Publications

2021

• Innovative Countermeasures to Defeat Cyber Attacks Against Blockchain Wallets
  
  • Urien Pascal
  , 2021, pp.49-54. Blockchain transactions are signed by private keys. Secure key storage and tamper resistant computing, are critical requirements for deployments of trusted infrastructure. In this paper we identify some threats against blockchain wallets, and we introduce a set of physical and logical countermeasures in order to defeat them. We introduce open software and hardware architectures based on secure elements, which enable detection of cloned device and corrupted software. These technologies are based on resistant computing (javacard), smartcard anti cloning, smartcard self content attestation, applicative firewall, bare metal architecture, remote attestation, dynamic PUF (Physical Unclonable Function), and programming token as root of trust. (10.1109/CSNet52717.2021.9614649)
  DOI : 10.1109/CSNet52717.2021.9614649
• Bridging the gap between debiasing and privacy for deep learning
  
  • Barbano Carlo Alberto
  • Tartaglione Enzo
  • Grangetto Marco
  , 2021, pp.3799-3808. (10.1109/iccvw54120.2021.00424)
  DOI : 10.1109/iccvw54120.2021.00424
• A Unified Objective for Novel Class Discovery
  
  • Fini Enrico
  • Sangineto Enver
  • Lathuilière Stéphane
  • Zhong Zhun
  • Nabi Moin
  • Ricci Elisa
  , 2021. In this paper, we study the problem of Novel Class Discovery (NCD). NCD aims at inferring novel object categories in an unlabeled set by leveraging from prior knowledge of a labeled set containing different, but related classes. Existing approaches tackle this problem by considering multiple objective functions, usually involving specialized loss terms for the labeled and the unlabeled samples respectively, and often requiring auxiliary regularization terms. In this paper, we depart from this traditional scheme and introduce a UNified Objective function (UNO) for discovering novel classes, with the explicit purpose of favoring synergy between supervised and unsupervised learning. Using a multi-view self-labeling strategy, we generate pseudo-labels that can be treated homogeneously with ground truth labels. This leads to a single classification objective operating on both known and unknown classes. Despite its simplicity, UNO outperforms the state of the art by a significant margin on several benchmarks (~+10% on CIFAR-100 and +8% on ImageNet). The project page is available at: https://ncd-uno.github.io.
• Variance-sensitive confidence intervals for parametric and offline bandits
  
  • Faury Louis
  , 2021. In this dissertation we present recent contributions to the problem of optimization under bandit feedback through the design of variance-sensitive confidence intervals. We tackle two distincts topics: (1) the regret minimization task in Generalized Linear Bandits (GLBs), a broad class of non-linear parametric bandits and (2) the problem of off-line policy optimization under bandit feedback. For (1) we study the effects of non-linearity in GLBs and challenge the current understanding that a high level of non-linearity is detrimental to the exploration-exploitation trade-off. We introduce improved algorithms as well as a novel analysis that prove that if correctly handled, the regret minimization task in GLBs is not necessarily harder than for their linear counterparts. It can even be easier for some important members of the GLB family such as the Logistic Bandit. Our approach leverages a new confidence set which captures the non-linearity of the reward signal through its variance, along with a local treatment of the non-linearity through a so-called self-concordance analysis. For (2) we leverage results from the distributionally robust optimization framework to construct asymptotic variance-sensitive confidence intervals for the counterfactual evaluation of policies. This allows to ensure conservatism (sought out by risk-averse agents) while searching off-line for promising policies. Our confidence intervals lead to new counterfactual objectives which, contrary to their predecessors, are more suited for practical deployment thanks to their convex and composite natures.
• Incremental k-Nearest Neighbors Using Reservoir Sampling for Data Streams
  
