Slide
Centre Interdisciplinaire
de Recherche et d’Innovation
en Cybersécurité et Société
de Recherche et d’Innovation
en Cybersécurité et Société
11.
Houda, Z. A. El; Moudoud, H.
Toward a Secure Zero-Touch Tactile Internet: Challenges and Opportunities Journal Article
In: IEEE Communications Magazine, vol. 64, no. 1, pp. 56–62, 2026, ISSN: 01636804 (ISSN).
Abstract | Links | BibTeX | Tags: Communication service, Communications networks, High availability, High reliability, Low latency, Network security, Real-time application, Remote surgery, Security aspects, Security challenges, Tactile sensors, Ultra-high, virtual reality
@article{elhoudaSecureZeroTouchTactile2026,
title = {Toward a Secure Zero-Touch Tactile Internet: Challenges and Opportunities},
author = {Z. A. El Houda and H. Moudoud},
url = {https://www.scopus.com/pages/publications/105025652094?origin=resultslist},
doi = {10.1109/MCOM.001.2500259},
issn = {01636804 (ISSN)},
year = {2026},
date = {2026-01-01},
journal = {IEEE Communications Magazine},
volume = {64},
number = {1},
pages = {56–62},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Tactile Internet (TI) is a new generation of communication networks, extending beyond 5G/6G capabilities to achieve ultra-low latency, ultra-responsive, ultra-high availability, and ultra-high reliability communication services and thus enabling a new era of real-time applications, such as virtual reality, remote surgery, and autonomous driving. TI is expected to revolutionize many industries, including healthcare, transportation, and manufacturing. Despite recent TI initiatives, security aspects remain largely unexplored. TI presents several security challenges due to the increased complexity of the network and the criticality of the applications it supports. Thus, security aspects must be carefully designed to ensure that the benefits of the TI are not outweighed by the risks. In this article, we will initially identify the leading factors that drive TI, with regard to their applications and related technological trends, with respect to ETSI's Zero-touch Service Management (ZSM) principles. Then, we shed light on the primary security challenges, the current missing building blocks toward secure TI, and propose some potential solutions. Finally, we conclude the article with several recommendations and observations for the roadmap toward secure TI. Ultimately, the focus of this article is to establish a foundation for more in-depth research on TI security. © 1979-2012 IEEE.},
keywords = {Communication service, Communications networks, High availability, High reliability, Low latency, Network security, Real-time application, Remote surgery, Security aspects, Security challenges, Tactile sensors, Ultra-high, virtual reality},
pubstate = {published},
tppubtype = {article}
}
Tactile Internet (TI) is a new generation of communication networks, extending beyond 5G/6G capabilities to achieve ultra-low latency, ultra-responsive, ultra-high availability, and ultra-high reliability communication services and thus enabling a new era of real-time applications, such as virtual reality, remote surgery, and autonomous driving. TI is expected to revolutionize many industries, including healthcare, transportation, and manufacturing. Despite recent TI initiatives, security aspects remain largely unexplored. TI presents several security challenges due to the increased complexity of the network and the criticality of the applications it supports. Thus, security aspects must be carefully designed to ensure that the benefits of the TI are not outweighed by the risks. In this article, we will initially identify the leading factors that drive TI, with regard to their applications and related technological trends, with respect to ETSI's Zero-touch Service Management (ZSM) principles. Then, we shed light on the primary security challenges, the current missing building blocks toward secure TI, and propose some potential solutions. Finally, we conclude the article with several recommendations and observations for the roadmap toward secure TI. Ultimately, the focus of this article is to establish a foundation for more in-depth research on TI security. © 1979-2012 IEEE.
12.
Sullivan, A.; Zakas, F.; Sklenarik, S.; Bouchard, S.; Potenza, M.; Astur, R.
A Brief Virtual Reality Intervention Reduces Gambling Symptoms and Gambling Engagement: A Pilot Study Journal Article
In: Journal of Gambling Studies, 2026, ISSN: 10505350 (ISSN).
@article{sullivanBriefVirtualReality2026,
title = {A Brief Virtual Reality Intervention Reduces Gambling Symptoms and Gambling Engagement: A Pilot Study},
author = {A. Sullivan and F. Zakas and S. Sklenarik and S. Bouchard and M. Potenza and R. Astur},
url = {https://www.scopus.com/pages/publications/105046108791?origin=resultslist},
doi = {10.1007/s10899-026-10538-8},
issn = {10505350 (ISSN)},
year = {2026},
date = {2026-01-01},
journal = {Journal of Gambling Studies},
publisher = {Springer},
keywords = {},
pubstate = {published},
tppubtype = {article}
}
13.
Abdollahzadeh, S.; Allili, M. S.; Boulmerka, A.; Lapointe, J. -F.
A Vision-Based Framework for Safe Landing Zone Mapping of UAVs in Dynamic Environments Journal Article
In: IEEE Open Journal of the Computer Society, vol. 7, pp. 492–503, 2026, ISSN: 26441268 (ISSN).
