Pith. sign in

REVIEW 4 major objections 5 minor 48 references

Contrastive Learning for Continuous Touch-Based Authentication

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read TouchSeqNet, a contrastive learning framework with masked-autoencoder pretraining, reports about 99% accuracy for touch-based user identification on three datasets.

desk verdict The architecture is plausible, but the reported SOTA numbers are unverifiable because the train/test split is never described; as written, the empirical claim doesn't hold. read the letter →

arxiv 2504.17271 v1 pith:DUMNMHWT submitted 2025-04-24 cs.CR

Mengyu Qiao , Yunpeng Zhai , Yang Wang This is my paper · ORCID
classification cs.CR
keywords continuousauthenticationtouchdynamicscontrastivelearningmaskedautoencodersiamesenetworkbehavioralbiometricsself-supervisedtemporalconvolutional
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that continuous touch-based authentication can be cast as a contrastive pair-classification problem: given two touch sequences, decide whether the same person produced both. To make that work, it pre-trains a Temporal Masked Autoencoder on unlabeled touch streams, transfers the encoder into a Siamese Temporal-Attentive Convolutional Network, and trains with a hybrid contrastive and cross-entropy loss. Reported accuracies reach 0.9769 on the new Ffinger dataset, 0.9902 on BioIdent, and 0.9908 on Touchalytics, beating TCN, Gate-Transformer, LSTM, InceptionTime, and TSLANet under the same pairing setup. If the results hold, the practical payoff is a non-disruptive authentication layer that keeps verifying identity throughout a session from natural swipes and taps.

What carries the argument

The load-bearing object is the two-stage TouchSeqNet pipeline. Stage one is a Temporal Masked Autoencoder (TMAE), which slices touch streams into windows, tokenizes them with a Gumbel-Softmax quantizer, masks 40% of windows, and learns by regressing masked representations against a momentum-updated encoder while predicting discrete codewords. Stage two is a Siamese Temporal-Attentive Convolutional Network (TACN), in which the pre-trained encoder feeds dilated causal convolutions, multi-head attention, and a finger-channel attention module; the two branches' pooled embeddings are concatenated and scored by a classification head under a contrastive-plus-cross-entropy loss.

What would settle it

Run TouchSeqNet under a strict leave-one-user-out protocol, where every sample of a held-out user appears only in the test set, and compare accuracy with the reported numbers; if accuracy collapses toward the no-pretraining baseline, the claimed user-level generalization was an artifact of sample-level train/test leakage.

Watch

Extended reading notes

Core claim

The central claim is that touch dynamics carry enough per-user temporal structure that a self-supervised pretraining stage plus contrastive fine-tuning can separate users nearly perfectly, without handcrafted features. TouchSeqNet pre-trains a Temporal Masked Autoencoder to reconstruct masked windows of raw multi-sensor touch sequences, then transfers the learned temporal encoder into a Siamese network whose Temporal-Attentive Convolutional Network combines dilated causal convolutions, multi-head self-attention, and channel attention. The model outputs a same-user probability for each pair of touch samples and achieves 0.9769 accuracy on Ffinger, 0.9902 on BioIdent, and 0.9908 on Touchalytics, outperforming five time-series baselines in a matched experimental setup.

Load-bearing premise

The reported results assume a user-disjoint train/test split, but the paper never describes the split, so if samples from the same user appear in both training and testing the model could memorize identity rather than learn touch style.

Editorial extensions

If this is right

  • A deployed system could verify identity continuously from normal touch interaction, with no extra user effort, because the model scores every pair of observed touch segments.
  • Self-supervised pretraining on unlabeled touch logs could let the system adapt to a new device or a new user's habits without collecting labeled impostor data.
  • The same encoder-plus-Siamese recipe transfers to other behavioral streams, since the paper's conclusion explicitly points to stylus, handwriting, and gesture input as extensions.
  • Near-perfect pair accuracy implies that impostor and genuine touch pairs are strongly separable in the learned embedding space, which is the property an authentication alarm needs.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A strict test would be leave-one-user-out evaluation: the paper never states whether the train/test split is user-disjoint, so the reported accuracies should be re-checked with all samples of each user confined to one side of the split.
  • Balanced pair accuracy can flatter real-world authentication, where impostor pairs are rare; an operational deployment would need a fixed decision threshold and a report of false-acceptance and false-rejection rates at that threshold.
  • The near-perfect scores on public benchmarks may partly reflect that contrastive pair classification is easier than one-class authentication, because negative pairs are randomly drawn from other users; testing on harder impostor distributions (same-gender, same-device, mimicry) would show whether the margin survives.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes TouchSeqNet, a contrastive-learning framework for continuous touch-based authentication. The architecture combines a Temporal Masked Autoencoder (TMAE) pretraining stage, a Siamese Temporal-Attentive Convolutional Network (TACN), and a hybrid contrastive/cross-entropy loss. The authors evaluate on a self-collected Ffinger dataset and on two public benchmarks, BioIdent and Touchalytics, reporting near-ceiling accuracies (0.9769, 0.9902, and 0.9908 respectively) and claiming state-of-the-art performance. They also provide comparisons against five generic time-series baselines and ablate the contribution of the attention and pretraining components.

