A Heterogeneous Model of TON_IoT and SideChannel-3D to Safeguard the CIA Triad
Open Access DepositedEnsemble Machine Learning for Securing 3D Printers
Alotaibi & Ilyas, 2023
What components of an ensemble machine learning (ML) model created on supervised ML models and trained on the TON_IoT and SideChannel-3D data can detect a subset of attacks that can affect a 3D printer’s CIA with a high truly malicious detection rate? A model experimentation methodology approach simulated individual and ensemble classifiers to accurately detect benign and malicious instances for relevant attack vectors with high rates of true malicious detection (i.e., recall). Two data sources were used to train these models, TON_IoT and SideChannel-3D. TON_IoT datasets are central to cybersecurity research for intrusion and detection. However, the datasets are not specific to 3D printers, or any other single IoT device, but rather includes IoT network sensors to generalize IoT devices. To leverage TON_IoT for specifically 3D printers, the praxis maps the properties of IoT devices with the characteristics of network 3D printers together with representing network traffic for a subset of known 3D printer attack vectors. Ultimately this dataset was used to train the confidentiality and availability components of the CIA triad. The SideChannel-3D which is specific to 3D printers trains the G-code segment data for tampered 3D printed objects affecting integrity. The results of the analysis displayed a complete success in constructing an ensemble stacking detector with a series of base classifiers and logistic regression meta model achieving a recall greater than the 98.2% benchmark (the average of 3 relevant sources). The high recall rate concludes that a high number of truly malicious instances and events are detected across all members of the CIA Triad. The mean recall observed in this praxis outcomes is 98.97%, performing better than the average of the 3 sources, (Booij et al., 2022
The Fourth Industrial Revolution (Industry 4.0) shifts the paradigms of manufacturing from automation to an interconnected system of manufacturing. Industry 4.0 encompasses Internet of Things (IoT), cloud computing, Artificial Intelligence (AI), Additive Manufacturing, and other emerging technologies. 3D printers can now integrate with other IoT devices wirelessly on networks, elevating them from simple stand-alone tools. 3D printers possess sensors and detectors that collect valuable data for analysis both in real-time and offline. Although these IoT features provide efficiency benefits and quality enhancements, they also invite malicious attackers to target all members of the confidentiality, integrity, and availability (CIA) triad.An Ensemble Machine Learning detection model based on IoT Attacks is a valuable capability to safeguard 3D printers against a segment of targeted attacks, ultimately preserving the CIA triad. One fundamental question addresses this
Yu et al., 2020). This praxis successfully filled gaps in prior research by (1) combining two established detection methods into a single ensemble stacking model (2) generating additional representative features for SideChannel-3D data (3) mapping characteristics of IoT devices to networked 3D printers proving they are IoT devices (4) constructing an attack vector schematic to map them to CIA risk and state (5) advancing the IoT detection capabilities using TON_IoT.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.