Abstract | Links | BibTeX | Tags: Aerial vehicle, Air navigation, Aircraft detection, Aircraft landing, Antennas, automatic UAV navigation, Computer vision, Dynamic environments, Forecasting, Homographies, Landing zones, Learning systems, Motion tracking, Object detection, Object recognition, Object Tracking, object trajectory prediction, Robotics, Safe landing, Safe landing zone, safe landing zones (SLZ), Semantic segmentation, Semantics, Trajectories, Trajectory forecasting, Uncrewed aerial vehicles (UAVs), Unmanned aerial vehicle, Unmanned aerial vehicles (UAV)
@article{abdollahzadehVisionBasedFrameworkSafe2026,
title = {A Vision-Based Framework for Safe Landing Zone Mapping of UAVs in Dynamic Environments},
author = {S. Abdollahzadeh and M. S. Allili and A. Boulmerka and J. -F. Lapointe},
url = {https://www.scopus.com/pages/publications/105029942397?origin=resultslist},
doi = {10.1109/OJCS.2026.3663268},
issn = {26441268 (ISSN)},
year = {2026},
date = {2026-01-01},
journal = {IEEE Open Journal of the Computer Society},
volume = {7},
pages = {492–503},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Identification safe landing zones (SLZ) for Uncrewed Aerial Vehicles (UAVs) is important to ensure reliable and safe navigation, especially when they are operated in complex and safety-critical environments. However, this is a challenging task due to obstacles and UAV motion. This paper proposes a vision-based framework that maps SLZs in dynamic scenes by integrating several functionalities for analyzing visually static and dynamic aspects of a scene. Static analysis is achieved through context-aware segmentation which divides the image into thematic classes enabling to identify suitable landing surfaces (e.g., roads, grass). For dynamic content analysis, we combine object detection, tracking, and trajectory prediction to determine object occupancy and identify regions free of obstacles. Trajectory prediction is performed through a novel encoder–decoder architecture taking past object positions to predict the most likely future locations. To ensure stable and robust trajectory prediction, we introduce an optimized homography computation using multi-scale image analysis and cumulative updates to compensate UAV motion. We tested our framework on different operational scenarios, including urban and natural scenes with moving objects like vehicles and pedestrians. Obtained results demonstrate its strong performance, and its significant potential for enabling autonomous and safe UAV navigation. © 2020 IEEE.},
keywords = {Aerial vehicle, Air navigation, Aircraft detection, Aircraft landing, Antennas, automatic UAV navigation, Computer vision, Dynamic environments, Forecasting, Homographies, Landing zones, Learning systems, Motion tracking, Object detection, Object recognition, Object Tracking, object trajectory prediction, Robotics, Safe landing, Safe landing zone, safe landing zones (SLZ), Semantic segmentation, Semantics, Trajectories, Trajectory forecasting, Uncrewed aerial vehicles (UAVs), Unmanned aerial vehicle, Unmanned aerial vehicles (UAV)},
pubstate = {published},
tppubtype = {article}
}
Identification safe landing zones (SLZ) for Uncrewed Aerial Vehicles (UAVs) is important to ensure reliable and safe navigation, especially when they are operated in complex and safety-critical environments. However, this is a challenging task due to obstacles and UAV motion. This paper proposes a vision-based framework that maps SLZs in dynamic scenes by integrating several functionalities for analyzing visually static and dynamic aspects of a scene. Static analysis is achieved through context-aware segmentation which divides the image into thematic classes enabling to identify suitable landing surfaces (e.g., roads, grass). For dynamic content analysis, we combine object detection, tracking, and trajectory prediction to determine object occupancy and identify regions free of obstacles. Trajectory prediction is performed through a novel encoder–decoder architecture taking past object positions to predict the most likely future locations. To ensure stable and robust trajectory prediction, we introduce an optimized homography computation using multi-scale image analysis and cumulative updates to compensate UAV motion. We tested our framework on different operational scenarios, including urban and natural scenes with moving objects like vehicles and pedestrians. Obtained results demonstrate its strong performance, and its significant potential for enabling autonomous and safe UAV navigation. © 2020 IEEE.
14.
Abdollahzadeh, S.; Allili, M. S.; Boulmerka, A.; Lapointe, J. -F.
Visual Safety Mapping for UAV Landings Using Ordinal Regression Networks Journal Article
In: IEEE Transactions on Artificial Intelligence, vol. 7, no. 6, pp. 3146–3159, 2026, ISSN: 26914581 (ISSN).