Significance. If the reported results are valid, the proposed combination of self-supervised masked autoencoding, Siamese contrastive learning, and attention-augmented temporal convolution would be a meaningful step for touch-based continuous authentication. The paper introduces a new dataset, Ffinger, and it is commendable that the authors evaluate multiple baseline models under a consistent pairing framework. However, the current manuscript does not provide enough experimental evidence to support these claims: the train/test protocol is unspecified, pretraining and evaluation may share data, and no code, data, error bars, or per-user statistics are provided. The significance is therefore conditional on a substantially strengthened evaluation.

major comments (4)
  1. [Sections 6.1.1-6.1.2, Table 1] The train/test protocol is not specified. The text defines positive and negative pairs only by same/different user and refers to "held-out test sets" without stating whether the split is by user, by session, or by individual sample. If the split is sample-level, samples from the same user can appear in both training and test sets, allowing the pair classifier to memorize user identity instead of generalizing across users. The near-ceiling results (0.9908 on Touchalytics, 0.9902 on BioIdent) are consistent with this leakage. Please specify the exact split origin, report per-user results, and give the number of users/sessions used for each dataset.
  2. [Section 6.2.1] Pre-training is conducted independently on each dataset and the model is then fine-tuned and evaluated on the same dataset. It is not stated whether the pre-training data are restricted to the training split. If the full dataset, including the test split, is used for masked-autoencoder pre-training, the "transferable representation" claim is confounded and the test numbers in Table 1 cannot be interpreted as generalization measurements. State explicitly which files/segments are used for pre-training, fine-tuning, and testing.
  3. [Table 2 and Section 6.2.3] The comparative evaluation uses generic time-series classifiers (TCN, Gate-Transformer, LSTM, InceptionTime, TSLANet) as baselines and does not compare with prior touch-biometric systems, such as the original Touchalytics evaluation or recent touch-dynamics authentication models. The conclusion that TouchSeqNet outperforms "state-of-the-art methods" is therefore not supported by the evidence. In addition, all numbers are single-run point estimates with no error bars, significance tests, or per-user variance; four-decimal precision is not meaningful without repeated runs or confidence intervals.
  4. [Section 3.1 and general availability] The Ffinger dataset is not released, and no code or pre-trained models are provided. Basic acquisition information is also absent: the number of sessions per participant, device models, gesture length distributions, and IRB/consent details are not reported. Without these, the Ffinger results cannot be reproduced or independently checked, which is a serious limitation for an empirical claim of state-of-the-art performance.
minor comments (5)
  1. [Sections 4.1 and 6.1.3] The slicing window is denoted by sigma in Section 4.1 but by delta in Section 6.1.3; please reconcile the notation.
  2. [Section 5.3] The FingerCA description begins "As illustrated in Figure 1," but FingerCA is shown in Figure 4; please correct the cross-reference.
  3. [Section 4.4] The text mentions auxiliary metrics Hits and NDCG@10, but these are never defined or reported anywhere in the experiments.
  4. [Abstract/Introduction] The phrase "self-supervised pertaining" in the introduction should read "self-supervised pretraining."
  5. [Sections 6.2.1 and 6.2.2] Section 6.2.1 says each model is evaluated on its corresponding validation set, while Section 6.2.2 reports held-out test sets; clarify which split is actually used for the reported numbers.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: TouchSeqNet is an empirical contrastive-learning system whose claimed gains are experimental results, not quantities derived from their own inputs.