  • Bahri Maroua
  • Bifet Albert
  , 2021, 12986, pp.122--137. The online and potentially infinite nature of data streams leads to the inability to store the flow in its entirety and thus restricts the storage to a part of – and/or synopsis information from – the stream. To process these evolving data, we need efficient and accurate methodologies and systems, such as window models (e.g., sliding windows) and summarization techniques (e.g., sampling, sketching, dimensionality reduction). In this paper, we propose, RW-kNN, a k-Nearest Neighbors (kNN) algorithm that employs a practical way to store information about past instances using the biased reservoir sampling to sample the input instances along with a sliding window to maintain the most recent instances from the stream. We evaluate our proposal on a diverse set of synthetic and real datasets and compare against state-of-the-art algorithms in a traditional test-then-train evaluation. Results show how our proposed RW-kNN approach produces high-predictive performance for both real and synthetic datasets while using a feasible amount of resources. (10.1007/978-3-030-88942-5_10)
  DOI : 10.1007/978-3-030-88942-5_10
• Click to Move: Controlling Video Generation with Sparse Motion
  
  • Ardino Pierfrancesco
  • de Nadai Marco
  • Lepri Bruno
  • Ricci Elisa
  • Lathuilière Stéphane
  , 2021. This paper introduces Click to Move (C2M), a novel framework for video generation where the user can control the motion of the synthesized video through mouse clicks specifying simple object trajectories of the key objects in the scene. Our model receives as input an initial frame, its corresponding segmentation map and the sparse motion vectors encoding the input provided by the user. It outputs a plausible video sequence starting from the given frame and with a motion that is consistent with user input. Notably, our proposed deep architecture incorporates a Graph Convolution Network (GCN) modelling the movements of all the objects in the scene in a holistic manner and effectively combining the sparse user motion information and image features. Experimental results show that C2M outperforms existing methods on two publicly available datasets, thus demonstrating the effectiveness of our GCN framework at modelling object interactions. The source code is publicly available at https://github.com/PierfrancescoArdino/C2M.
• Multi-View Radar Semantic Segmentation
  
  • Ouaknine Arthur
  • Newson Alasdair
  • Pérez Patrick
  • Tupin Florence
  • Rebut Julien
  , 2021. Understanding the scene around the ego-vehicle is key to assisted and autonomous driving. Nowadays, this is mostly conducted using cameras and laser scanners, despite their reduced performance in adverse weather conditions. Automotive radars are low-cost active sensors that measure properties of surrounding objects, including their relative speed, and have the key advantage of not being impacted by rain, snow or fog. However, they are seldom used for scene understanding due to the size and complexity of radar raw data and the lack of annotated datasets. Fortunately, recent open-sourced datasets have opened up research on classification, object detection and semantic segmentation with raw radar signals using end-to-end trainable models. In this work, we propose several novel architectures, and their associated losses, which analyse multiple "views" of the range-angle-Doppler radar tensor to segment it semantically. Experiments conducted on the recent CARRADA dataset demonstrate that our best model outperforms alternative models, derived either from the semantic segmentation of natural images or from radar scene understanding, while requiring significantly fewer parameters. Both our code and trained models are available at https://github.com/valeoai/MVRSS.
• Execution Trace Analysis for a Precise Understanding of Latency Violations
  
  • Zoor Maysam
  • Apvrille Ludovic
  • Pacalet Renaud
  , 2021. Despite the amount of proposed works for the verification of diverse model properties, understanding the root cause of latency requirements violation in execution traces is still an open-issue especially for complex HW/SW system-level designs: is it due to an unfavorable real-time scheduling, to contentions on buses, to the characteristics of functional algorithms or hardware components? This identification is particularly at stake when adding new features in a model, e.g., a new security countermeasure. The paper introduces PLAN, a new trace analysis technique whose objective is to classify execution transactions according to their impact on latency. To do so, we rely first on a model transformation that builds up a dependency graph from an allocation model, thus including hardware and software aspects of a system model. Then, from this graph and an execution trace, our analysis can highlight how software or hardware elements contributed to the latency violation. The paper first formalizes the problem before applying our approach to simulation traces of SysML models. A case study defined in the AQUAS European project illustrates the interest of our approach.
• Diminisher: A Linux Kernel based Countermeasure for TAA Vulnerability
  