Abstract | Links | BibTeX | Tags: Aerial vehicle, Air navigation, Aircraft landing, Antennas, Automatic uncrewed aerial vehicle (UAV) navigation, Automatic unmanned aerial vehicle navigation, Autonomous navigation, Behavioral research, Computer vision, decision making, Deep learning, deep ordinal regression, Emergency services, Landing zones, Learning systems, Ordinal regression, regression analysis, Safe landing, Safe landing zone, safe landing zones (SLZs), Semantic segmentation, Semantics, Textures, Three dimensional computer graphics, Unmanned aerial vehicles (UAV), Vehicle navigation, Visual servoing
@article{abdollahzadehVisualSafetyMapping2026,
title = {Visual Safety Mapping for UAV Landings Using Ordinal Regression Networks},
author = {S. Abdollahzadeh and M. S. Allili and A. Boulmerka and J. -F. Lapointe},
url = {https://www.scopus.com/pages/publications/105023324811?origin=resultslist},
doi = {10.1109/TAI.2025.3635093},
issn = {26914581 (ISSN)},
year = {2026},
date = {2026-01-01},
journal = {IEEE Transactions on Artificial Intelligence},
volume = {7},
number = {6},
pages = {3146–3159},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {As uncrewed aerial vehicles (UAVs) see growing use in civilian applications, reliably identifying safe landing zones (SLZs) in varied environments is essential for autonomous navigation and emergency response. Passive vision sensors offer a low cost, lightweight solution for real-time terrain analysis and 3D scene reconstruction, making them ideal for onboard systems. We introduce ordinal regression safe landing zones netwrok (OR-SLZNet), an original deep learning model based on ordinal regression to predict SLZs from UAV imagery. Unlike prior approaches, OR-SLZNet produces dense, multilevel safety maps by jointly leveraging photometric (e.g., color and texture) and geometric cues (e.g., flatness, slope, and depth), assigning each pixel an ordinal safety score that reflects landing suitability. With real-time inference (textbackslashsim0.02 s/frame), the model supports onboard deployment and rapid decision-making in time-critical situations. Extensive experiments on five diverse datasets demonstrate OR-SLZNet effectiveness and strong generalization across a wide range of structural complexities. © 2020 IEEE.},
keywords = {Aerial vehicle, Air navigation, Aircraft landing, Antennas, Automatic uncrewed aerial vehicle (UAV) navigation, Automatic unmanned aerial vehicle navigation, Autonomous navigation, Behavioral research, Computer vision, decision making, Deep learning, deep ordinal regression, Emergency services, Landing zones, Learning systems, Ordinal regression, regression analysis, Safe landing, Safe landing zone, safe landing zones (SLZs), Semantic segmentation, Semantics, Textures, Three dimensional computer graphics, Unmanned aerial vehicles (UAV), Vehicle navigation, Visual servoing},
pubstate = {published},
tppubtype = {article}
}
As uncrewed aerial vehicles (UAVs) see growing use in civilian applications, reliably identifying safe landing zones (SLZs) in varied environments is essential for autonomous navigation and emergency response. Passive vision sensors offer a low cost, lightweight solution for real-time terrain analysis and 3D scene reconstruction, making them ideal for onboard systems. We introduce ordinal regression safe landing zones netwrok (OR-SLZNet), an original deep learning model based on ordinal regression to predict SLZs from UAV imagery. Unlike prior approaches, OR-SLZNet produces dense, multilevel safety maps by jointly leveraging photometric (e.g., color and texture) and geometric cues (e.g., flatness, slope, and depth), assigning each pixel an ordinal safety score that reflects landing suitability. With real-time inference (textbackslashsim0.02 s/frame), the model supports onboard deployment and rapid decision-making in time-critical situations. Extensive experiments on five diverse datasets demonstrate OR-SLZNet effectiveness and strong generalization across a wide range of structural complexities. © 2020 IEEE.
15.
Nouboukpo, A.; Allili, M. S.
Semi-supervised flow-augmented Gaussian mixture VAEs with weighted contrastive learning for out-of-distribution detection Journal Article
In: Knowledge-Based Systems, vol. 347, 2026, ISSN: 09507051 (ISSN).
Abstract | Links | BibTeX | Tags: Auto encoders, Benchmarking, Classification (of information), Clusterings, Contrastive Learning, Flow based, Flow-based GMM clustering, Gaussian distribution, Gaussian-mixtures, Learning algorithms, Learning systems, Out-of-distribution detection, Semantics, Semi-supervised, Semi-supervised learning, uncertainty, Variational autoencoder, Variational autoencoders (VAEs)
@article{nouboukpoSemisupervisedFlowaugmentedGaussian2026,
title = {Semi-supervised flow-augmented Gaussian mixture VAEs with weighted contrastive learning for out-of-distribution detection},
author = {A. Nouboukpo and M. S. Allili},
url = {https://www.scopus.com/pages/publications/105040505017?origin=resultslist},
doi = {10.1016/j.knosys.2026.116312},
issn = {09507051 (ISSN)},
year = {2026},
date = {2026-01-01},
journal = {Knowledge-Based Systems},
volume = {347},
publisher = {Elsevier B.V.},
abstract = {Out-of-distribution (OOD) detection is essential for the safe deployment of machine learning systems, particularly in high-stakes domains where identifying inputs that deviate from the in-distribution (ID) is critical. However, in real-world scenarios, true OOD data are typically unavailable at training time, and only auxiliary datasets acting as imperfect proxies can be used. This makes fully supervised approaches impractical and potentially biased toward specific anomaly types. To address this limitation, we propose FLoW-ssGMVAE, a semi-supervised generative–discriminative framework that leverages limited proxy anomaly supervision together with abundant unlabeled data for robust OOD detection. Our model integrates a flow-augmented Gaussian Mixture VAE, enabling flexible latent modeling that captures nonlinear and multimodal intra-class distributions. Unlike standard VAEs with unimodal Gaussian priors, it uses class-conditional normalizing flows to better represent complex ID and OOD variability. To further refine the latent space and strengthen ID/OOD separation, our model integrates a weighted contrastive learning objective guided by sample-wise attention scores derived from likelihood uncertainty. This mechanism emphasizes ambiguous or hard-to-classify instances, reinforcing semantic boundaries and reducing false detections. By combining expressive generative modeling with uncertainty-aware discriminative training, our method constructs a structured and interpretable latent space that reliably identifies OOD samples under limited supervision. Experiments on standard benchmarks validate the effectiveness and scalability of our approach, demonstrating state-of-the-art performance with reasonable computational cost. Our implementation will be available at: FLoW-ssGMVAE. © 2026},
