Publications

2024

• Explainable by-design Audio Segmentation through Non-Negative Matrix Factorization and Probing
  
  • Lebourdais Martin
  • Mariotte Théo
  • Almudévar Antonio
  • Tahon Marie
  • Ortega Alfonso
  , 2024. Audio segmentation is a key task for many speech technologies, most of which are based on neural networks, usually considered as black boxes, with high-level performances.However, in many domains, among which health or forensics, there is not only a need for good performance but also for explanations about the output decision.Explanations derived directly from latent representations need to satisfy "good" properties, such as informativeness, compactness, or modularity, to be interpretable.In this article, we propose an explainable-by-design audio segmentation model based on non-negative matrix factorization (NMF) which is a good candidate for the design of interpretable representations.This paper shows that our model reaches good segmentation performances, and presents deep analyses of the latent representation extracted from the non-negative matrix.The proposed approach opens new perspectives toward the evaluation of interpretable representations according to "good" properties.
• Speech dereverberation constrained on room impulse response characteristics
  
  • Bahrman Louis
  • Fontaine Mathieu
  • Le Roux Jonathan
  • Richard Gaël
  , 2024. Single-channel speech dereverberation aims at extracting a dry speech signal from a recording affected by the acoustic reflections in a room. However, most current deep learning-based approaches for speech dereverberation are not interpretable for room acoustics, and can be considered as black-box systems in that regard. In this work, we address this problem by regularizing the training loss using a novel physical coherence loss which encourages the room impulse response (RIR) induced by the dereverberated output of the model to match the acoustic properties of the room in which the signal was recorded. Our investigation demonstrates the preservation of the original dereverberated signal alongside the provision of a more physically coherent RIR.
• STOCHASTIC MIRROR DESCENT FOR NONPARAMETRIC ADAPTIVE IMPORTANCE SAMPLING
  
  • Bianchi Pascal
  • Delyon Bernard
  • Portier François
  • Priser Victor
  , 2024. This paper addresses the problem of approximating an unknown probability distribution with density $f$ - which can only be evaluated up to an unknown scaling factor - with the help of a sequential algorithm that produces at each iteration $n\geq 1$ an estimated density $q_n$. The proposed method optimizes the Kullback-Leibler divergence using a mirror descent (MD) algorithm directly on the space of density functions, while a stochastic approximation technique helps to manage between algorithm complexity and variability. One of the key innovations of this work is the theoretical guarantee that is provided for an algorithm with a fixed MD learning rate \(\eta \in (0,1 )\). The main result is that the sequence \(q_n\) converges almost surely to the target density \(f\) uniformly on compact sets. Through numerical experiments, we show that fixing the learning rate \(\eta \in (0,1 )\) significantly improves the algorithm's performance, particularly in the context of multi-modal target distributions where a small value of $\eta$ allows to increase the chance of finding all modes. Additionally, we propose a particle subsampling method to enhance computational efficiency and compare our method against other approaches through numerical experiments.
• Meta-learners for few-shot weakly-supervised medical image segmentation
  
  • Oliveira Hugo
  • Gama Pedro H.T.
  • Bloch Isabelle
  • Cesar Roberto Marcondes
  Pattern Recognition, Elsevier, 2024, 153, pp.110471-1:110471-13. Most uses of Meta-Learning in visual recognition are very often applied to image classification, with a relative lack of work in other tasks such as segmentation and detection. We propose a new generic Meta-Learning framework for few-shot weakly supervised segmentation in medical imaging domains. The proposed approach includes a meta-training phase that uses a meta-dataset. It is deployed on an out-of-distribution few-shot target task, where a single highly generalizable model, trained via a selective supervised loss function, is used as a predictor. The model can be trained in several distinct ways, such as second-order optimization, metric learning, and late fusion. Some relevant improvements of existing methods that are part of the proposed approach are presented. We conduct a comparative analysis of meta-learners from distinct paradigms adapted to few-shot image segmentation in different sparsely annotated radiological tasks. The imaging modalities include 2D chest, mammographic, and dental X-rays, as well as 2D slices of volumetric tomography and resonance images. Our experiments consider in total 9 meta-learners, 4 backbones, and multiple target organ segmentation tasks. We explore small-data scenarios in radiology with varying weak annotation styles and densities. Our analysis shows that metric-based meta-learning approaches achieve better segmentation results in tasks with smaller domain shifts compared to the meta-training datasets, while some gradient- and fusion-based meta-learners are more generalizable to larger domain shifts. Guidelines learned from the comparative performance assessment of the analyzed methods are summarized to support those readers interested in the field. (10.1016/j.patcog.2024.110471)
  DOI : 10.1016/j.patcog.2024.110471
• Simulation Secure Multi-input Quadratic Functional Encryption
  
