EEG-based identification and cryptographic key generation system using extracted features from transformer-based models

Biometric systems use the unique behavioral or physical characteristics of a user to verify their claimed identity. Due to the high probability of forgery or theft of traditional passwords and keys, there is a decreasing tendency to use them in security systems. By using biometric indicators, it becomes impossible to forge or steal them. Electroencephalogram (EEG) signals meet the basic requirements of biometric indicators, making them suitable for use in authentication and crypto-biometric systems. In this paper, the first step involves extracting features from recorded EEG signals using transformer-based models within an identification system. In the second step, the extracted features are imported into a key generation system. The proposed method maps the features of each user to different segments. The distributions of the segment indexes are then used to generate repeatable keys from EEG features in future sessions. The Transformer-based identification system achieved a mean accuracy of 99.8%, and the key generation system achieved a 0.1% mean Half Total Error Rate (HTER) using five different categories of visual stimulus. The high accuracy of the proposed identification system and the low error rate of the proposed key generation system indicate that features extracted by the Transformers are a good choice for visual stimulus EEG-based biometric systems.