University of Thessaly (UTH)
Ευφυές υπολογιστικό σύστημα για την ανίχνευση ελαττωμάτων και ανωμαλιών στην Ηλεκτρολογία
Abstract
dc:descriptionIn our days there is a need for even more and more sophisticated circuits. The design companies to reduce the operating costs and facilitate mass production of integrated circuits, outsource their fabrication to third-party foundries. This process increases the risk of intrusion attacks in the form of hardware viruses, also known as hardware trojans (HTs) viruses. HTs viruses are a critical problem that has the potential to become an outbreak in the coming years, presenting a significant threat both technologically and socially.The majority of the studies are concerned with the development of countermeasures against HTs for Field-Programmable Gate Array (FPGA) circuits at the post-silicon stage. Also, there is limited information and published studies for the Application-Specific Integrated Circuits (ASICs) and specifically for the pre-silicon stage. ASICs are challenging due to the variety of design phases especially on the pre-silicon stage and for the need for professional tools for the design of each phase.In this thesis, we studied several phases for the design process on ASICs and we found that there is a general lack of free benchmark circuits and also there is a high imbalance problem between uninfected and infected benchmark circuits. We used and designed all the limited benchmark circuits for the Gate-Level Netlist (GLN) phase of ASICs with a professional tool and extracted area, power and time analysis features. We developed our Machine Learning (ML) classification models based on this limited data and we observed that the lack of samples leads to the development of imbalanced and no robust ML-based classification approaches against HTs viruses. We solved the problem of the limited data with the development of our Deep Learning (DL) - Generative Adversarial Networks (GANs) models which were able to synthesize new generated data based on our real limited data. GANs are novel DL algorithms that are used in the computer vision field for generating artificial images and it was the first time that GANs were used in this research field. Based on our new generated data we developed a robust ML-based classifier as a countermeasure against HTs at the GLN phase and compared it with existing methods. Finally, we turned our generative model into a free tool to be used as a solution for dealing with the limited number of data.
Degree
thesis:*- Grantor dc:publisher
- University of Thessaly (UTH)
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Liakos, Konstantinos
- Λιάκος, Κωνσταντίνος
Subjects
dc:subject × 32- Ασφάλεια υλικού
- Ολοκληρωμένα κυκλώματα
- ολοκληρωμένα κυκλώματα ειδικής εφαρμογής
- ιοί δουρείου ίππου συσκευών
- Αντίμετρα
- Τεχνητή νοημοσύνη
- Μηχανική μάθηση
- μάθηση σε βάθος
- παραγωγική μάθηση
- παραγωγικά αντιφατικά δίκτυα
- προ πυριτίου στάδιο
- φάση επιπέδου πύλης
- Hardware security
- Integrated circuits
- application-specific integrated circuits
- hardware trojan viruses
- countermeasures
- Artificial intelligence
- Machine learning
- Deep learning
- generative learning
- Generative adversarial networks
- pre-silicon stage
- gate-level netlist phase
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Υπολογιστές, Υλικό (hardware) και Αρχιτεκτονική
- Αυτοματισμοί και Συστήματα ελέγχου
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Computer science, Hardware and Architecture
- Automation and control systems
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/51720