  • Alborch Escobar Ferran
  • Canard Sébastien
  • Laguillaumie Fabien
  , 2025, 15516, pp.26-53. Multi-input functional encryption is a primitive that allows for the evaluation of an -ary function over multiple ciphertexts, without learning any information about the underlying plaintexts. This type of computation is useful in many cases where one has to compute over encrypted data, such as privacy-preserving cloud services, federated learning, or more generally delegation of computation from multiple clients. In this work we propose the first secret-key multi-input quadratic functional encryption scheme satisfying simulation security. On contrary, current constructions supporting quadratic functionalities, proposed by Agrawal et al. in CRYPTO '21 and TCC '22, only reach indistinguishibility-based security. Our proposed construction is generic, and for a concrete instantiation, we propose a new function-hiding innerproduct functional encryption scheme proven simulation secure against one challenge ciphertext in the standard model, which is of independent interest. (10.1007/978-3-031-82852-2_2)
  DOI : 10.1007/978-3-031-82852-2_2
• Leveraging Reusable Code and Proofs to Design Complex DRAM Controllers - A Case Study
  
  • Malaquias Felipe Lisboa
  • Asavoae Mihail
  • Brandner Florian
  , 2024, pp.298-305. Critical real-time systems are getting more and more complex and require ever more computing power. Multi-core platforms, GPUs, and custom accelerators promise to deliver this needed performance. However, these platforms are notoriously hard to analyze and lack predictability in terms of timing properties. Computer architectures and platforms that offer both predictability and performance are thus needed. This work investigates the use of the interactive proof assistant Coq in order to model complex DRAM memory controllers (MCs) for multi-core platforms. The design of predictable high- performance MCs is particularly challenging, since memory requests have to be processed efficiently, while facing interference from other cores in the system. The problem is exacerbated by the complexity of DRAM devices and the various timing constraints they impose. Specifically, this work extends a previous Coq framework by focusing on reusability, which allows designers to develop and prove complex MCs. As a use-case, we present TDMShelve, an MC balancing performance and isolation. (10.1109/DSD64264.2024.00047)
  DOI : 10.1109/DSD64264.2024.00047
• Device-independent lower bounds on the conditional von Neumann entropy
  
  • Brown Peter
  • Fawzi Hamza
  • Fawzi Omar
  Quantum, Verein zur Förderung des Open Access Publizierens in den Quantenwissenschaften, 2024, 8, pp.1445. The rates of several device-independent (DI) protocols, including quantum key-distribution (QKD) and randomness expansion (RE), can be computed via an optimization of the conditional von Neumann entropy over a particular class of quantum states. In this work we introduce a numerical method to compute lower bounds on such rates. We derive a sequence of optimization problems that converge to the conditional von Neumann entropy of systems defined on general separable Hilbert spaces. Using the Navascu\'es-Pironio-Ac\'in hierarchy we can then relax these problems to semidefinite programs, giving a computationally tractable method to compute lower bounds on the rates of DI protocols. Applying our method to compute the rates of DI-RE and DI-QKD protocols we find substantial improvements over all previous numerical techniques, demonstrating significantly higher rates for both DI-RE and DI-QKD. In particular, for DI-QKD we show a new minimal detection efficiency threshold which is within the realm of current capabilities. Moreover, we demonstrate that our method is capable of converging rapidly by recovering instances of known tight analytical bounds. Finally, we note that our method is compatible with the entropy accumulation theorem and can thus be used to compute rates of finite round protocols and subsequently prove their security. (10.22331/q-2024-08-27-1445)
  DOI : 10.22331/q-2024-08-27-1445
• Multifrequency Highly Oscillating Aperiodic Amplitude Estimation for Nonlinear Chirp Signal
  
