MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking

Md Kamrul Hossain; Walid Aljoby

doi:10.1109/NOMS69089.2026.11668334

MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking

Md Kamrul Hossain^*
, Walid Aljoby

^*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

2 Scopus citations

Abstract

In multi-intent intent-based networks, a single fault can trigger co-drift where multiple intents exhibit symptomatic KPI degradation, creating ambiguity about the true root-cause intent. We present MILD, a proactive framework that reformulates intent assurance from reactive drift detection to fixed-horizon failure prediction with intent-level disambiguation. MILD uses a teacher-augmented Mixture-of-Experts where a gated disambiguation module identifies the root-cause intent while per-intent heads output calibrated risk scores. On a benchmark with non-linear failures and co-drifts, MILD provides 3. 8 % - 9 2. 5 % longer remediation lead time and improves intent-level root-cause disambiguation accuracy by 9. 4 % - 4 5. 8 % over baselines. MILD also provides per-alert KPI explanations, enabling actionable diagnosis.

Original language	English
Title of host publication	Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026
Editors	Pierpaolo Loreti, Stefano Salsano, Eiji Oki, Lorenzo Bracciale, Marc-Oliver Pahl, Carlos Raniery Paula dos Santos, Nattapong Kitsuwan, Vinicius Fulber-Garcia
Publisher	Institute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)	9798331592684
DOIs	https://doi.org/10.1109/NOMS69089.2026.11668334
State	Published - 2026
Event	IEEE/IFIP Network Operations and Management Symposium, NOMS 2026 - Rome, Italy Duration: 18 May 2026 → 22 May 2026

Publication series

Name	Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026

Conference

Conference	IEEE/IFIP Network Operations and Management Symposium, NOMS 2026
Country/Territory	Italy
City	Rome
Period	18/05/26 → 22/05/26

Bibliographical note

Publisher Copyright:
© 2026 IEEE.

Keywords

Intent Drift
Intent-Based Networking (IBN)
Mixture-of-Experts (MoE)
Root-Cause Disambiguation

ASJC Scopus subject areas

Artificial Intelligence
Computer Networks and Communications
Modeling and Simulation
Information Systems and Management

Access to Document

10.1109/NOMS69089.2026.11668334

Cite this

Hossain, M. K., & Aljoby, W. (2026). MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking. In P. Loreti, S. Salsano, E. Oki, L. Bracciale, M.-O. Pahl, C. R. P. dos Santos, N. Kitsuwan, & V. Fulber-Garcia (Eds.), Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026 (Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/NOMS69089.2026.11668334

Hossain, Md Kamrul ; Aljoby, Walid. / MILD : Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking. Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026. editor / Pierpaolo Loreti ; Stefano Salsano ; Eiji Oki ; Lorenzo Bracciale ; Marc-Oliver Pahl ; Carlos Raniery Paula dos Santos ; Nattapong Kitsuwan ; Vinicius Fulber-Garcia. Institute of Electrical and Electronics Engineers Inc., 2026. (Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026).

@inproceedings{8851bcd521c3490fb8cdc5db5e49df57,

title = "MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking",

abstract = "In multi-intent intent-based networks, a single fault can trigger co-drift where multiple intents exhibit symptomatic KPI degradation, creating ambiguity about the true root-cause intent. We present MILD, a proactive framework that reformulates intent assurance from reactive drift detection to fixed-horizon failure prediction with intent-level disambiguation. MILD uses a teacher-augmented Mixture-of-Experts where a gated disambiguation module identifies the root-cause intent while per-intent heads output calibrated risk scores. On a benchmark with non-linear failures and co-drifts, MILD provides 3. 8 \% - 9 2. 5 \% longer remediation lead time and improves intent-level root-cause disambiguation accuracy by 9. 4 \% - 4 5. 8 \% over baselines. MILD also provides per-alert KPI explanations, enabling actionable diagnosis.",

keywords = "Intent Drift, Intent-Based Networking (IBN), Mixture-of-Experts (MoE), Root-Cause Disambiguation",

author = "Hossain, \{Md Kamrul\} and Walid Aljoby",

note = "Publisher Copyright: {\textcopyright} 2026 IEEE.; IEEE/IFIP Network Operations and Management Symposium, NOMS 2026 ; Conference date: 18-05-2026 Through 22-05-2026",