  • Hamza Ameer
  • Mushtaq Maria
  • Khurram Bhatti Muhammad
  • Novo David
  • Bruguier Florent
  • Benoit Pascal
  , 2022, 13106, pp.477-495. TSX Asynchronous Abort (TAA) vulnerability is a class of Side-Channel Attack (SCA) that allows an application to leak data from internal CPU buffers through asynchronous Transactional Synchronization Extension (TSX) aborts that are exploited by the recent Microarchitectural Data Sampling (MDS) attacks. Cross-core TAA attacks can be prevented through microcode updates where CPU buffers are flushed during Operating System (OS) context switching, but there is no solution to our knowledge that exists for hyper-threaded TAA attacks in which the attacker leaks data from sibling hardware threads through asynchronous abort. In this work, we have proposed Diminisher, a Linux kernel-based detection and mitigation solution for both hyper-threaded and cross-core TAA attacks. Diminisher can be logically divided into three phases, i.e., scheduling, detection, and mitigation. Diminisher is a lightweight tool to prevent TAA vulnerability. The novelty lies in the methodology that we propose enabling easy extensions to cover other hyper-threaded attacks for which no satisfactory solutions exist yet. Diminisher detects and mitigates the TAA attacks around 99% of the time at a low-performance overhead of 2.5%. (10.1007/978-3-030-95484-0_28)
  DOI : 10.1007/978-3-030-95484-0_28
• Derived Terms without Derivation -- A shifted perspective on the derived-term automaton
  
  • Lombardy Sylvain
  • Sakarovitch Jacques
  Journal of Computer Science and Cybernetics, Vietnamese Academy of Science and Technology, 2021, 37 (3), pp.201-221. We present here a construction for the derived term automaton (aka partial derivative, or Antimirov, automaton) of a rational (or regular) expression based on a sole induction on the depth of the expression and without making reference to an operation of derivation of the expression. It is particularly well-suited to the case of weighted rational expressions. (10.15625/1813-9663/37/3/16263)
  DOI : 10.15625/1813-9663/37/3/16263
• Analyzing and Repairing Concept Drift Adaptation in Data Stream Classification
  
  • Halstead Ben
  • Koh Yun Sing
  • Riddle Patricia
  • Pears Russel
  • Pechenizkiy Mykola
  • Bifet Albert
  • Olivares Gustavo
  • Coulson Guy
  , 2021, pp.1--2. Data collected over time often exhibit changes in distribution, or concept drift, caused by changes in hidden context relevant to the classification task, e.g. weather conditions. Adaptive learning methods are able to retain performance in changing conditions by explicitly detecting concept drift and changing the classifier used to make predictions. However, in realworld conditions, existing methods often select classifiers which poorly represent current data due to adaptation errors, where change in context is misidentified. We propose the AiRStream system, which uses a novel repair algorithm to identify and correct adaptation errors. We identify errors by periodically testing the performance of inactive classifiers. If an error is identified, a backtracking procedure repairs training done under the misidentified context. AiRStream achieves higher accuracy compared to baseline methods and selects classifiers which better match changes in context. A case study on a real-world air quality inference task shows that AiRStream is able to build a robust model of environmental conditions, allowing the adaptions made to concept drift to be analysed and related to changes in weather. (10.1109/DSAA53316.2021.9564191)
  DOI : 10.1109/DSAA53316.2021.9564191
• Individual Survival Curves with Conditional Normalizing Flows
  
  • Clémençon Stéphan
  • Ausset Guillaume
  • Ciffreo Tom
  • Portier François
  • Papin Timothee
  , 2021, pp.1-10. Survival analysis, or time-to-event modelling, is a classical statistical problem that has garnered a lot of interest for its practical use in epidemiology, demographics or actuarial sciences. Recent advances on the subject from the point of view of machine learning have been concerned with precise per-individual predictions instead of population studies, driven by the rise of individualized medicine. We introduce here a conditional normalizing flow based estimate of the time-to-event density as a way to model highly flexible and individualized conditional survival distributions. We use a novel hierarchical formulation of normalizing flows to enable efficient fitting of flexible conditional distributions without overfitting and show how the normalizing flow formulation can be efficiently adapted to the censored setting. We experimentally validate the proposed approach on a synthetic dataset as well as four open medical datasets and an example of a common financial problem. (10.1109/DSAA53316.2021.9564222)
  DOI : 10.1109/DSAA53316.2021.9564222
• Testing and Reliability Enhancement of Security Primitives
  