  • Emelchenkov Anton
  • Fontaine Mathieu
  • Grenier Yves
  • Mahé Hervé
  • Roueff François
  , 2024. This paper addresses the challenge of estimating multiple highly oscillating amplitudes within the nonlinear chirp signal model. The problem is analogous to the mode detection task with fixed instantaneous frequencies, where the oscillating amplitudes signify mechanical vibrations concealing crucial information for predictive maintenance. Existing methods often focus on single-frequency estimation, employ simple amplitude functions, or impose strong noise assumptions. Furthermore, these methods frequently rely on arbitrarily chosen hyperparameters, leading to sub-optimal generalization for a diverse range of amplitudes. To address these limitations, our approach introduces two estimators, based on Capon filters and negative log-likelihood approaches respectively, that leverage locally stationary assumptions and incorporate hyperparameters estimation. The results demonstrate that, even under challenging conditions, these estimators yield competitive outcomes across various noisy scenarios, mitigating the drawbacks associated with existing methods. (10.23919/EUSIPCO63174.2024.10715060)
  DOI : 10.23919/EUSIPCO63174.2024.10715060
• Deep unrolling of the multiplicative updates algorithm for blind source separation, with application to hyperspectral unmixing
  
  • Kervazo Christophe
  • Chetoui Abdelkhalak
  • Cohen Jérémy E.
  , 2024. Blind Source Separation (BSS) has gained a large interest in many fields, including hyperspectral unmixing which is broadly used in remote sensing and astrophysics. BSS being an ill-posed problem, many strategies have been proposed to solve it, ranging from model-based to deep-learning ones. While modelbased algorithms are in general interpretable, in contrast with neural networks, these algorithms often require a large number of iterations and obtain worse unmixing results that their deeplearning counterparts. To try to obtain the best of both worlds, in this work we unroll the multiplicative updates algorithm, leading to two new algorithms. The first one, NALMU, learns some parameters which are fixed once the training is over. The second one, ALMU, enables the parameters of the unrolled algorithm to be predicted by small neural networks, making the whole algorithm adaptative to the specific datasets considered in the test phase. We conduct experiments on two astrophysic datasets, and show that our approach enables to largely outperform the other unmixing unrolled algorithms, while largely reducing the number of iterations compared to the original multiplicative updates algorithm.
• Unsupervised radiometric change detection from synthetic aperture radar images
  
  • Bultingaire Thomas
  • Meraoumia Inès
  • Kervazo Christophe
  • Denis Loïc
  • Tupin Florence
  , 2024. Change detection is an important data processing task in remote sensing, with applications such as deforestation monitoring or natural disaster assessment. Synthetic Aperture Radar (SAR) imaging offers key advantages for change detection, in particular due to its robustness to weather condition and cloud coverage. Because of the speckle phenomenon, the intensity of SAR images suffer from strong fluctuations, making the detection of radiometric changes challenging. Our method builds on a recently introduced self-supervised despeckling technique. It estimates despeckling uncertainty to better identify meaningful differences between two despeckled images. Conformal prediction permits to approach the change detection problem from the angle of anomaly detection. Thus, we develop a fully unsupervised change detection approach with a controlled probability of false alarm. Experimental results on TerraSAR-X satellite images with metric resolution show the capability of our method to detect changes without any supervision.
• Using Random Codebooks for Audio Neural AutoEncoders
  
  • Giniès Benoît
  • Bie Xiaoyu
  • Fercoq Olivier
  • Richard Gaël
  , 2024. Latent representation learning has been an active field of study for decades in numerous applications. Inspired among others by the tokenization from Natural Language Processing and motivated by the research of a simple data representation, recent works have introduced a quantization step into the feature extraction. In this work, we propose a novel strategy to build the neural discrete representation by means of random codebooks. These codebooks are obtained by randomly sampling a large, predefined fixed codebook. We experimentally show the merits and potential of our approach in a task of audio compression and reconstruction. (10.23919/EUSIPCO63174.2024.10715290)
  DOI : 10.23919/EUSIPCO63174.2024.10715290
• Towards a Perfect Reconstruction Theory
  
  • Rioul Olivier
  • Souloumiac Antoine
  , 2024, pp.2547-2551. This paper presents some preliminary considerations on the general problems of missing/complementary information and perfect reconstruction, with the hope to attract the attention of the signal processing researchers that the rigorous derivation of a general theory of reconstruction should be desirable and possible. (10.23919/EUSIPCO63174.2024.10715178)
  DOI : 10.23919/EUSIPCO63174.2024.10715178
• Invariance-based layer regularization for sound event detection
  
