Evaluating Poisoning Resistance and Blockchain Provenance in Federated Intrusion Detection Across Institutions

Srijib Samanta

Evaluating Poisoning Resistance and Blockchain Provenance in Federated Intrusion Detection Across Institutions

Author(s)	Mr. Srijib Samanta
Country	India
Abstract	Abstract: Institutions can share intrusion detection models without pooling their network records, but federated learning remains vulnerable to corrupted training updates. Collaborative training also creates a need to reconstruct how a deployed model was produced and approved. This conceptual paper develops an architecture and evaluation protocol that examine poisoning resistance and auditability as separate outcomes. A focused review informs an institutional threat model, off-chain update screening, robust aggregation and permissioned records of model lineage. The protocol compares ordinary federated averaging, independently witnessed signed logs and blockchain-backed coordination under matched training conditions. Public intrusion detection datasets support controlled institutional partitions, which are distinguished from real deployments. The analysis shows why authenticated provenance cannot establish label correctness or update safety. It identifies targeted attack success, worst-institution detection performance, honest-client exclusion, evidence reconstruction and infrastructure cost as complementary evaluation endpoints. No training experiments or measured improvements are reported. The contribution is a controlled study design for determining whether shared ledger custody offers practical accountability beyond a strong signed-log alternative. Keywords: federated learning; intrusion detection; model poisoning; permissioned blockchain; provenance; institutional cybersecurity.
Keywords	Abstract: Institutions can share intrusion detection models without pooling their network records, but federated learning remains vulnerable to corrupted training updates. Collaborative training also creates a need to reconstruct how a deployed model was produced and approved. This conceptual paper develops an architecture and evaluation protocol that examine poisoning resistance and auditability as separate outcomes. A focused review informs an institutional threat model, off-chain update screening, robust aggregation and permissioned records of model lineage. The protocol compares ordinary federated averaging, independently witnessed signed logs and blockchain-backed coordination under matched training conditions. Public intrusion detection datasets support controlled institutional partitions, which are distinguished from real deployments. The analysis shows why authenticated provenance cannot establish label correctness or update safety. It identifies targeted attack success, worst-institution detection performance, honest-client exclusion, evidence reconstruction and infrastructure cost as complementary evaluation endpoints. No training experiments or measured improvements are reported. The contribution is a controlled study design for determining whether shared ledger custody offers practical accountability beyond a strong signed-log alternative. Keywords: federated learning; intrusion detection; model poisoning; permissioned blockchain; provenance; institutional cybersecurity.
Field	Engineering
Published In	Volume 8, Issue 5, September-October 2026
Published On	2026-10-21

View / Download PDF File

About IJFMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijfmr.com

International Journal For Multidisciplinary Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Evaluating Poisoning Resistance and Blockchain Provenance in Federated Intrusion Detection Across Institutions

Share this