  • Danger Jean-Luc
  • Hasan Anik Toufiq
  • Diankha Omar
  • Ebrahimabadi Mohammad
  • Frisch Christoph
  • Guilley Sylvain
  • Karimi Naghmeh
  • Pehl Michael
  • Takarabt Sofiane
  , 2021, pp.1-8. (10.1109/DFT52944.2021.9568297)
  DOI : 10.1109/DFT52944.2021.9568297
• A Perceptual Study of the Decoding Process of the SoftCast Wireless Video Broadcast Scheme
  
  • Trioux Anthony
  • Valenzise Giuseppe
  • Cagnazzo Marco
  • Kieffer Michel
  • Coudoux François-Xavier
  • Corlay Patrick
  • Gharbi M
  , 2021, pp.6 pages. The SoftCast scheme has been proposed as a promising alternative to traditional video broadcasting systems in wireless environments. In its current form, SoftCast performs image decoding at the receiver side by using a Linear Least Square Error (LLSE) estimator. Such approach maximizes the reconstructed quality in terms of Peak Signal-to-Noise Ratio (PSNR). However, we show that the LLSE induces an annoying blur effect at low Channel Signal-to-Noise Ratio (CSNR) quality. To cancel this artifact, we propose to replace the LLSE estimator by the Zero-Forcing (ZF) one. In order to better understand the perceived quality offered by these two estimators, a mathematical characterization as well as an objective and subjective studies are performed. Results show that the gains brought by the LLSE estimator, in terms of PSNR and Structural SIMiliraty (SSIM), are limited and quickly tend to null value as the CSNR increases. However, higher gains are obtained by the ZF estimator when considering the recent Video Multi-method Assessment Fusion (VMAF) metric proposed by Netflix, which evaluates the perceptual video quality. This result is confirmed by the subjective assessment. (10.1109/MMSP53017.2021.9733474)
  DOI : 10.1109/MMSP53017.2021.9733474
• Dynamic Graph Convolutional LSTM application for traffic flow estimation from error-prone measurements: results and transferability analysis
  
  • Clémençon Stéphan
  • Boudabous Safa
  • Labiod Houda
  • Garbiso Julian
  , 2021, pp.1-10. The technological advances in the transportation and automotive industry led to the use of new types of sensing systems more cost-effective and adapted to large-scale dense deployment. Those sensing techniques allow continuously gathering traffic measurements times series in different geospatial locations. The accuracy of the obtained raw measurements is often hindered by different factors related to the sensing environment and the sensing process itself and thus fail to capture the short-term traffic variations crucial for real-time traffic monitoring. In this paper, we propose the DGC-LSTM model for area-wide traffic estimation from error-prone measurements time series. The backbone of the DGC-LSTM model is a graph convolutional Long Short Term Memory model with a dynamic adjacency matrix. The adjacency matrix is learned and optimized during the model training. The adjacency matrix values are estimated from the set of contextual features that impact the dynamicity of the dependencies in both the spatial and temporal dimensions. Experiments on a realistic synthetic labelled Bluetooth counts dataset is used for model evaluation. Lastly, we highlight the importance of transfer learning methods to improve the model applicability by ensuring model adaptation to the new deployment site while avoiding the extensive data-labelling effort. (10.1109/DSAA53316.2021.9564245)
  DOI : 10.1109/DSAA53316.2021.9564245
• Performances 5G: étude comparée en zones rurales et urbaines
  
  • Coupechoux Marceau
  , 2021. Ce rapport propose une méthodologie permettant de comparer les débits moyens obtenus par les utilisateurs 5G en zone rurale et en zone urbaine. Ces environnements sont caractérisés par des conditions de propagation radio et des choix de paramétrage de la technologie 5G spécifiques à chacune des zones. Par exemple, en zone rurale, la propagation radio est beaucoup plus favorable qu'en zone urbaine, les fréquences 5G utilisées sont plus basses, les largeurs de bandes sont plus petites et la charge est généralement plus faible. L'hypothèse principale est que la bande 700MHz est utilisée en zone rurale avec des largeurs de bande de 10 à 30 MHz et que la bande 3.5GHz est utilisée en zone urbaine avec des largeurs de bandes de 50 ou 100 MHz.
• Reasoning with Transformer-based Models: Deep Learning, but Shallow Reasoning
  