keywords = {Auto encoders, Benchmarking, Classification (of information), Clusterings, Contrastive Learning, Flow based, Flow-based GMM clustering, Gaussian distribution, Gaussian-mixtures, Learning algorithms, Learning systems, Out-of-distribution detection, Semantics, Semi-supervised, Semi-supervised learning, uncertainty, Variational autoencoder, Variational autoencoders (VAEs)},
pubstate = {published},
tppubtype = {article}
}
Out-of-distribution (OOD) detection is essential for the safe deployment of machine learning systems, particularly in high-stakes domains where identifying inputs that deviate from the in-distribution (ID) is critical. However, in real-world scenarios, true OOD data are typically unavailable at training time, and only auxiliary datasets acting as imperfect proxies can be used. This makes fully supervised approaches impractical and potentially biased toward specific anomaly types. To address this limitation, we propose FLoW-ssGMVAE, a semi-supervised generative–discriminative framework that leverages limited proxy anomaly supervision together with abundant unlabeled data for robust OOD detection. Our model integrates a flow-augmented Gaussian Mixture VAE, enabling flexible latent modeling that captures nonlinear and multimodal intra-class distributions. Unlike standard VAEs with unimodal Gaussian priors, it uses class-conditional normalizing flows to better represent complex ID and OOD variability. To further refine the latent space and strengthen ID/OOD separation, our model integrates a weighted contrastive learning objective guided by sample-wise attention scores derived from likelihood uncertainty. This mechanism emphasizes ambiguous or hard-to-classify instances, reinforcing semantic boundaries and reducing false detections. By combining expressive generative modeling with uncertainty-aware discriminative training, our method constructs a structured and interpretable latent space that reliably identifies OOD samples under limited supervision. Experiments on standard benchmarks validate the effectiveness and scalability of our approach, demonstrating state-of-the-art performance with reasonable computational cost. Our implementation will be available at: FLoW-ssGMVAE. © 2026
16.
Mehrban, A.; Moudoud, H.; Brik, B.; Khoukhi, L.; Houda, Z. A. El
Game-Theoretic Security Orchestration for Cross-RIC Policy Conflicts in O-RAN Proceedings Article
In: IEEE Int Conf Commun, Institute of Electrical and Electronics Engineers Inc., 2026, ISBN: 15503607 (ISSN); 979-831954209-0 (ISBN), (Journal Abbreviation: IEEE Int Conf Commun).
Abstract | Links | BibTeX | Tags: Computation theory, Computer system firewalls, Cryptography, Game theory, Game-theoretic, Multiple vendors, Network architecture, Network security, Networks security, Open radio access network, Open RAN, Policy conflict, Radio access networks, Security orchestration, Service management, Service orchestration, SMO, Xapp/rapp conflict, xApp/rApp conflicts
@inproceedings{mehrbanGameTheoreticSecurityOrchestration2026,
title = {Game-Theoretic Security Orchestration for Cross-RIC Policy Conflicts in O-RAN},
author = {A. Mehrban and H. Moudoud and B. Brik and L. Khoukhi and Z. A. El Houda},
url = {https://www.scopus.com/pages/publications/105045343135?origin=resultslist},
doi = {10.1109/ICC59461.2026.11587926},
isbn = {15503607 (ISSN); 979-831954209-0 (ISBN)},
year = {2026},
date = {2026-01-01},
booktitle = {IEEE Int Conf Commun},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {In Open Radio Access Network (Open RAN), disaggregated network components can interact through specialized xApps or rApps from multiple vendors. Despite the flexibility of operations and the dynamism of functions enabled by the open architecture, this openness can cause conflicts in security policy enforcement (e.g., encryption levels, key rotation, firewall restrictiveness), formally defined as thresholded divergences on shared or interdependent security parameters, where one app wants to set a stricter rule while another wants a more lenient policy simultaneously. The current solutions of "first-come"or "last-write"make the change one-sided for the benefit of a single app, while other apps may be harmed by the result, and at a network level, security and stability can be compromised. In this paper, we propose a novel Game-Theoretic Security Orchestrator (GTSO) that operates at the Service Management and Orchestration (SMO) layer, with cross-RIC scope, and is responsible for resolving parameter conflicts across xApps and rApps. GTSO treats xApps/rApps as strategic players and negotiates convergent solutions using game-theoretic models such as non-cooperative Nash games, cooperative bargaining, and leader-follower (Stackelberg) that guide competing control applications toward stable equilibrium points, while exchanging strategy summaries over secure channels to protect sensitive details and enhance overall network security. © 2026 IEEE.},
note = {Journal Abbreviation: IEEE Int Conf Commun},
keywords = {Computation theory, Computer system firewalls, Cryptography, Game theory, Game-theoretic, Multiple vendors, Network architecture, Network security, Networks security, Open radio access network, Open RAN, Policy conflict, Radio access networks, Security orchestration, Service management, Service orchestration, SMO, Xapp/rapp conflict, xApp/rApp conflicts},
pubstate = {published},
tppubtype = {inproceedings}
}
