Print Email Facebook Twitter Semi-Supervised Learning In Side-Channel Attacks Title Semi-Supervised Learning In Side-Channel Attacks Author Slooff, Tom (TU Delft Electrical Engineering, Mathematics and Computer Science; TU Delft Cyber Security) Contributor Picek, S. (mentor) Erkin, Z. (graduation committee) van Gemert, J.C. (graduation committee) Degree granting institution Delft University of Technology Programme Computer Science | Cyber Security Date 2021-10-22 Abstract One of the most potent attacks against cryptographic implementations nowadays is side-channel attacks. Side-channel attacks use unintended leakages in the implementation, for example, electromagnetic radiation, to retrieve the secret key. Over time side-channel attacks have become more powerful, and recently the community has shifted towards using deep learning methods. These powerful side-channel attacks operate in the profiled setting, meaning the attacker has complete control over an identical device as the one they are attacking, which they can use to train the model. In this research, the attacker model is changed to the semi-supervised setting, meaning the attacker is limited in the profiling stage.This thesis proposes a new semi-supervised approach for side-channel attacks, which uses a clustering and a labeling model in a hierarchical fashion. The two common leakage models of side-channel attacks are used to maximize the captured leakage. Furthermore, the newly proposed technique is optimized on several hyperparameters. The approach is compared against two state-of-the-art models in several settings in which five common countermeasures are considered. The results show that the new approach is up to par with the state-of-the-art against desynchronization and clock jitter.Lastly, data augmentation is utilized to improve the performance of the new approach as well as the state-of-the-art. With the usage of data augmentation, the new approach exceeds the performance of the state-of-the-art against all countermeasures except Gaussian noise. Furthermore, the new approach with data augmentation outperforms the state-of-the-art with data augmentation against the most challenging countermeasures of clock jitter, and random delay interrupts. Subject deep learningside-channel attackssemi-supervised learning To reference this document use: http://resolver.tudelft.nl/uuid:5df16274-54ed-45e5-bed7-e5258ce8ad92 Part of collection Student theses Document type master thesis Rights © 2021 Tom Slooff Files PDF thesis_tom_slooff.pdf 10.62 MB Close viewer /islandora/object/uuid:5df16274-54ed-45e5-bed7-e5258ce8ad92/datastream/OBJ/view