  • Helwe Chadi
  • Clavel Chloé
  • Suchanek Fabian
  , 2021. Recent years have seen impressive performance of transformer-based models on different natural language processing tasks. However, it is not clear to what degree the transformers can reason on natural language. To shed light on this question, this survey paper discusses the performance of transformers on different reasoning tasks, including mathematical reasoning, commonsense reasoning, and logical reasoning. We point out successes and limitations, of both empirical and theoretical nature.
• Permissionless and Asynchronous Asset Transfer
  
  • Kuznetsov Petr
  • Pignolet Yvonne-Anne
  • Ponomarev Pavel
  • Tonkikh Andrei
  , 2021. Most modern asset transfer systems use consensus to maintain a totally ordered chain of transactions. It was recently shown that consensus is not always necessary for implementing asset transfer. More efficient, asynchronous solutions can be built using reliable broadcast instead of consensus. This approach has been originally used in the closed (permissioned) setting. In this paper, we extend it to the open (permissionless) environment. We present Pastro, a permissionless and asynchronous asset-transfer implementation, in which quorum systems, traditionally used in reliable broadcast, are replaced with a weighted Proof-of-Stake mechanism. Pastro tolerates a dynamic adversary that is able to adaptively corrupt participants based on the assets owned by them. (10.4230/LIPIcs.DISC.2021.28)
  DOI : 10.4230/LIPIcs.DISC.2021.28
• Anomalous Cluster Detection in Large Networks with Diffusion-Percolation Testing
  
  • Larroche Corentin
  • Mazel Johan
  • Clémençon Stephan
  , 2021, pp.1-6. We propose a computationally efficient procedure for elevated mean detection on a connected subgraph of a network with node-related scalar observations. Our approach relies on two intuitions: first, a significant concentration of high observations in a connected subgraph implies that the subgraph induced by the nodes associated with the highest observations has a large connected component. Secondly, a greater detection power can be obtained in certain cases by denoising the observations using the network structure. Numerical experiments show that our procedure's detection performance and computational efficiency are both competitive.
• Aquadiff
  
  • Zayana Karim
  • Bernard Jean-Noël
  • Boyer Ivan
  • Rabiet Victor
  CultureMath, ENS, 2021. Quand l'eau se fait source d'équations différentielles... Au lycée comme à l'université, il est classique d'introduire les équations différentielles en invoquant la charge d'un condensateur en électricité, l'oscillation du pendule en mécanique, la décroissance radioactive ou encore la cinétique d'une réaction chimique. Pour motivantes qu'elles soient, ces situations nécessitent un bagage scientifique solide doublé d'un réel sens physique. Plus simple et quotidien, le problème dit de la vidange ou de la purge d'un récipient conduit naturellement à des modèles souvent du premier ordre et accessibles dès la classe de Terminale. Les contextes d'application sont variés : réservoir sans couvercle type baignoire ou avec couvercle type bidon / cubi / cuve ; réservoir sous pression type fusée à eau, spray aérosol ou bonbonne de gaz, ou soumis à une force extérieure type vessie ou bouée dégonflée par constriction.
• Investigations on c -(Almost) Perfect Nonlinear Functions
  
  • Mesnager Sihem
  • Riera Constanza
  • Stanica Pantelimon
  • Yan Haode
  • Zhou Zhengchun
  IEEE Transactions on Information Theory, Institute of Electrical and Electronics Engineers, 2021, 67 (10), pp.6916-6925. (10.1109/TIT.2021.3081348)
  DOI : 10.1109/TIT.2021.3081348
• Network slicing with load-balancing for task offloading in vehicular edge computing
  
  • Hejja Khaled
  • Berri Sara
  • Labiod Houda
  Vehicular Communications, Elsevier, 2021, pp.100419. (10.1016/j.vehcom.2021.100419)
  DOI : 10.1016/j.vehcom.2021.100419
• IEEE 802.11p performance enhancement based on Markov chain and neural networks for safety applications
  