  • Perera David
  • Essid Slim
  • Gaël Richard
  , 2024. Experimental and theoretical evidences suggest that invariance constraints can improve the performance and generalization capabilities of a classification model. While invariance-based regularization has become part of the standard tool-belt of machine learning practitioners, this regularization is usually applied near the decision layers or at the end of the feature extracting layers of a deep classification network. However, the optimal placement of invariance constraints inside a deep classifier is yet an open question. In particular, it would be beneficial to link it to the structural properties of the network (\textit{e.g.} its architecture), or its dynamical properties (\textit{e.g.} the effectively used volume of its latent spaces). The purpose of this article is to initiate an investigation on these aspects. We use the experimental framework of the DCASE 2023 Task 4A challenge, which considers the training of a sound event classifier in a semi-supervised manner. We show that the optimal placement of invariance constraints improves the performance of the standard baseline for this task. (10.23919/EUSIPCO63174.2024.10715380)
  DOI : 10.23919/EUSIPCO63174.2024.10715380
• Generating synthetic data to train a deep unrolled network for Hyperspectral Unmixing
  
  • Hadjeres Rassim
  • Kervazo Christophe
  • Tupin Florence
  , 2024. Hyperspectral unmixing is an essential tool for analyzing hyperspectral data, especially in remote sensing. Many approaches have been developed for this problem, ranging from model-based to deep learning-based, and (hybrid) unrolled methods. However, the development of supervisedly trained deep learning-based unmixing methods is hindered by the lack of available labeled training datasets. In this paper, to enable the supervised training of neural networks for hyperspectral unmixing, we propose a methodology to construct a synthetic training database directly from the hyperspectral image to unmix. We use this data generation approach to train an unrolled unmixing method LPALM. The trained LPALM is assessed on two real hyperspectral datasets and shows the best performances compared to other classical, unrolled, and autoencoder-based unmixing methods. The code of this work will be available at https://github.com/rhadjeres/Synthetic
• Anomalous Sound Detection For Road Surveillance Based On Graph Signal Processing
  
  • Mnasri Zied
  • Giraldo Zuluaga Jhony Heriberto
  • Bouwmans Thierry
  , 2024. Recently, Anomalous Sound Detection (ASD) has emerged as a promising method for road surveillance. However, since the ratio of anomalous events is generally too small, anomaly detection in general, and ASD in particular, are mainly treated as one-class classification problems. Besides, a common problem in road surveillance is the background noise, which makes it difficult to train effective models based on normal sounds only. Therefore, this work aims to experiment with the use of graph signal processing (GSP) to improve ASD performance. Thus, we propose a Graph-based One-Class SVM technique (GOC-SVM) where the features extracted from audio signals are firstly embedded on graphs, and then filtered through a graph filterbank, before computing their joint Fourier transform magnitude. Subsequently, they are fed into a one-class SVM classifier trained on normal data only. Evaluation results show a threefold advantage of using graph embedding and filtering for ASD: (a) improving the anomaly detection results in comparison to plain features, (b) outperforming the classical OC-SVM baseline, (c) enhancing the performance of the proposed semi-supervised GOC-SVM, so as to reach a comparable level of performance of the fully-supervised binary classification SVM, yielding 0.91 of Area-under-the-curve (AUC), 98% of overall accuracy, 99% and 88% of F1 score for normal and anomalous classes, respectively. Such a performance proves the potential of using GSP to solve the ASD problem in road traffic monitoring. (10.23919/EUSIPCO63174.2024.10715291)
  DOI : 10.23919/EUSIPCO63174.2024.10715291
• Shrinkage MMSE estimators of covariances beyond the zero-mean and stationary variance assumptions
  
