{"id":"https://openalex.org/W7148327945","doi":"https://doi.org/10.48550/arxiv.2604.01127","title":"Multi-Agent LLM Governance for Safe Two-Timescale Reinforcement Learning in SDN-IoT Defense","display_name":"Multi-Agent LLM Governance for Safe Two-Timescale Reinforcement Learning in SDN-IoT Defense","publication_year":2026,"publication_date":"2026-04-01","ids":{"openalex":"https://openalex.org/W7148327945","doi":"https://doi.org/10.48550/arxiv.2604.01127"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.01127","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.01127","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.01127","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5132797074","display_name":"Saeid Jamshidi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jamshidi, Saeid","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132824003","display_name":"Negar Shahabi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shahabi, Negar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5132805079","display_name":"Foutse Khomh","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Khomh, Foutse","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001203568","display_name":"Carol Fung","orcid":"https://orcid.org/0000-0001-5726-4371"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fung, Carol","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5042033117","display_name":"Mohammad Hamdaqa","orcid":"https://orcid.org/0000-0003-4927-2755"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hamdaqa, Mohammad","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10714","display_name":"Software-Defined Networks and 5G","score":0.9922000169754028,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10714","display_name":"Software-Defined Networks and 5G","score":0.9922000169754028,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.000699999975040555,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10273","display_name":"IoT and Edge/Fog Computing","score":0.0005000000237487257,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.7282000184059143},{"id":"https://openalex.org/keywords/control","display_name":"Control (management)","score":0.46299999952316284},{"id":"https://openalex.org/keywords/controller","display_name":"Controller (irrigation)","score":0.45579999685287476},{"id":"https://openalex.org/keywords/corporate-governance","display_name":"Corporate governance","score":0.39579999446868896},{"id":"https://openalex.org/keywords/quality-of-service","display_name":"Quality of service","score":0.37700000405311584},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.35040000081062317}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.7282000184059143},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.5295000076293945},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.4837999939918518},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4812999963760376},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.46299999952316284},{"id":"https://openalex.org/C203479927","wikidata":"https://www.wikidata.org/wiki/Q5165939","display_name":"Controller (irrigation)","level":2,"score":0.45579999685287476},{"id":"https://openalex.org/C39389867","wikidata":"https://www.wikidata.org/wiki/Q380767","display_name":"Corporate governance","level":2,"score":0.39579999446868896},{"id":"https://openalex.org/C5119721","wikidata":"https://www.wikidata.org/wiki/Q220501","display_name":"Quality of service","level":2,"score":0.37700000405311584},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.3603000044822693},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.35040000081062317},{"id":"https://openalex.org/C2780378061","wikidata":"https://www.wikidata.org/wiki/Q25351891","display_name":"Service (business)","level":2,"score":0.3325999975204468},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.33169999718666077},{"id":"https://openalex.org/C160403385","wikidata":"https://www.wikidata.org/wiki/Q220543","display_name":"Queue","level":2,"score":0.28690001368522644},{"id":"https://openalex.org/C106189395","wikidata":"https://www.wikidata.org/wiki/Q176789","display_name":"Markov decision process","level":3,"score":0.28139999508857727},{"id":"https://openalex.org/C21593369","wikidata":"https://www.wikidata.org/wiki/Q1032176","display_name":"Countermeasure","level":2,"score":0.2639000117778778},{"id":"https://openalex.org/C42475967","wikidata":"https://www.wikidata.org/wiki/Q194292","display_name":"Operations research","level":1,"score":0.26109999418258667},{"id":"https://openalex.org/C2780264999","wikidata":"https://www.wikidata.org/wiki/Q7445032","display_name":"Security domain","level":2,"score":0.25870001316070557}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.01127","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.01127","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.01127","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.01127","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/16","display_name":"Peace, Justice and strong institutions","score":0.4966057538986206}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Software-Defined":[0],"Networking":[1],"(SDN)":[2],"is":[3,22],"increasingly":[4],"adopted":[5],"to":[6,12,130,220],"secure":[7],"Internet-of-Things":[8],"(IoT)":[9],"networks":[10],"due":[11],"its":[13],"centralized":[14],"control":[15,26,48],"and":[16,39,55,69,112,142,153,158,174,184,199],"programmable":[17],"forwarding.":[18],"However,":[19],"SDN-IoT":[20,86],"defense":[21,87],"inherently":[23],"a":[24,83,119],"closed-loop":[25],"problem":[27],"in":[28],"which":[29,136],"mitigation":[30,44,91,108],"actions":[31],"impact":[32],"controller":[33,67,194],"workload,":[34],"queue":[35],"dynamics,":[36],"rule-installation":[37],"delay,":[38],"future":[40],"traffic":[41,173],"observations.":[42],"Aggressive":[43],"may":[45],"destabilize":[46],"the":[47,97,116,131],"plane,":[49],"degrade":[50],"Quality":[51],"of":[52,179],"Service":[53],"(QoS),":[54],"amplify":[56],"systemic":[57],"risk.":[58],"Existing":[59],"learning-based":[60],"approaches":[61],"prioritize":[62],"detection":[63],"accuracy":[64],"while":[65],"neglecting":[66],"coupling":[68],"short-horizon":[70],"Reinforcement":[71],"Learning":[72],"(RL)":[73],"optimization":[74],"without":[75,165],"structured,":[76],"auditable":[77,163],"policy":[78,94,133],"evolution.":[79],"This":[80],"paper":[81],"introduces":[82],"self-reflective":[84],"two-timescale":[85],"solution":[88],"separating":[89],"fast":[90,98],"from":[92,218],"slow":[93,117],"governance.":[95],"At":[96,115],"timescale,":[99,118],"per-switch":[100],"Proximal":[101],"Policy":[102,209],"Optimization":[103],"(PPO)":[104],"agents":[105],"perform":[106],"controller-aware":[107],"under":[109,170,206],"safety":[110,140,159],"constraints":[111],"action":[113],"masking.":[114],"multi-agent":[120],"Large":[121],"Language":[122],"Model":[123],"(LLM)":[124],"governance":[125],"engine":[126],"generates":[127],"machine-parsable":[128],"updates":[129],"global":[132],"constitution":[134],"Pi,":[135],"encodes":[137],"admissible":[138],"actions,":[139],"thresholds,":[141],"reward":[143],"priorities.":[144],"Updates":[145],"(Delta":[146],"Pi)":[147],"are":[148],"validated":[149],"through":[150],"stress":[151,176],"testing":[152],"deployed":[154],"only":[155],"with":[156],"non-regression":[157],"guarantees,":[160],"ensuring":[161],"an":[162],"evolution":[164,210],"retraining":[166],"RL":[167],"agents.":[168],"Evaluation":[169],"heterogeneous":[171],"IoT":[172],"adversarial":[175],"shows":[177],"improvements":[178],"9.1%":[180],"Macro-F1":[181],"over":[182,186],"PPO":[183],"15.4%":[185],"static":[187],"baselines.":[188],"Worst-case":[189],"degradation":[190],"drops":[191],"by":[192,197],"36.8%,":[193],"backlog":[195],"peaks":[196],"42.7%,":[198],"RTT":[200],"p95":[201],"inflation":[202],"remains":[203],"below":[204],"5.8%":[205],"high-intensity":[207],"attacks.":[208],"converges":[211],"within":[212],"five":[213],"cycles,":[214],"reducing":[215],"catastrophic":[216],"overload":[217],"11.6%":[219],"2.3%.":[221]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-03T00:00:00"}