full rationale

I examined the derivation chain in Sections 3-6. The claimed contributions are architectural and empirical: TMAE pretext tasks reconstruct masked windows and predict discrete codewords (Eqs. 18-19), TouchSeqNet combines a Siamese TACN with a hybrid contrastive/cross-entropy loss (Eqs. 24-26), and the reported accuracies in Tables 1-3 are measured classification results. None of these equations defines a predicted quantity in terms of the same quantity at training time, and no fitted parameter is renamed as a prediction. The contrastive pair labeling (same-user vs. different-user) defines the task rather than smuggling the target into the input. The self-supervised pretraining is performed on the same datasets used for fine-tuning (Section 6.2.1), which raises a legitimate generalization/leakage concern, but that is an experimental-design risk, not circular reasoning. I found no author self-citation chain that carries the load-bearing argument, and no invocation of a uniqueness theorem or ansatz-by-citation. The SOTA claims stand or fall on the adequacy of the experimental protocol, not on a circular derivation. Score 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central performance claim rests on several unstated experimental choices and hyperparameters. The architecture itself reuses standard modules such as Transformer, TCN, SE attention, and Gumbel-Softmax, so no new physical entities are introduced. The main unacknowledged assumptions are about data preprocessing, the self-supervised pretraining benefit, and the pair-classification evaluation protocol.

free parameters (4)
  • Window size sigma for TMAE slicing = unknown, selected per dataset from {4, 8, 12}
    Controls tokenization granularity; final value per dataset is not reported in Section 6.1.3.
  • TCN kernel size = unknown, selected from {4, 5, 7}
    Chosen on validation per dataset; final value is not reported in Section 6.1.3.
  • Loss weights alpha, beta, lambda1, lambda2 = not specified
    Weights of the four loss terms are introduced in Equations 18, 19, 24, and 26, but no values are given.
  • Contrastive margin m = not specified
    Margin in Equation 24 is never reported.
assumptions (5)
  • domain assumption The 5-channel preprocessed input (time, x, y, pressure, area) with first-order differencing preserves enough identity information for authentication.
    Used as model input in Section 3.2; no feature-selection or ablation evidence is given.
  • domain assumption Self-supervised masked reconstruction on unlabeled touch data produces representations useful for identity classification.
    Assumed in Section 4.3 and relied on by transfer learning in Section 5.1; only an ablation versus no pretraining is shown, with no error bars.
  • domain assumption Binary same-user versus different-user pair classification is a valid proxy for continuous authentication.
    Defines the whole evaluation in Section 6.1.1; no EER or session-aware protocol is reported.
  • standard math Momentum encoder EMA targets provide stable regression targets.
    Standard practice from the MoCo and MAE line of work; invoked in Equation 17.
  • standard math Gumbel-Softmax gives a differentiable approximation of discrete token selection.
    Used in Equations 4 and 5 without proof, but standard in the literature.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Contrastive Learning for Continuous Touch-Based Authentication." pith.science (2026). https://pith.science/paper/DUMNMHWT

@misc{pith2026250417271,
  author       = {Pith},
  title        = {Pith review of: Contrastive Learning for Continuous Touch-Based Authentication},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DUMNMHWT}},
  note         = {Machine review of arXiv:2504.17271}
}
read the original abstract

Smart mobile devices have become indispensable in modern daily life, where sensitive information is frequently processed, stored, and transmitted-posing critical demands for robust security controls. Given that touchscreens are the primary medium for human-device interaction, continuous user authentication based on touch behavior presents a natural and seamless security solution. While existing methods predominantly adopt binary classification under single-modal learning settings, we propose a unified contrastive learning framework for continuous authentication in a non-disruptive manner. Specifically, the proposed method leverages a Temporal Masked Autoencoder to extract temporal patterns from raw multi-sensor data streams, capturing continuous motion and gesture dynamics. The pre-trained TMAE is subsequently integrated into a Siamese Temporal-Attentive Convolutional Network within a contrastive learning paradigm to model both sequential and cross-modal patterns. To further enhance performance, we incorporate multi-head attention and channel attention mechanisms to capture long-range dependencies and optimize inter-channel feature integration. Extensive experiments on public benchmarks and a self-collected dataset demonstrate that our approach outperforms state-of-the-art methods, offering a reliable and effective solution for user authentication on mobile devices.

Figures

Figures reproduced from arXiv: 2504.17271 by the authors.