  • Flasseur Olivier
  • Thiébaut Éric
  • Denis Loïc
  • Langlois Maud
  , 2024. We tackle covariance estimation in low-sample scenarios, employing a structured covariance matrix with shrinkage methods. These involve convexly combining a low-bias/highvariance empirical estimate with a biased regularization estimator, striking a bias-variance trade-off. Literature provides optimal settings of the regularization amount through risk minimization between the true covariance and its shrunk counterpart. Such estimators were derived for zero-mean statistics with i.i.d. diagonal regularization matrices accounting for the average sample variance solely. We extend these results to regularization matrices accounting for the sample variances both for centered and noncentered samples. In the latter case, the empirical estimate of the true mean is incorporated into our shrinkage estimators. Introducing confidence weights into the statistics also enhance estimator robustness against outliers. We compare our estimators to other shrinkage methods both on numerical simulations and on real data to solve a detection problem in astronomy.
• InSAR2InSAR: A Self-Supervised Method for InSAR Parameters Estimation
  
  • Geara Carla
  • Gelas Colette
  • De Vitry Louis
  • Colin Elise
  • Tupin Florence
  , 2024. <div><p>Interferometric Synthetic Aperture Radar (InSAR) is a remote sensing tool that provides comprehensive information about the Earth's surface. However, InSAR parameters are highly corrupted by speckle, which limits their utility. Deep learning methods have recently achieved promising results in improving the reliability of InSAR parameters. Most of the proposed methods are fully supervised. These methods are usually trained on synthetic data, which are not able to fully take into account all the properties of real images. In this paper, we address this issue by extending the self-supervised denoising approach Noise2Noise, previously proposed by Lehtinen et al. in 2018, for the joint estimation of InSAR parameters. Additionally, the proposed method uses a loss function that is adapted to the InSAR noise model, making it well-suited for the problem we are addressing.</p></div>
• Diffusion-based image inpainting with internal learning
  
  • Cherel Nicolas
  • Almansa Andrés
  • Gousseau Yann
  • Newson Alasdair
  , 2024, pp.446-450. Diffusion models are now the undisputed state-of-the-art for image generation and image restoration. However, they require large amounts of computational power for training and inference. In this paper, we propose lightweight diffusion models for image inpainting that can be trained on a single image, or a few images. We show that our approach competes with large state-of-the-art models in specific cases. We also show that training a model on a single image is particularly relevant for image acquisition modality that differ from the RGB images of standard learning databases. We show results in three different contexts: texture images, line drawing images, and materials BRDF, for which we achieve state-of-the-art results in terms of realism, with a computational load that is greatly reduced compared to concurrent methods. (10.23919/EUSIPCO63174.2024.10714982)
  DOI : 10.23919/EUSIPCO63174.2024.10714982
• Silicone-Based Haptic Interfaces: Enhancing Multimodal Interactions through Pneumatic Tactile Feedback
  
  • Liu Yang
  • Shangguan Zhegong
  • Tapus Adriana
  • Safin Stéphane
  • Détienne Françoise
  • Lecolinet Eric
  , 2024, pp.140-143. This paper explores the potential of pneumatic haptic interfaces in enhancing human-computer interaction. We present two projects: one augmenting movie experiences with emotion-synchronized haptic feedback, and another integrating pneumatic interfaces into steering wheels for improved driver takeover in autonomous vehicles. The movie experience project demonstrated enhanced emotional engagement, while the automotive application showed improved safety and user trust. These applications highlight the versatility of pneumatic haptic technology across entertainment and safety contexts. We discuss the advantages of the Baromorph technique and outline future research directions, including comparative studies with vibrotactile feedback and machine learning approaches. This work contributes to the development of more insightful and emotionally intelligent interactive systems. (10.1109/IHMSC62065.2024.00039)
  DOI : 10.1109/IHMSC62065.2024.00039
• Bounds on Petz-Rényi Divergences and their Applications for Device-Independent Cryptography
  
  • Hahn Thomas
  • Tan Ernest
  • Brown Peter
  , 2024. Variational techniques have been recently developed to find tighter bounds on the von Neumann entropy in a completely device-independent (DI) setting. This, in turn, has led to significantly improved key rates of DI protocols, in both the asymptotic limit as well as in the finite-size regime. In this paper, we discuss two approaches towards applying these variational methods for Petz-Rényi divergences instead. We then show how this can be used to further improve the finite-size key rate of DI protocols, utilizing a fully-Rényi entropy accumulation theorem developed in a partner work. Petz-Rényi divergences can also be applied to study DI advantage distillation, in which two-way communication is used to improve the noise tolerance of quantum key distribution (QKD) protocols. We implement these techniques to derive increased noise tolerances for DIQKD protocols, which surpass all previous known bounds. (10.48550/arXiv.2408.12313)
  DOI : 10.48550/arXiv.2408.12313
• Platooning in Connected Vehicles: A Review of Current Solutions, Standardization Activities, Cybersecurity, and Research Opportunities
  
