Back to search

Clemson University

A Generalizable and Privacy-Preserving Framework for Anomaly Detection in Heterogeneous IoT Environments

Abstract

dc:description.abstract

With the rapid growth of Internet of Things (IoT) devices across various sectors, detecting anomalies in such systems has become increasingly challenging. IoT environments produce diverse and evolving data streams, often lacking labeled examples, which limits the effectiveness of traditional machine learning models. These models typically require frequent retraining and struggle to adapt to new deployment conditions. This thesis proposes a flexible, privacy-aware framework for anomaly detection in multivariate time series data generated by heterogeneous IoT systems. The approach integrates a long short-term memory variational autoencoder (LSTM-VAE) with contrastive learning and adversarial adaptation, enabling the model to generalize across domains, even in few-shot or zero-shot situations where labeled target data is unavailable. A central contribution of this work is the use of domain-specific adapter layers that allow the model to adapt to new environments without needing access to the raw target data, thus preserving privacy. Additionally, the framework segments traffic based on destination IP addresses instead of fixed time windows, retaining communication context and enhancing the relevance of input sequences. The proposed method is evaluated using real-world IoT datasets from industrial, civilian, and military settings. Results demonstrate improved anomaly detection accuracy and strong cross-domain adaptability. Overall, this thesis introduces an effective, scalable, and privacy-respecting solution for securing dynamic IoT networks, capable of addressing the limitations of conventional methods and defending against emerging cyber threats.

Degree

thesis:*
Name thesis:degree_name
Master of Science (MS)
Level thesis:degree_level
Thesis
Discipline thesis:degree_discipline
Electrical and Computer Engineering (Holcomb Dept. of)

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Rezakhani, Mahshid
Contributors dc:contributor
  • Dr. Fatemeh Afghah
  • Dr. Richard Brooks
  • Dr. Yongkai Wu

Subjects

dc:subject × 7

Identifiers

dc:identifier.*
Repository record dc:identifier
https://open.clemson.edu/all_theses/4575

Chain of custody

source
Harvested from
Clemson University
Base URL
open.clemson.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Rezakhani, Mahshid. A Generalizable and Privacy-Preserving Framework for Anomaly Detection in Heterogeneous IoT Environments. Thesis thesis, https://open.clemson.edu/all_theses/4575