Learning-enabled Intrusion Detection in IoT: Current Challenges and Future Directions
Bibliographic record
Abstract
Machine and deep learning-enabled intrusion detection in the Internet of Things (IoT) still pose a significant challenge, primarily due to the heterogeneity of sensors, diverse nature of generated data, and perpetual evolution of attack vectors. This paper presents a comparative study of supervised and unsupervised learning techniques using two state-of-the-art IoT security datasets: TON-IoT and Edge-IIoT. We analyze the performance of six supervised algorithms (Random Forest, XGBoost, SVM, KNN, Logistic Regression, and Naive Bayes), three unsupervised detection methods (K-Means, Isolation Forest, and One-Class SVM), and a multilayer perceptron (MLP) deep learning model. The findings indicate that supervised approaches, particularly XGBoost and Random Forest, exhibit superior performance compared to unsupervised methods on both datasets. The discrepancies in the empirical results on the studied datasets can be attributed to the inherent characteristics of the data: namely, its raw, noisy, and heterogeneous nature for TON-IoT, as opposed to its preprocessed and more homogeneous state for Edge-IIoTset. Moreover, our thorough investigation reveals several noteworthy limitations, including significant class imbalance, variability across sensors, and the absence of discernible intrusion signals on certain devices. Finally, the practical implications for deploying an intrusion detection system in real IoT environments are discussed, including the need for lightweight, federated models capable of exploiting multi-sensor fusion.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.007 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.004 | 0.011 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.003 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".