Figure 1
Figure 1. TMAE Model Architecture In order to capture meaningful temporal dependencies across dif￾ferent time scales. We employ a window-slicing strategy to segment the continuous touch sequence into smaller, fixed-length sub-series [16]. Formally, let the touch sequence be represented as 𝑋 = {𝑥1, 𝑥2, . . . , 𝑥𝑇 } ∈ R 𝑇 ×𝐶, where𝑇 is the length of the sequence and 𝐶 is the number of channels (features such as x and y coordina… view at source ↗
Figure 2
Figure 2. Comparison of touch dynamics for different ges [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. TouchSeqNet 4.3.4 Momentum Encoder Updates. The momentum encoder is updated via an exponential moving average of the primary encoder weights [7]: 𝜃𝑚 ← 𝜇𝜃𝑚 + (1 − 𝜇)𝜃𝑒 (17) Here, 𝜃𝑚 and 𝜃𝑒 denote the parameters of the momentum and primary encoders, respectively. A high smoothing factor 𝜇 (e.g., 0.99) ensures stable target representations for regression. 4.4 Self-supervised Loss The self-supervised objective of our mo… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: The structure of FingerCA module 𝐹 (𝑠) = 𝑘 ∑︁−1 𝑖=0 𝑓 (𝑖) · 𝑋𝑠−𝑑·𝑖 . (21) By setting 𝑑 = 2 𝑙 for the 𝑙-th layer, the receptive field grows rapidly with depth while preserving the input length, which is essential for temporal alignment in authentication tasks. Each resi…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 32 canonical work pages

  1. [1]

    Alejandro Acien, Aythami Morales, John V Monaco, Ruben Vera-Rodriguez, and Julian Fierrez. 2021. TypeNet: Deep learning keystroke biometrics. IEEE Transactions on Biometrics, Behavior, and Identity Science 4, 1 (2021), 57–70

    work page 2021

  2. [2]

    Sara Amini, Vahid Noroozi, Amit Pande, Satyajit Gupte, Philip S Yu, and Chris Kanich. 2018. Deepauth: A framework for continuous user re-authentication in mobile apps. In Proceedings of the 27th ACM International Conference on Informa- tion and Knowledge Management . 2027–2035

    work page 2018

  3. [3]

    Margit Antal, Zsolt Bokor, and László Zsolt Szabó. 2015. Information revealed from scrolling interactions on mobile devices. Pattern Recognition Letters 56 (2015), 7–13

    work page 2015

  4. [4]

    Asrar Bajaber, Mai Fadel, and Lamiaa Elrefaei. 2022. Evaluation of Deep Learning Models for Person Authentication Based on Touch Gesture. Computer Systems Science & Engineering 42, 2 (2022)

    work page 2022

  5. [5]

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. 2020. Language models are few-shot learners. Advances in neural Contrastive Learning for Continuous Touch-Based Authentication Conference acronym ’XX, June 03–05, 2018, Woodstock, NY information p...

    work page 2020

  6. [6]

    Mario Parreño Centeno, Yu Guan, and Aad van Moorsel. 2018. Mobile based con- tinuous authentication using deep features. InProceedings of the 2nd international workshop on embedded and mobile deep learning . 19–24

    work page 2018

  7. [7]

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geoffrey Hinton. 2020. A simple framework for contrastive learning of visual representations. In Interna- tional conference on machine learning . PmLR, 1597–1607

    2020

  8. [8]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. Bert: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers). 4171–4186

    2019

Show all 48 references
  1. [9]

    Emadeldeen Eldele, Mohamed Ragab, Zhenghua Chen, Min Wu, Chee Keong Kwoh, Xiaoli Li, and Cuntai Guan. 2021. Time-series representation learning via temporal and contextual contrasting. arXiv preprint arXiv:2106.14112 (2021)

    2021 arXiv

  2. [10]

    Emadeldeen Eldele, Mohamed Ragab, Zhenghua Chen, Min Wu, and Xiaoli Li

  3. [11]

    Julian Fierrez, Ada Pozo, Marcos Martinez-Diaz, Javier Galbally, and Aythami Morales. 2018. Benchmarking touchscreen biometrics for mobile authentication. IEEE transactions on information forensics and security 13, 11 (2018), 2720–2733

    2018

  4. [12]

    Mario Frank, Ralf Biedert, Eugene Ma, Ivan Martinovic, and Dawn Song. 2012. Touchalytics: On the applicability of touchscreen input as a behavioral biometric for continuous authentication. IEEE transactions on information forensics and security 8, 1 (2012), 136–148

    2012

  5. [13]

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick

  6. [14]

    Pradeep Hewage, Ardhendu Behera, Marcello Trovati, Ella Pereira, Morteza Ghahremani, Francesco Palmieri, and Yonghuai Liu. 2020. Temporal convolutional neural (TCN) network for an effective weather forecasting using time-series data from the local weather station. Soft Computi...