In Open Radio Access Network (Open RAN), disaggregated network components can interact through specialized xApps or rApps from multiple vendors. Despite the flexibility of operations and the dynamism of functions enabled by the open architecture, this openness can cause conflicts in security policy enforcement (e.g., encryption levels, key rotation, firewall restrictiveness), formally defined as thresholded divergences on shared or interdependent security parameters, where one app wants to set a stricter rule while another wants a more lenient policy simultaneously. The current solutions of "first-come"or "last-write"make the change one-sided for the benefit of a single app, while other apps may be harmed by the result, and at a network level, security and stability can be compromised. In this paper, we propose a novel Game-Theoretic Security Orchestrator (GTSO) that operates at the Service Management and Orchestration (SMO) layer, with cross-RIC scope, and is responsible for resolving parameter conflicts across xApps and rApps. GTSO treats xApps/rApps as strategic players and negotiates convergent solutions using game-theoretic models such as non-cooperative Nash games, cooperative bargaining, and leader-follower (Stackelberg) that guide competing control applications toward stable equilibrium points, while exchanging strategy summaries over secure channels to protect sensitive details and enhance overall network security. © 2026 IEEE.
17.
Pimentel, A.; Zhu, Y.; Falk, T. H.
Partial Audio Deepfake Detection: Are We Really Detecting Synthetic Media or Just Dataset Content Biases? Proceedings Article
In: IEEE Conf. Artif. Intell., CAI, pp. 1846–1851, Institute of Electrical and Electronics Engineers Inc., 2026, ISBN: 979-833156039-3 (ISBN), (Journal Abbreviation: IEEE Conf. Artif. Intell., CAI).
Abstract | Links | BibTeX | Tags: Audio signal processing, Computational linguistics, Condition, Detection models, Information integrity, Language model, Large datasets, Learning models, Optimistics, Performance, Self-supervised learning, Speech communication, Speech models, Speech recognition, Speech signals, Synthetic media, Transcription
@inproceedings{pimentelPartialAudioDeepfake2026,
title = {Partial Audio Deepfake Detection: Are We Really Detecting Synthetic Media or Just Dataset Content Biases?},
author = {A. Pimentel and Y. Zhu and T. H. Falk},
url = {https://www.scopus.com/pages/publications/105042045476?origin=resultslist},
doi = {10.1109/CAI68641.2026.11536622},
isbn = {979-833156039-3 (ISBN)},
year = {2026},
date = {2026-01-01},
booktitle = {IEEE Conf. Artif. Intell., CAI},
pages = {1846–1851},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {In recent years, the generation of highly realistic audio deepfakes has raised significant concerns regarding privacy and information integrity. While most research has focused on fully bonafide or spoofed speech, partial deep-fakes, where only segments of an utterance are manipulated, remain less explored. Given the nature of the task, existing datasets are relying on large language models to manipulate bonafide speech signals into partial deepfakes by altering, deleting, or replacing segments with synthetic content. These manipulations may alter the semantics and sentiment of the generated content, creating biases that can be captured by deepfake detection models, leading to overly optimistic performance and poor generalizability. In this work, we investigate the presence of such biases in two popular partial deepfake datasets, namely AV-Deepfake1M and PartialEdit. We explore four self-supervised learning models, two relying only on text transcriptions (RoBERTa and RoBERTa-sentiment) and two relying on universal speech representations (WavLM Large and Wav2vec2-XLSR). Our experiments show that, while speech models achieve higher in-domain performance, they do not generalize to out-of-domain conditions. In turn, models trained on only transcribed speech can effectively distinguish manipulated content, achieving up to 97.8% AUC in-domain and nearly 70% AUC out-of-domain. These results suggest that linguistic patterns and dataset confounds may indeed be biasing partial deepfake detection models, leading to poor generalizability. © 2026 IEEE.},
note = {Journal Abbreviation: IEEE Conf. Artif. Intell., CAI},
keywords = {Audio signal processing, Computational linguistics, Condition, Detection models, Information integrity, Language model, Large datasets, Learning models, Optimistics, Performance, Self-supervised learning, Speech communication, Speech models, Speech recognition, Speech signals, Synthetic media, Transcription},
pubstate = {published},
tppubtype = {inproceedings}
}
In recent years, the generation of highly realistic audio deepfakes has raised significant concerns regarding privacy and information integrity. While most research has focused on fully bonafide or spoofed speech, partial deep-fakes, where only segments of an utterance are manipulated, remain less explored. Given the nature of the task, existing datasets are relying on large language models to manipulate bonafide speech signals into partial deepfakes by altering, deleting, or replacing segments with synthetic content. These manipulations may alter the semantics and sentiment of the generated content, creating biases that can be captured by deepfake detection models, leading to overly optimistic performance and poor generalizability. In this work, we investigate the presence of such biases in two popular partial deepfake datasets, namely AV-Deepfake1M and PartialEdit. We explore four self-supervised learning models, two relying only on text transcriptions (RoBERTa and RoBERTa-sentiment) and two relying on universal speech representations (WavLM Large and Wav2vec2-XLSR). Our experiments show that, while speech models achieve higher in-domain performance, they do not generalize to out-of-domain conditions. In turn, models trained on only transcribed speech can effectively distinguish manipulated content, achieving up to 97.8% AUC in-domain and nearly 70% AUC out-of-domain. These results suggest that linguistic patterns and dataset confounds may indeed be biasing partial deepfake detection models, leading to poor generalizability. © 2026 IEEE.