  • Braiteh Farah-Emma
  • Bassi Francesca
  • Khatoun Rida
  IEEE Transactions on Intelligent Vehicles, Institute of Electrical and Electronics Engineers, 2024, pp.1-23. Over the past few years, we have witnessed notable advancements in connected vehicle technologies, illustrating significant progress in their interactions with drivers, as well as with various other networks and devices. One remarkable advancement is the Cooperative Adaptive Cruise Control (CACC) and the specialized service known as the platoon system, which entails vehicles traveling closely together in a convoy-like manner. Numerous pilot projects, academic studies, and standardization initiatives have been dedicated to exploring platooning functions. In this paper, we provide an overview of platooning and CACC protocols, examining them from both academic and industry perspectives. We analyze and compare some of the recently published research works and projects related to Platooning. Furthermore, we discuss and analyze the cybersecurity requirements and potential threats associated with platooning. Finally, we discuss some key research challenges and open research issues in the platooning context that require further investigation in the future by the academic and the industrial communities. (10.1109/TIV.2024.3447916)
  DOI : 10.1109/TIV.2024.3447916
• Formal security proofs via Doeblin coefficients: Optimal side-channel factorization from noisy leakage to random probing
  
  • Béguinot Julien
  • Cheng Wei
  • Guilley Sylvain
  • Rioul Olivier
  , 2024, 14925, pp.389-426. Masking is one of the most popular countermeasures to side- channel attacks, because it can offer provable security. However, depend- ing on the adversary’s model, useful security guarantees can be hard to provide. At first, masking has been shown secure against t-threshold probing adversaries by Ishai et al. at Crypto’03. It has then been shown secure in the more generic random probing model by Duc et al. at Euro- crypt’14. Prouff and Rivain have introduced the noisy leakage model to capture more realistic leakage at Eurocrypt’13. Reduction from noisy leakage to random probing has been introduced by Duc et al. at Euro- crypt’14, and security guarantees were improved for both models by Prest et al. at Crypto’19, Duc et al. in Eurocrypt’15/J. Cryptol’19, and Masure and Standaert at Crypto’23. Unfortunately, as it turns out, we found that previous proofs in either random probing or noisy leakage models are flawed, and such flaws do not appear easy to fix. In this work, we show that the Doeblin coefficient allows one to over- come these flaws. In fact, it yields optimal reductions from noisy leakage to random probing, thereby providing a correct and usable metric to properly ground security proofs. This shows the inherent inevitable cost of a reduction from the noisy leakages to the random probing model. We show that it can also be used to derive direct formal security proofs using the subsequence decomposition of Prouff and Rivain. (10.1007/978-3-031-68391-6_12)
  DOI : 10.1007/978-3-031-68391-6_12
• Public-Key Anamorphism in (CCA-Secure) Public-Key Encryption and Beyond
  
  • Persiano Giuseppe
  • Phan Duong Hieu
  • Yung Moti
  , 2024, 14921, pp.422-455. The notion of (Receiver-) Anamorphic Encryption was put forth recently to show that a dictator (i.e., an overreaching government), which demands to get the receiver’s private key and even dictates messages to the sender, cannot prevent the receiver from getting an additional covert anamorphic message from a sender. The model required an initial private collaboration to share some secret. There may be settings though where an initial collaboration may be impossible or performance-wise prohibitive, or cases when we need an immediate message to be sent without private key generation (e.g., by any casual sender in need). This situation, to date, somewhat limits the applicability of anamorphic encryption. To overcome this, in this work, we put forth the new notion of “public-key anamorphic encryption,” where, without any initialization, any sender that has not coordinated in any shape or form with the receiver, can nevertheless, under the dictator control of the receiver’s private key, send the receiver an additional anamorphic secret message hidden from the dictator. We define the new notion with its unique new properties, and then prove that, quite interestingly, the known CCA-secure Koppula-Waters (KW) system is, in fact, public-key anamorphic. We then describe how a public-key anamorphic scheme can support a new hybrid anamorphic encapsulation mode (KDEM) where the public-key anamorphic part serves a bootstrapping mechanism to activate regular anamorphic messages in the same ciphertext, thus together increasing the anamorphic channel capacity. Looking at the state of research thus far, we observe that the initial system (Eurocrypt’22) that was shown to have regular anamorphic properties is the CCA-secure Naor-Yung (and other related schemes). Here we identify that the KW CCA-secure scheme also provides a new type of anamorphism. Thus, this situation is hinting that there may be a connection between some types of CCA-secure schemes and some type of anamorphic schemes (in spite of the fact that the goals of the two primitives are fundamentally different); this question is foundational in nature. Given this, we identify a sufficient condition for a “CCA-secure scheme which is black-box reduced from a CPA secure scheme” to directly give rise to an “anamorphic encryption scheme!” Furthermore, we identify one extra property of the reduction, that yields a public-key anamorphic scheme as defined here. (10.1007/978-3-031-68379-4_13)
  DOI : 10.1007/978-3-031-68379-4_13
• DynSplit: A Dynamic Split Learning Scheme for 5G-Enpowered Metaverse
  