    2020

  7. [15]

    Jie Hu, Li Shen, and Gang Sun. 2018. Squeeze-and-excitation networks. InProceed- ings of the IEEE conference on computer vision and pattern recognition . 7132–7141

    2018

  8. [16]

    Shima Imani and Eamonn Keogh. 2021. Multi-window-finder: domain agnostic window size for time series data. Proceedings of the MileTS 21 (2021)

    2021

  9. [17]

    Hassan Ismail Fawaz, Benjamin Lucas, Germain Forestier, Charlotte Pelletier, Daniel F Schmidt, Jonathan Weber, Geoffrey I Webb, Lhassane Idoumghar, Pierre- Alain Muller, and François Petitjean. 2020. Inceptiontime: Finding alexnet for time series classification. Data Mining an...

    2020

  10. [18]

    Parker Lamb, Alexander Millar, and Ramon Fuentes. 2020. Swipe dynamics as a means of authentication: Results from a bayesian unsupervised approach. In 2020 IEEE International Joint Conference on Biometrics (IJCB) . IEEE, 1–9

    2020

  11. [19]

    Colin Lea, Michael D Flynn, Rene Vidal, Austin Reiter, and Gregory D Hager

  12. [20]

    Seunghan Lee, Taeyoung Park, and Kibok Lee. 2023. Learning to embed time series patches independently. arXiv preprint arXiv:2312.16427 (2023)

    2023 arXiv

  13. [21]

    Seunghan Lee, Taeyoung Park, and Kibok Lee. 2023. Soft contrastive learning for time series. arXiv preprint arXiv:2312.16424 (2023)

    2023

  14. [22]

    Lin Lin, Jinlei Wu, Song Fu, Sihao Zhang, Changsheng Tong, and Lizheng Zu

  15. [23]

    Minghao Liu, Shengqi Ren, Siyuan Ma, Jiahui Jiao, Yizhou Chen, Zhiguang Wang, and Wei Song. 2021. Gated transformer networks for multivariate time series classification. arXiv preprint arXiv:2103.14438 (2021)

    2021 arXiv

  16. [24]

    Mingsheng Long, Jianmin Wang, Yue Cao, Jiaguang Sun, and Philip S Yu. 2016. Deep learning of transferable representation for scalable domain adaptation.IEEE Transactions on Knowledge and Data Engineering 28, 8 (2016), 2027–2040

    2016

  17. [25]

    Sakorn Mekruksavanich and Anuchit Jitpattanakul. 2021. Deep learning ap- proaches for continuous authentication based on activity patterns using mobile sensing. Sensors 21, 22 (2021), 7519

    2021

  18. [26]

    Advanced Engineering Informatics 60 (2024), 102372

    Channel attention & temporal attention based temporal convolutional network: A dual attention framework for remaining useful life prediction of the aircraft engines. Advanced Engineering Informatics 60 (2024), 102372

    2024

  19. [27]

    Natalia Neverova, Christian Wolf, Griffin Lacey, Lex Fridman, Deepak Chandra, Brandon Barbello, and Graham Taylor. 2016. Learning human identity from motion patterns. IEEE Access 4 (2016), 1810–1820

    2016

  20. [28]

    Xuran Pan, Chunjiang Ge, Rui Lu, Shiji Song, Guanfu Chen, Zeyi Huang, and Gao Huang. 2022. On the integration of self-attention and convolution. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition . 815–825

    2022

  21. [29]

    Chao Shen, Yuanxun Li, Yufei Chen, Xiaohong Guan, and Roy A Maxion. 2017. Performance analysis of multi-motion sensor behavior for active smartphone authentication. IEEE Transactions on Information Forensics and Security 13, 1 (2017), 48–62

    2017

  22. [30]

    Asif Iqbal Middya, Sarbani Roy, Saptarshi Mandal, and Rahul Talukdar. 2021. Privacy protected user identification using deep learning for smartphone-based participatory sensing applications. Neural Computing and Applications 33 (2021), 17303–17313