18.
Moradi, A.; Falk, T. H.
Benchmarking Foundation Models for Cross-Domain Speaker Profiling Proceedings Article
In: IEEE Conf. Artif. Intell., CAI, pp. 78–84, Institute of Electrical and Electronics Engineers Inc., 2026, ISBN: 979-833156039-3 (ISBN), (Journal Abbreviation: IEEE Conf. Artif. Intell., CAI).
Abstract | Links | BibTeX | Tags: Benchmarking, Continuous speech recognition, Cross-domain, Forecasting, Formant frequency, Foundation models, Learning systems, Linguistics, Multi-attributes, Multi-task learning, Paralinguistic, Performance, Speaker identification, Speaker verification, Speech communication, State of the art, Verification task
@inproceedings{moradiBenchmarkingFoundationModels2026,
title = {Benchmarking Foundation Models for Cross-Domain Speaker Profiling},
author = {A. Moradi and T. H. Falk},
url = {https://www.scopus.com/pages/publications/105042046048?origin=resultslist},
doi = {10.1109/CAI68641.2026.11536565},
isbn = {979-833156039-3 (ISBN)},
year = {2026},
date = {2026-01-01},
booktitle = {IEEE Conf. Artif. Intell., CAI},
pages = {78–84},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Speech conveys both linguistic and paralinguistic content. While pre-trained speech foundation models have been widely explored for linguistic tasks, such as speech recognition, and for speaker identification and verification tasks, very limited work has been done to test their usefulness for multi-attribute speaker profiling, i.e., simultaneous prediction of biological sex, age, and height from speech. In this paper, we aim to benchmark the performance of four state-of-the-art self-supervised foundation models, namely, WavLM, Wav2vec2, HuBERT, and XLSR-53, under both within- and cross-domain conditions. Each model is employed as a frozen pre-trained feature extractor, with lightweight task-specific heads trained jointly in a multi-task learning framework, and their feature extraction latency is empirically analyzed to assess practical deployment considerations. Experiments on two datasets show that WavLM achieves consistently strong within- and cross-domain performance for biological sex prediction, while Wav2vec2 and XLSR-53 exhibit more consistent performance for the age and height regression tasks in cross-domain settings. Interpretability analysis based on correlations with key acoustic features show (1) WavLM internal representations correlating highly with pitch and formant frequencies, corroborating the improved performance on biological sex prediction, and (2) XLSR-53 correlating highly with the second formant frequency, corroborating the results obtained for age and height. Overall, our analysis shows that pre-trained speech foundation models could serve as useful tools for cross-domain speaker profiling tasks. While no model stood out as a clear winner across all tested physical traits, future work could explore the use of ensemble methods for improved generalizability. © 2026 IEEE.},
note = {Journal Abbreviation: IEEE Conf. Artif. Intell., CAI},
keywords = {Benchmarking, Continuous speech recognition, Cross-domain, Forecasting, Formant frequency, Foundation models, Learning systems, Linguistics, Multi-attributes, Multi-task learning, Paralinguistic, Performance, Speaker identification, Speaker verification, Speech communication, State of the art, Verification task},
pubstate = {published},
tppubtype = {inproceedings}
}
Speech conveys both linguistic and paralinguistic content. While pre-trained speech foundation models have been widely explored for linguistic tasks, such as speech recognition, and for speaker identification and verification tasks, very limited work has been done to test their usefulness for multi-attribute speaker profiling, i.e., simultaneous prediction of biological sex, age, and height from speech. In this paper, we aim to benchmark the performance of four state-of-the-art self-supervised foundation models, namely, WavLM, Wav2vec2, HuBERT, and XLSR-53, under both within- and cross-domain conditions. Each model is employed as a frozen pre-trained feature extractor, with lightweight task-specific heads trained jointly in a multi-task learning framework, and their feature extraction latency is empirically analyzed to assess practical deployment considerations. Experiments on two datasets show that WavLM achieves consistently strong within- and cross-domain performance for biological sex prediction, while Wav2vec2 and XLSR-53 exhibit more consistent performance for the age and height regression tasks in cross-domain settings. Interpretability analysis based on correlations with key acoustic features show (1) WavLM internal representations correlating highly with pitch and formant frequencies, corroborating the improved performance on biological sex prediction, and (2) XLSR-53 correlating highly with the second formant frequency, corroborating the results obtained for age and height. Overall, our analysis shows that pre-trained speech foundation models could serve as useful tools for cross-domain speaker profiling tasks. While no model stood out as a clear winner across all tested physical traits, future work could explore the use of ensemble methods for improved generalizability. © 2026 IEEE.