  • Shu Yunmeng
  • Gu Pengwenlong
  • Adjih Cédric
  • Chen Chung Shue
  • Serhrouchni Ahmed
  , 2024, pp.214-221. The Metaverse is a virtual world based on numerous technologies, which enables users to interact socially in a persistent online 3-D virtual environment. To generate high-level imaginary environments, extremely low latency data transmission and learning-based sensor data analysis are required. With the development of 5G techniques, processing and learning methods, both the transmission delay and high-quality scene generation have been significantly improved in meta-applications. However, many Metaverse devices are battery-powered, and local processes and learning are still too costly. To address this issue, in this paper, by taking full architectural advantage of 5G networks, we propose a novel dynamic split learning scheme for enabled Metaverse systems. In our proposed scheme, each neural network is split into two segments, and the upper segment is stored at the base station (BS) side. Thus, between two segments, multiple pathways are featured, each with distinct compression ratios, accompanied by a gating mechanism that intelligently guides the selection of paths for each input data. This design excels in adapting to diverse Metaverse applications and network conditions, enhancing both the learning and computing phases of split models. Simulation results underscore the efficacy of our proposed scheme, revealing that it does not impede the convergence of split learning models. Furthermore, the scheme demonstrates notable performance gains in terms of communication overhead, prediction accuracy, and adaptability to resource constraints. (10.1109/MetaCom62920.2024.00043)
  DOI : 10.1109/MetaCom62920.2024.00043
• Deep Reinforcement Learning for Uplink Scheduling in NOMA-URLLC Networks
  
  • Robaglia Benoît-Marie
  • Coupechoux Marceau
  • Tsilimantos Dimitrios
  IEEE Transactions on Machine Learning in Communications and Networking, Institute of Electrical and Electronics Engineers, 2024, 2, pp.1142-1158. This article addresses the problem of Ultra Reliable Low Latency Communications (URLLC) in wireless networks, a framework with particularly stringent constraints imposed by many Internet of Things (IoT) applications from diverse sectors. We propose a novel Deep Reinforcement Learning (DRL) scheduling algorithm, named NOMA-PPO, to solve the Non-Orthogonal Multiple Access (NOMA) uplink URLLC scheduling problem involving strict deadlines. The challenge of addressing uplink URLLC requirements in NOMA systems is related to the combinatorial complexity of the action space due to the possibility to schedule multiple devices, and to the partial observability constraint that we impose to our algorithm in order to meet the IoT communication constraints and be scalable. Our approach involves 1) formulating the NOMA-URLLC problem as a Partially Observable Markov Decision Process (POMDP) and the introduction of an agent state, serving as a sufficient statistic of past observations and actions, enabling a transformation of the POMDP into a Markov Decision Process (MDP); 2) adapting the Proximal Policy Optimization (PPO) algorithm to handle the combinatorial action space; 3) incorporating prior knowledge into the learning agent with the introduction of a Bayesian policy. Numerical results reveal that not only does our approach outperform traditional multiple access protocols and DRL benchmarks on 3GPP scenarios, but also proves to be robust under various channel and traffic configurations, efficiently exploiting inherent time correlations. (10.1109/TMLCN.2024.3437351)
  DOI : 10.1109/TMLCN.2024.3437351