Please use this identifier to cite or link to this item:
https://hdl.handle.net/2440/139311
Citations | ||
Scopus | Web of Science® | Altmetric |
---|---|---|
?
|
?
|
Full metadata record
DC Field | Value | Language |
---|---|---|
dc.contributor.author | Goel, D. | - |
dc.contributor.author | Neumann, A. | - |
dc.contributor.author | Neumann, F. | - |
dc.contributor.author | Nguyen, H. | - |
dc.contributor.author | Guo, M. | - |
dc.contributor.editor | Paquete, L. | - |
dc.date.issued | 2023 | - |
dc.identifier.citation | Proceedings of the Genetic and Evolutionary Computation Conference (GECCO '23), 2023 / Paquete, L. (ed./s), pp.1348-1356 | - |
dc.identifier.isbn | 9798400701191 | - |
dc.identifier.uri | https://hdl.handle.net/2440/139311 | - |
dc.description.abstract | We study a Stackelberg game between one attacker and one defender in a configurable environment. The defender picks a specific environment configuration. The attacker observes the configuration and attacks via Reinforcement Learning (RL trained against the observed environment). The defender's goal is to find the environment with minimum achievable reward for the attacker. We apply Evolutionary Diversity Optimization (EDO) to generate diverse population of environments for training. Environments with clearly high rewards are killed off and replaced by new offsprings to avoid wasting training time. Diversity not only improves training quality but also fits well with our RL scenario: RL agents tend to improve gradually, so a slightly worse environment earlier on may become better later. We demonstrate the effectiveness of our approach by focusing on a specific application, Active Directory (AD). AD is the default security management system for Windows domain networks. AD environment describes an attack graph, where nodes represent computers/accounts/etc., and edges represent accesses. The attacker aims to find the best attack path to reach the highest-privilege node. The defender can change the graph by removing a limited number of edges (revoke accesses). Our approach generates better defensive plans than the existing approach and scales better. | - |
dc.description.statementofresponsibility | Diksha Goel, Aneta Neumann, Frank Neumann, Hung Nguyen, Mingyu Guo | - |
dc.language.iso | en | - |
dc.publisher | Association for Computing Machinery | - |
dc.rights | © 2023 by the Association for Computing Machinery, Inc. (ACM). | - |
dc.source.uri | https://dl.acm.org/doi/proceedings/10.1145/3583131 | - |
dc.subject | Active directory; reinforcement learning; evolutionary diversity optimization; attack graph | - |
dc.title | Evolving Reinforcement Learning Environment to Minimize Learner's Achievable Reward: An Application on Hardening Active Directory Systems | - |
dc.type | Conference paper | - |
dc.contributor.conference | Genetic and Evolutionary Computation Conference (GECCO 2023) (15 Jul 2023 - 15 Jul 2023 : Lisbon, Portugal) | - |
dc.identifier.doi | 10.1145/3583131.3590436 | - |
dc.publisher.place | New York, NY | - |
dc.relation.grant | http://purl.org/au-research/grants/arc/DP190103894 | - |
dc.relation.grant | http://purl.org/au-research/grants/arc/FT200100536 | - |
pubs.publication-status | Published | - |
dc.identifier.orcid | Goel, D. [0000-0001-8212-8793] | - |
dc.identifier.orcid | Neumann, A. [0000-0002-0036-4782] | - |
dc.identifier.orcid | Neumann, F. [0000-0002-2721-3618] | - |
dc.identifier.orcid | Nguyen, H. [0000-0003-1028-920X] | - |
dc.identifier.orcid | Guo, M. [0000-0002-3478-9201] | - |
Appears in Collections: | Computer Science publications |
Files in This Item:
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.