    2021

  23. [31]

    Zhihao Shen, Shun Li, Xi Zhao, and Jianhua Zou. 2022. MMAuth: A continuous authentication framework on smartphones using multiple modalities. IEEE Transactions on Information Forensics and Security 17 (2022), 1450–1465

    2022

  24. [32]

    Onsiri Silasai and Wachana Khowfa. 2020. The study on using biometric authen- tication on mobile device. NU Int. J. Sci 17 (2020), 90–110

    2020

  25. [33]

    Yunpeng Song and Zhongmin Cai. 2022. Integrating handcrafted features with deep representations for smartphone authentication. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies 6, 1 (2022), 1–27

    2022

  26. [34]

    Chao Shen, Yong Zhang, Xiaohong Guan, and Roy A Maxion. 2015. Performance analysis of touch-interaction behavior for active smartphone authentication. IEEE Transactions on Information Forensics and Security 11, 3 (2015), 498–513

    2015

  27. [35]

    Giuseppe Stragapede, Paula Delgado-Santos, Ruben Tolosana, Ruben Vera- Rodriguez, Richard Guest, and Aythami Morales. 2023. Mobile keystroke biomet- rics using transformers. In 2023 IEEE 17th International Conference on Automatic Face and Gesture Recognition (FG) . IEEE, 1–6

    2023

  28. [36]

    Giuseppe Stragapede, Ruben Vera-Rodriguez, Ruben Tolosana, Aythami Morales, Alejandro Acien, and Gaël Le Lan. 2022. Mobile behavioral biometrics for passive authentication. Pattern Recognition Letters 157 (2022), 35–41

    2022

  29. [37]

    Yu Sun, Qiyuan Gao, Xiaofan Du, and Zhao Gu. 2019. Smartphone User Authen- tication Based on Holding Position and Touch-Typing Biometrics. Computers, Materials & Continua 61, 3 (2019)

    2019

  30. [38]

    Yan Song, Shengyao Gao, Yibin Li, Lei Jia, Qiqiang Li, and Fuzhen Pang. 2020. Distributed attention-based temporal convolutional network for remaining useful life prediction. IEEE Internet of Things Journal 8, 12 (2020), 9594–9602

    2020

  31. [39]

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems 30 (2017)

    2017

  32. [40]

    Cong Wang, Yanru Xiao, Xing Gao, Li Li, and Jun Wang. 2021. A framework for behavioral biometric authentication using deep metric learning on mobile devices. IEEE Transactions on Mobile Computing 22, 1 (2021), 19–36

    2021

  33. [41]

    Huanran Wang, Hui He, Chen Song, Hao Tang, Yanwei Sun, Yanchen Qiao, and Weizhe Zhang. 2022. Who Is Using the Phone? Representation-Learning- Based Continuous Authentication on Smartphones. Security and Communication Networks 2022, 1 (2022), 6339407

    2022

  34. [42]

    Ruben Tolosana, Ruben Vera-Rodriguez, Julian Fierrez, and Javier Ortega-Garcia

  35. [43]

    Iqra Zahid, Yue Chang, Tharindu Madusanka, Youcheng Sun, and Riza Theresa Batista-Navarro. 2024. Multi-Loss Fusion: Angular and Contrastive Integra- tion for Machine-Generated Text Detection. In Findings of the Association for Computational Linguistics: EMNLP 2024 . 7189–7202....

    2024

  36. [47]

    Jingyun Xiao, Ran Liu, and Eva L Dyer. 2024. Gaformer: Enhancing timeseries transformers through group-aware embeddings. In The Twelfth International Conference on Learning Representations

    2024

  37. [2017]

    In proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Temporal convolutional networks for action segmentation and detection. In proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 156–165

  38. [2020]

    BioTouchPass2: Touchscreen password biometrics using time-aligned recurrent neural networks.IEEE Transactions on Information Forensics and Security 15 (2020), 2616–2628

    2020

  39. [2021]

    arXiv preprint arXiv:2111.06377 (2021)

    Mae: Masked autoencoders are scalable vision learners. arXiv preprint arXiv:2111.06377 (2021)

    2021 arXiv

  40. [2024]

    arXiv preprint arXiv:2404.08472 (2024)

    Tslanet: Rethinking transformers for time series representation learning. arXiv preprint arXiv:2404.08472 (2024)

    2024 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.