Chronic Study on Brainwave Authentication in a Real-Life Setting: An LSTM-Based Bagging Approach

Abstract

With the advent of the digital age, concern about how to secure authorized access to sensitive data is increasing. Besides traditional authentication methods, there is an interest in biometric traits such as fingerprints, the iris, facial characteristics, and, recently, brainwaves, primarily based on electroencephalography (EEG). Current work on EEG-based authentication focuses on acute recordings in laboratory settings using high-end equipment, typically equipped with 64 channels and operating at a high sampling rate. In this work, we validated the feasibility of EEG-based authentication in a real-world, out-of-laboratory setting using a commercial dry-electrode EEG headset and chronic recordings on a population of 15 healthy people. We used an LSTM-based network with bootstrap aggregating (bagging) to decode our recordings in response to a multitask scheme consisting of performed and imagined motor tasks, and showed that it improved the performance of the standard LSTM approach. We achieved an authentication accuracy, false acceptance rate (FAR), and false rejection rate (FRR) of 92.6%, 2.5%, and 5.0% for the performed motor task; 92.5%, 2.6%, and 4.9% for the imagined motor task; and 93.0%, 1.9%, and 5.1% for the combined tasks, respectively. We recommend the proposed method for time- and data-limited scenarios.

Biometrics for Industry 4.0: a survey of recent applications

“…( 2022 ) Authentication Fingerprint Public dataset: FVC-2004 DB3 Information extraction on minutiae Encryption methods EER: 30% Accuracy: 70% Bhowal et al. ( 2022 ) Authentication Signature Public datasets: SVC 2004 MCYT-100 Physical, frequency-based, and statistical feature extraction A two-tier ensemble approach for classification SVC 2004 Accuracy: 98.43% EER: 2.20 GAR@0.01FAR: 94.50% MCYT-100 Accuracy: 97.87% EER: 2.84 GAR@0.01FAR: 92.90% Libert and Van Hulle ( 2021 ) Authentication EEG Private dataset: 15 subjects STFT for feature extraction LSTM-based network with bootstrap aggregating Performed motor task Accuracy: 92.6% FAR: 2.5% FRR: 5% Imagined motor task Accuracy: 92,5% FAR: 2.6% FRR: 4.9% Combined tasks Accuracy: 93% FAR: 1.9% FRR: 5,1% Kaczmarek et al. ( 2018 ) De-authentication Identification Posture Patterns Private dataset: 30 subjects ML algorithms applied to time series of the force on the sensors True positive rate: 91.0% False positive rate: 0,33% False negative rate: 8.68% Multiple sessions True positive rate: 22%, False positive rate: 5.2%, False negative rate: 72.7% Castiglione et al.…”

“…Among them, Libert and Van Hulle ( 2021 ) in their work proposed a brainwave authentication system. In particular, the results obtained in this work showed the potential of using this biometric trait to authenticate the identity of a subject in a real-life context (which could also be the work environment) even with only the use of commercial and not particularly accurate hardware (a commercial EEG headset with a dry electrode and chronic recordings).…”

“… 2021 EEG L L M M L H L WHP-QWL-Security Bacevice and Ducao 2021 ; Ramírez-Moreno et al. 2021 ; Libert and Van Hulle 2021 Galvanic skin response M M L L L H H WHP-QWL Girardi et al. 2021 ; Soto et al.…”

Improving Generalization of CNN-Based Motor-Imagery EEG Decoders via Dynamic Convolutions

“…In fact, these differences are so distinct that previous works have shown that the identification of a specific subject out-of-many is actually feasible (e.g. [31], [32], [33]). Therefore, modern DLbased BCIs tend to fail to generalize well in unseen subjects due to this type of data distribution shift.…”

Feature Selection Assists BLSTM for the Ultrasensitive Detection of Bioflavonoids in Different Biological Matrices Based on the 3D Fluorescence Spectra of Gold Nanoclusters

“…Due to the problem of gradient vanishing and exploding associated with RNN, an advanced architecture of RNN called long-term memory (LSTM) was designed. ,, LSTM has good efficiency in predicting time series because of its unique advantage in solving gradient problems (vanishing and exploding). Also, the term (long-term memory) makes it able to deal with long sequences and predict time series more accurately with more parameters compared to RNN. − LSTM depends on the previous input, but in fact, the outputs also depend on the latter inputs, which increases the number of inputs available to the model. Therefore, a new model with an advanced architecture of LSTM has been designed.…”