Virginia Tech
Tenko: A Zero Trust Inspired Framework for Real-Time Network Defense via Intelligent Thresholding and Node-Level Anomaly Scoring
Abstract
dc:description.abstractThis thesis introduces Tenko, a lightweight, online, and unsupervised anomaly detection framework that redefines the scoring and management of network intrusions in real time. While Kitsune—the current state-of-the-art unsupervised Network Intrusion Detection System (NIDS)—performs well in controlled environments, its design assumptions prove unsustainable for online deployment. Kitsune's reliance on domain-specific and optimum thresholds and per-packet anomaly scores, coupled with the absence of a mechanism to track node behavior over time, results in noisy alerts, unsmoothed attack patterns, and limited interpretability of node trust. Tenko addresses these limitations by implementing a node scoring mechanism that evolves over time through a lightweight streaming algorithm inspired by Misra-Gries sketches. This approach accumulates each node's historical anomaly signals and applies exponential damping, allowing recent behavior to weigh more heavily while preserving long-term trends. This method smooths transient spikes in anomaly scores and enables meaningful differentiation between sustained malicious behavior and benign fluctuations. The system replaces Kitsune's per-observation anomaly detection with dynamic, node-specific trust evolution, updated in real time. To support scalable and distributed trust tracking, Tenko integrates a blockchain-based offloading mechanism, enhancing the secure storage and retrieval of dormant or migrating node histories across network slices or sessions. This decentralized layer ensures that historical trust is portable, persistent, and scalable with distributivity , aligning with key principles of Zero Trust without requiring prior knowledge or labeled data. The proposed system is evaluated using the complete Kitsune dataset, encompassing nine real-world attack scenarios. Compared to the original Kitsune implementation, Tenko demonstrates significantly improved anomaly detection performance, including higher true positive rates, lower false positives and false negatives, and better per-node attribution in online unsupervised conditions. These findings establish Tenko as a robust and practical anomaly detection solution for dynamic, real-world network environments.
Degree
thesis:*- Name thesis:degree_name
- Master of Science
- Level thesis:degree_level
- masters
- Discipline thesis:degree_discipline
- Computer Science & Applications
- Department dc:contributor.department
- Computer Science and Applications
- Grantor dc:publisher
- Virginia Tech
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bhola, Sahil
- Chairs dc:contributor.committeechair
-
- Burger, Eric William
- Cameron, Melissa
- Committee member dc:contributor.committeemember
-
- Ji, Bo
Subjects
dc:subject × 2Rights
dc:rights- Statement dc:rights
-
- In Copyright
- Licence dc:rights.uri
- Language dc:language.iso
- en
Identifiers
dc:identifier.*- Dc Identifier Other
- vt_gsexam:44185