-Thanh, Long
e0666669-d34b-460e-950d-e8b139fab16c
Norman, Timothy J.
663e522f-807c-4569-9201-dc141c8eb50d
14 July 2018
Güneş, Taha D.
19edfddb-8c8e-49f1-83eb-f7fca56e737a
Tran-Thanh, Long
e0666669-d34b-460e-950d-e8b139fab16c
Norman, Timothy J.
663e522f-807c-4569-9201-dc141c8eb50d
Güneş, Taha D., Tran-Thanh, Long and Norman, Timothy J.
(2018)
Strategic attacks on trust models via bandit optimization.
CEUR Workshop Proceedings, 2154, .
Abstract
Trust and reputation systems are designed to mitigate risks associated with decisions to rely upon systems over which there is no direct control. The effectiveness of trust models are typically evaluated against relatively shallow metrics regarding the sophistication of potential attacks. In reality, such systems may be open to strategic attacks, which need to be investigated in-depth if trust model resilience is to be more fully understood. Here, we devise an orchestrated attack strategy for a specific state-of-the-art statistical trust model (HABIT). We evaluate how these intelligent attack strategies can influence predictions of target trustworthiness by this model. Our conjecture is that this approach represents a stronger benchmark for the assessment of trust models in general.
Text
paper8
- Version of Record
Available under License Other.
More information
Published date: 14 July 2018
Identifiers
Local EPrints ID: 423211
URI: http://eprints.soton.ac.uk/id/eprint/423211
ISSN: 1613-0073
PURE UUID: 70ef5624-71d2-4a6f-9723-edb45ecbd23e
Catalogue record
Date deposited: 19 Sep 2018 16:30
Last modified: 06 Jun 2024 01:55
Export record
Contributors
Author:
Taha D. Güneş
Author:
Long Tran-Thanh
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics