Publications

Exploring WavLM back-ends for speech spoofing and deepfake detection

Abstract

This paper describes our submitted systems to the ASVspoof 5 Challenge Track 1: Speech Deepfake Detection - Open Condition, which consists of a stand-alone speech deepfake (bonafide vs spoof) detection task. Recently, large-scale self-supervised models become a standard in Automatic Speech Recognition (ASR) and other speech processing tasks. Thus, we leverage a pre-trained WavLM as a front-end model and pool its representations with different back-end techniques. The complete framework is fine-tuned using only the trained dataset of the challenge, similar to the close condition. Besides, we adopt data-augmentation by adding noise and reverberation using MUSAN noise and RIR datasets. We also experiment with codec augmentations to increase the performance of our method. Ultimately, we use the Bosaris toolkit for score calibration and system fusion to get better Cllr scores. Our fused system achieves $0.0937$ minDCF, $3.42%$ EER, $0.1927$ Cllr, and $0.1375$ actDCF.

Continue reading

Navigating pharmacophore space to identify activity discontinuities: A case study with BCR-ABL

Abstract

Abstract The exploration of chemical space is a fundamental aspect of chemoinformatics, particularly when one explores a large compound data set to relate chemical structures with molecular properties. In this study, we extend our previous work on chemical space visualization at the pharmacophoric level. Instead of using conventional binary classification of affinity (active vs inactive), we introduce a refined approach that categorizes compounds into four distinct classes based on their activity levels: super active, very active, active, and inactive. This classification enriches the color scheme applied to pharmacophore space, where the color representation of a pharmacophore hypothesis is driven by the associated compounds. Using the BCR-ABL tyrosine kinase as a case study, we identified intriguing regions corresponding to pharmacophore activity discontinuities, providing valuable insights for structure-activity relationships analysis.

Continue reading

Towards supervised performance on speaker verification with self-supervised learning by leveraging large-scale ASR models

Abstract

Recent advancements in Self-Supervised Learning (SSL) have shown promising results in Speaker Verification (SV). However, narrowing the performance gap with supervised systems remains an ongoing challenge. Several studies have observed that speech representations from large-scale ASR models contain valuable speaker information. This work explores the limitations of fine-tuning these models for SV using an SSL contrastive objective in an end-to-end approach. Then, we propose a framework to learn speaker representations in an SSL context by fine-tuning a pre-trained WavLM with a supervised loss using pseudo-labels. Initial pseudo-labels are derived from an SSL DINO-based model and are iteratively refined by clustering the model embeddings. Our method achieves 0.99% EER on VoxCeleb1-O, establishing the new state-of-the-art on self-supervised SV. As this performance is close to our supervised baseline of 0.94% EER, this contribution is a step towards supervised performance on SV with SSL.

Continue reading

A generic, efficient, and interactive approach to image processing with applications in mathematical morphology

Abstract

Continue reading

Informatique mathématique. Une photographie en 2023

Abstract

Continue reading

A CP-based automatic tool for instantiating truncated differential characteristics

Abstract

An important criteria to assert the security of a cryptographic primitive is its resistance against differential cryptanalysis. For word-oriented primitives, a common technique to determine the number of rounds required to ensure the immunity against differential distinguishers is to consider truncated differential characteristics and to count the number of active S-boxes. Doing so allows to provide an upper bound on the probability of the best differential characteristic with a reduced computational cost. However, in order to design very efficient primitives, it might be needed to evaluate the probability more accurately. This is usually done in a second step, during which one tries to instantiate truncated differential characteristics with actual values and computes its corresponding probability. This step is usually done with ad-hoc algorithms and CP or MILP models for generic solvers. In this paper, we present a generic approach for automatically generating these models to handle all word-oriented ciphers. Furthermore the running times to solve these models are very competitive with all the previous dedicated approaches.

Continue reading

Ce que nous savons sur (les) sciences du jeu : Analyse bibliométrique et lexicométrique des articles de la revue (octobre 2013 - mai 2022)

Abstract

On its website, Sciences du jeu describes itself as an “international and interdisciplinary journal whose mission is to develop and promote French-speaking researc[h] on play”, “to foster dialogue between social sciences and set off debates on this particular subject” Created in 2013 following a study day in tribute to the work of Jacques Henriot, its scientific program clearly follows in his footsteps. Indeed, Sciences du jeu defines itself not only as “open to all approaches or methods”, but also to “every aspect of play” (including, but not exclusively, video games) and to “researc[h] from various fields related to play in a broad sense (objects, structures, situations, experiences, attitudes)”.. Ten years after the publication of the first issue of the journal, we may well ask to what extent the articles published to date reflect the original approach of play originally promoted by Henriot. What about the references to this author and to the concepts he developed in his work? More generally, what are the bibliographical references most frequently used by the journal’s authors? What do they tell us about their conception of play and how they approach it? On which disciplinary approaches and methods are their analyses most often based? What types of games, themes and/or fields are most frequently studied? What are the gray areas and less visible fields? Finally, who are the authors of these papers (in terms of gender and status), where do they come from (in terms of affiliation and disciplinary roots) and how does this influence their perspective on play? To answer these questions, this paper draws on a bibliometric, lexicometric and sociological analysis based on a corpus comprising all the articles published in the first seventeen issues of the journal.

Continue reading

Closure and decision properties for higher-dimensional automata

Abstract

Continue reading

Performance evaluation of container management tasks in OS-level virtualization platforms

Abstract

Cloud computing is a method for accessing and managing computing resources over the internet, providing flexibility, scalability, and cost-efficiency. Cloud computing relies more and more on OS-level virtualization tools such as Docker and Podman, enabling users to create and run containers, which are widely used for application management. Given its significance in cloud infrastructures, it is crucial to have a better understanding of OS-level virtualization performance, especially in tasks related to container management (ex: creation, destruction). In this paper, we conducted benchmarking tests on Docker and Podman to evaluate their performance in various container management scenarios and with different image sizes. The results revealed that Podman excels in quickly instantiating small-sized containers, while Docker demonstrates superior performance with larger-sized containers.

Continue reading

An improved spectral extraction method for JWST/NIRSpec fixed slit observations

Abstract

The James Webb Space Telescope is performing beyond our expectations. Its Near Infrared Spectrograph (NIRSpec) provides versatile spectroscopic capabilities in the 0.6-5.3 micrometre wavelength range, where a new window is opening for studying Trans-Neptunian objects in particular. We propose a spectral extraction method for NIRSpec fixed slit observations, with the aim of meeting the superior performance on the instrument with the most advanced data processing. We applied this method on the fixed slit dataset of the guaranteed-time observation program 1231, which targets Plutino 2003 AZ84. We compared the spectra we extracted with those from the calibration pipeline.

Continue reading