Please use this identifier to cite or link to this item:
http://dx.doi.org/10.25673/124539| Title: | Multi Layer Security Framework for Internet of Things Systems |
| Author(s): | Saud, Rasha Aboud Ajlan, Iman Kadhim |
| Granting Institution: | Hochschule Anhalt |
| Issue Date: | 2026-06 |
| Extent: | 1 Online-Ressource (11 Seiten) |
| Language: | English |
| Abstract: | The exponentially growing presence of Internet of Things (IoT) devices has immensely revolutionized many application areas like healthcare, manufacturing, smart cities, etc. but it also faces serious security issues, because most of the IoT devices are not equipped with formidable computational resources and security features so make them vulnerable for malware threats, signature- and rule-based solutions are ineffective against zero day threats and polymorphic threats and deep learning solutions are computationally expensive for resource constrained edge devices. In this paper, a lightweight, multi-layer IoT malware detection framework that applies the hybrid ensemble learning with a new external data protection architecture is proposed. On the 14-variant hybrid ensemble model of XGBoost and RF classifier, the weighted fusion scheme with empirically optimized weight (=0.6, by 5-fold cross-validation grid-search) was used to optimize the detection precision with relative minimal complexity. To deal with the class imbalanced connection-oriented intrusions in the released TON_IoT dataset, the SMOTE oversampling technique was applied to the training set only, protecting minority-class samples by data-privacy without contaminating the test set. A Data Protection Layer performs IP address anonymization, field masked using MD5 and data integrity verification using SHA-256 without affecting detection performance.10 attack-types. The detection performance on the TON IOT dataset of 211,043 network traffic samples is99.56%, an F1-score of 0.9956, and a training time of only 15.83 seconds. These results show a statistically significant improvement over the MLP-4Layers deep learning baseline (McNemar's test: =2372.76, p<0.001) and is 21.9 times faster to train than the deep learning baseline. This framework effectively detects, backdoor, DDoS, DoS, injection, brute -force, scanning, ransomware, XSS and man-in-the-middle attacks. Its detection rate is near perfect which makes this framework practically feasible for deployment in real-world IoT security environment. |
| URI: | https://opendata.uni-halle.de//handle/1981185920/126473 http://dx.doi.org/10.25673/124539 |
| Open Access: | Open access publication |
| License: | (CC BY-SA 4.0) Creative Commons Attribution ShareAlike 4.0 |
| Appears in Collections: | International Conference on Applied Innovations in IT (ICAIIT) |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| ICAIIT_2026_14(3)-1-3.pdf | 1.6 MB | Adobe PDF | ![]() View/Open |
Open access publication