19.
Kadi, A.; Moudoud, H.; Khoukhi, L.; Houda, Z. A. El
Quantum-Enhanced LSTM for Sequential Network Flow Analysis: A Hybrid Approach to DDoS Detection Proceedings Article
In: Proc. - Int. Conf. Quantum Commun., Netw., Comput., QCNC, pp. 830–834, Institute of Electrical and Electronics Engineers Inc., 2026, ISBN: 979-833156110-9 (ISBN), (Journal Abbreviation: Proc. - Int. Conf. Quantum Commun., Netw., Comput., QCNC).
Abstract | Links | BibTeX | Tags: Denialof- service attacks, Distributed computer systems, Distributed denial of service, Distributed denial-of-service, Distributed Denial-of-Service (DDoS), Hybrid approach, Intrusion Detection, Learning systems, Logic gates, Long short-term memory, Machine-learning, Memory architecture, Network architecture, Network flow analysis, Network security, QLSTM, Quantum entanglement, Quantum machine learning, Quantum Machine Learning(QML), Quantum machines, Qubits, short term memory
@inproceedings{kadiQuantumEnhancedLSTMSequential2026,
title = {Quantum-Enhanced LSTM for Sequential Network Flow Analysis: A Hybrid Approach to DDoS Detection},
author = {A. Kadi and H. Moudoud and L. Khoukhi and Z. A. El Houda},
url = {https://www.scopus.com/pages/publications/105040813271?origin=resultslist},
doi = {10.1109/QCNC69040.2026.00136},
isbn = {979-833156110-9 (ISBN)},
year = {2026},
date = {2026-01-01},
booktitle = {Proc. - Int. Conf. Quantum Commun., Netw., Comput., QCNC},
pages = {830–834},
publisher = {Institute of Electrical and Electronics Engineers Inc.},
abstract = {Distributed Denial-of-Service (DDoS) attacks continue to grow at a rapid rate, making timely and reliable detection a challenge for intrusion detection systems (IDSs). Classical deep models such as Long Short-Term Memory (LSTM) can capture temporal dependencies; however, they still struggle with highly nonlinear and noisy flow dynamics, especially when attack patterns shift. To address this, we propose QLSTM-E, an enhanced hybrid quantum-classical architecture for network intrusion detection that uses Variational Quantum Circuits (VQCs) and LSTM modeling to learn complex temporal dependencies in network traffic. QLSTM-E exploits quantum characteristics, including superposition and entanglement, to improve the representation of nonlinear patterns in sequential flow features. To balance expressivity and circuit cost on near-term devices, we adopt an angle-based encoding strategy and a star-topology entangling layout that reduces two-qubit gate overhead while preserving effective quantum correlations. We implement QLSTM-E using PennyLane and Qiskit and evaluate it on the CIC-DDoS2019 dataset under a fully simulated setting, requiring no access to quantum hardware. Experimental results demonstrate strong detection effectiveness, achieving 99.7% accuracy and F1-score, and show strong robustness under depolarizing noise up to p= 0.1. © 2026 IEEE.},
note = {Journal Abbreviation: Proc. - Int. Conf. Quantum Commun., Netw., Comput., QCNC},
keywords = {Denialof- service attacks, Distributed computer systems, Distributed denial of service, Distributed denial-of-service, Distributed Denial-of-Service (DDoS), Hybrid approach, Intrusion Detection, Learning systems, Logic gates, Long short-term memory, Machine-learning, Memory architecture, Network architecture, Network flow analysis, Network security, QLSTM, Quantum entanglement, Quantum machine learning, Quantum Machine Learning(QML), Quantum machines, Qubits, short term memory},
pubstate = {published},
tppubtype = {inproceedings}
}
Distributed Denial-of-Service (DDoS) attacks continue to grow at a rapid rate, making timely and reliable detection a challenge for intrusion detection systems (IDSs). Classical deep models such as Long Short-Term Memory (LSTM) can capture temporal dependencies; however, they still struggle with highly nonlinear and noisy flow dynamics, especially when attack patterns shift. To address this, we propose QLSTM-E, an enhanced hybrid quantum-classical architecture for network intrusion detection that uses Variational Quantum Circuits (VQCs) and LSTM modeling to learn complex temporal dependencies in network traffic. QLSTM-E exploits quantum characteristics, including superposition and entanglement, to improve the representation of nonlinear patterns in sequential flow features. To balance expressivity and circuit cost on near-term devices, we adopt an angle-based encoding strategy and a star-topology entangling layout that reduces two-qubit gate overhead while preserving effective quantum correlations. We implement QLSTM-E using PennyLane and Qiskit and evaluate it on the CIC-DDoS2019 dataset under a fully simulated setting, requiring no access to quantum hardware. Experimental results demonstrate strong detection effectiveness, achieving 99.7% accuracy and F1-score, and show strong robustness under depolarizing noise up to p= 0.1. © 2026 IEEE.
20.
Delangle, M.; Moïse-Richard, A.; Leclercq, A. -L.; Labbé, D.; Bouchard, S.; Andrews, S.; Ménard, L.