year = "2026",

doi = "10.1109/NOMS69089.2026.11668334",

language = "English",

series = "Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

editor = "Pierpaolo Loreti and Stefano Salsano and Eiji Oki and Lorenzo Bracciale and Marc-Oliver Pahl and \{dos Santos\}, \{Carlos Raniery Paula\} and Nattapong Kitsuwan and Vinicius Fulber-Garcia",

booktitle = "Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026",

address = "United States",

}

Hossain, MK & Aljoby, W 2026, MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking. in P Loreti, S Salsano, E Oki, L Bracciale, M-O Pahl, CRP dos Santos, N Kitsuwan & V Fulber-Garcia (eds), Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026. Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026, Institute of Electrical and Electronics Engineers Inc., IEEE/IFIP Network Operations and Management Symposium, NOMS 2026, Rome, Italy, 18/05/26. https://doi.org/10.1109/NOMS69089.2026.11668334

MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking. / Hossain, Md Kamrul; Aljoby, Walid.
Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026. ed. / Pierpaolo Loreti; Stefano Salsano; Eiji Oki; Lorenzo Bracciale; Marc-Oliver Pahl; Carlos Raniery Paula dos Santos; Nattapong Kitsuwan; Vinicius Fulber-Garcia. Institute of Electrical and Electronics Engineers Inc., 2026. (Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - MILD

T2 - IEEE/IFIP Network Operations and Management Symposium, NOMS 2026

AU - Hossain, Md Kamrul

AU - Aljoby, Walid

PY - 2026

Y1 - 2026

N2 - In multi-intent intent-based networks, a single fault can trigger co-drift where multiple intents exhibit symptomatic KPI degradation, creating ambiguity about the true root-cause intent. We present MILD, a proactive framework that reformulates intent assurance from reactive drift detection to fixed-horizon failure prediction with intent-level disambiguation. MILD uses a teacher-augmented Mixture-of-Experts where a gated disambiguation module identifies the root-cause intent while per-intent heads output calibrated risk scores. On a benchmark with non-linear failures and co-drifts, MILD provides 3. 8 % - 9 2. 5 % longer remediation lead time and improves intent-level root-cause disambiguation accuracy by 9. 4 % - 4 5. 8 % over baselines. MILD also provides per-alert KPI explanations, enabling actionable diagnosis.

AB - In multi-intent intent-based networks, a single fault can trigger co-drift where multiple intents exhibit symptomatic KPI degradation, creating ambiguity about the true root-cause intent. We present MILD, a proactive framework that reformulates intent assurance from reactive drift detection to fixed-horizon failure prediction with intent-level disambiguation. MILD uses a teacher-augmented Mixture-of-Experts where a gated disambiguation module identifies the root-cause intent while per-intent heads output calibrated risk scores. On a benchmark with non-linear failures and co-drifts, MILD provides 3. 8 % - 9 2. 5 % longer remediation lead time and improves intent-level root-cause disambiguation accuracy by 9. 4 % - 4 5. 8 % over baselines. MILD also provides per-alert KPI explanations, enabling actionable diagnosis.

KW - Intent Drift

KW - Intent-Based Networking (IBN)

KW - Mixture-of-Experts (MoE)

KW - Root-Cause Disambiguation

UR - https://www.scopus.com/pages/publications/105050592769

U2 - 10.1109/NOMS69089.2026.11668334

DO - 10.1109/NOMS69089.2026.11668334

M3 - Conference contribution

AN - SCOPUS:105050592769

T3 - Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026

BT - Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026

A2 - Loreti, Pierpaolo

A2 - Salsano, Stefano

A2 - Oki, Eiji

A2 - Bracciale, Lorenzo

A2 - Pahl, Marc-Oliver

A2 - dos Santos, Carlos Raniery Paula

A2 - Kitsuwan, Nattapong

A2 - Fulber-Garcia, Vinicius

PB - Institute of Electrical and Electronics Engineers Inc.

Y2 - 18 May 2026 through 22 May 2026

ER -

Hossain MK, Aljoby W. MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking. In Loreti P, Salsano S, Oki E, Bracciale L, Pahl MO, dos Santos CRP, Kitsuwan N, Fulber-Garcia V, editors, Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026. Institute of Electrical and Electronics Engineers Inc. 2026. (Proceedings of IEEE/IFIP Network Operations and Management Symposium, NOMS 2026). doi: 10.1109/NOMS69089.2026.11668334

MILD: Multi-Intent Learning and Disambiguation for Proactive Failure Prediction in Intent-Based Networking

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this