  • Chbib Fadlallah
  • Fahs Walid
  • Haydar Jamal
  • Khoukhi Lyes
  • Khatoun Rida
  Annals of Telecommunications - annales des télécommunications, Springer, 2021, 76 (9-10), pp.617-632. Vehicular communication is recently considered as one of the key future technology to improve the safety of vehicles, the efficiency of traffic and the comfort for both drivers and pedestrians. Vehicular communications, based on IEEE 802.11p, use the Enhanced Distributed Channel Access (EDCA) algorithm to support different levels of Quality of Service (QoS). In this paper, a machine learning neural network with Markov chain approach is proposed to ensure the delivery of urgent safety messages to the receiver whatever the situation of the network. We propose to control the rate of periodic messages in Control Channel (CCH), by modifying the back-off parameters according to the state of the buffer. We also use Radial Basis Function Neural Network (RBFNN) to adjust the EDCA back-off parameters, using the following parameters: the priority of message (P), the sensitivity of road (S), the threshold of buffer (T), and the type of vehicle (V). Our simulation is done using SUMO 0.22 simulator, NS 2.34 and awk scripts; the simulation was applied on Hamra area (Lebanon). The results show that our proposed models perform better compared to the IEEE 802.11p in terms of packet delivery ratio, throughput and end-to-end delay. (10.1007/s12243-021-00846-y)
  DOI : 10.1007/s12243-021-00846-y
• Brief Announcement: Accountability and Reconfiguration -Self-Healing Lattice Agreement
  
  • Freitas de Souza Luciano
  • Kuznetsov Petr
  • Rieutord Thibault
  • Tucci-Piergiovanni Sara
  , 2021. An accountable distributed system provides means to detect deviations of system components from their expected behavior. It is natural to complement fault detection with a reconfiguration mechanism, so that the system could heal itself, by replacing malfunctioning parts with new ones. In this paper, we describe a framework that can be used to implement a large class of accountable and reconfigurable replicated services. We build atop the fundamental lattice agreement abstraction lying at the core of storage systems and cryptocurrencies. Our asynchronous implementation of accountable lattice agreement ensures that every violation of consistency is followed by an undeniable evidence of misbehavior of a faulty replica. The system can then be seamlessly reconfigured by evicting faulty replicas, adding new ones and merging inconsistent states. We believe that this paper opens a direction towards asynchronous "self-healing" systems that combine accountability and reconfiguration. (10.4230/LIPIcs.DISC.2021.54)
  DOI : 10.4230/LIPIcs.DISC.2021.54
• On Hulls of Some Primitive BCH Codes and Self-Orthogonal Codes
  
  • Gan Chunyu
  • Li Chengju
  • Mesnager Sihem
  • Qian Haifeng
  IEEE Transactions on Information Theory, Institute of Electrical and Electronics Engineers, 2021, 67 (10), pp.6442-6455. Self-orthogonal codes are an important type of linear codes due to their wide applications in communication and cryptography. The Euclidean (or Hermitian) hull of a linear code is defined to be the intersection of the code and its Euclidean (or Hermitian) dual. It is clear that the hull is self-orthogonal. The main goal of this paper is to obtain self-orthogonal codes by investigating the hulls. Let C(r,rm−1,δ,b) be the primitive BCH code over Fr of length rm−1 with designed distance δ , where Fr is the finite field of order r . In this paper, we will present Euclidean (or Hermitian) self-orthogonal codes and determine their parameters by investigating the Euclidean (or Hermitian) hulls of some primitive BCH codes. Several sufficient and necessary conditions for primitive BCH codes with large Hermitian hulls are developed by presenting lower and upper bounds on their designed distances. Furthermore, some Hermitian self-orthogonal codes are proposed via the hulls of BCH codes and their parameters are also investigated. In addition, we determine the dimensions of the code C(r,r2−1,δ,1) and its hull in both Hermitian and Euclidean cases for 2≤δ≤r2−1 . We also present two sufficient and necessary conditions on designed distances such that the hull has the largest dimension. (10.1109/TIT.2021.3076878)
  DOI : 10.1109/TIT.2021.3076878