Speaking face-to-face with a virtual avatar to reduce anxiety in students who stutter: Tool development and pilot study results Journal Article
In: Journal of Fluency Disorders, vol. 87, 2026, ISSN: 0094730X (ISSN).
Abstract | Links | BibTeX | Tags: Adolescent, adult, Anxiety, Arousal, article, Avatar, avoidance behavior, Canada, Child, clinical article, cognitive behavioral therapy, complication, controlled study, distance learning, ecological validity, electrocardiogram, electrodermal response, exposure, female, Heart Rate, heart rate variability, human, Humans, Liebowitz Social Anxiety Scale, Likert scale, male, nonverbal communication, physiological stress, Pilot Projects, pilot study, procedures, psychology, questionnaire, randomized controlled trial, role playing, School-age children, Self Concept, Self Efficacy, self report, Signal processing, skin conductance, social anxiety, speech, student, Students, Stuttering, Therapy, treatment outcome, virtual reality, virtual reality exposure therapy
@article{delangleSpeakingFacetofaceVirtual2026,
title = {Speaking face-to-face with a virtual avatar to reduce anxiety in students who stutter: Tool development and pilot study results},
author = {M. Delangle and A. Moïse-Richard and A. -L. Leclercq and D. Labbé and S. Bouchard and S. Andrews and L. Ménard},
url = {https://www.scopus.com/pages/publications/105028302364?origin=resultslist},
doi = {10.1016/j.jfludis.2026.106194},
issn = {0094730X (ISSN)},
year = {2026},
date = {2026-01-01},
journal = {Journal of Fluency Disorders},
volume = {87},
publisher = {Elsevier Inc.},
abstract = {Purpose Speaking in class is challenging for students who stutter. Cognitive-behavioral therapy (CBT) with exposure in virtual reality (VR) emerges as a promising intervention for treating speaking anxiety in pediatric populations. This pilot study tested if real-time avatar-based VR can elicit anxiety responses while remaining acceptable to youth who stutter. Method Twelve students who stutter (aged 9–18) were randomly assigned to a single training session conducted either (1) in VR with a realistic avatar controlled live by their SLP, or (2) in role-play with their SLP, before facing a real actor. We assessed system acceptability, anxiety levels and perceived self-efficacy. Results The VR system was well accepted and elicited physiological arousal comparable to real-life interactions. Although participants reported experiencing less anxiety during VR, skin conductance level showed higher arousal; suggesting a divergence between the subjective report and physiological response. Finally, one training session (either in VR or with the SLP) did not produce gains in self-efficacy or decrease in anxiety related to the final real-actor conversation. Conclusion This study demonstrates evidence that the potential use of immersive VR could represent an acceptable and viable complementary strategy for SLP treatment, that could control exposure parameters while evoking physiological responses similar to real-life contexts. The differences between subjective and physiological measures suggest that VR is inducing anxiety responses differently than it was perceived. Further research could investigate the use of VR as anxiety interventions for students who stutter and should be explored across multi-session studies to understand their therapeutic effect. © 2026 The Authors.},
keywords = {Adolescent, adult, Anxiety, Arousal, article, Avatar, avoidance behavior, Canada, Child, clinical article, cognitive behavioral therapy, complication, controlled study, distance learning, ecological validity, electrocardiogram, electrodermal response, exposure, female, Heart Rate, heart rate variability, human, Humans, Liebowitz Social Anxiety Scale, Likert scale, male, nonverbal communication, physiological stress, Pilot Projects, pilot study, procedures, psychology, questionnaire, randomized controlled trial, role playing, School-age children, Self Concept, Self Efficacy, self report, Signal processing, skin conductance, social anxiety, speech, student, Students, Stuttering, Therapy, treatment outcome, virtual reality, virtual reality exposure therapy},
pubstate = {published},
tppubtype = {article}
}
Purpose Speaking in class is challenging for students who stutter. Cognitive-behavioral therapy (CBT) with exposure in virtual reality (VR) emerges as a promising intervention for treating speaking anxiety in pediatric populations. This pilot study tested if real-time avatar-based VR can elicit anxiety responses while remaining acceptable to youth who stutter. Method Twelve students who stutter (aged 9–18) were randomly assigned to a single training session conducted either (1) in VR with a realistic avatar controlled live by their SLP, or (2) in role-play with their SLP, before facing a real actor. We assessed system acceptability, anxiety levels and perceived self-efficacy. Results The VR system was well accepted and elicited physiological arousal comparable to real-life interactions. Although participants reported experiencing less anxiety during VR, skin conductance level showed higher arousal; suggesting a divergence between the subjective report and physiological response. Finally, one training session (either in VR or with the SLP) did not produce gains in self-efficacy or decrease in anxiety related to the final real-actor conversation. Conclusion This study demonstrates evidence that the potential use of immersive VR could represent an acceptable and viable complementary strategy for SLP treatment, that could control exposure parameters while evoking physiological responses similar to real-life contexts. The differences between subjective and physiological measures suggest that VR is inducing anxiety responses differently than it was perceived. Further research could investigate the use of VR as anxiety interventions for students who stutter and should be explored across multi-session studies to understand their therapeutic effect. © 2026 The Authors.