Publications

2021

• Multi-source Fault Injection Detection Using Machine Learning and Sensor Fusion
  
  • Shrivastwa Ritu-Ranjan
  • Guilley Sylvain
  • Danger Jean-Luc
  , 2021, 1497, pp.93-107. (10.1007/978-3-030-90553-8_7)
  DOI : 10.1007/978-3-030-90553-8_7
• Side-channel Analysis of CRYSTALS-Kyber and A Novel Low-Cost Countermeasure
  
  • Hamoudi Meziane
  • Bel Korchi Amina
  • Guilley Sylvain
  • Takarabt Sofiane
  • Karray Khaled
  • Souissi Youssef
  , 2021, 1497, pp.30-46. In this paper, we propose a vertical side-channel leakage detection on the decryption function of the third round implementation of CPA-secure public-key encryption scheme underlying CRYSTALS-Kyber, a lattice-based key encapsulation mechanism, which is a candidate to the NIST Post-Quantum Cryptography standardization project. Using a leakage assessment metric, we show that the side-channel information can be efficiently used to pinpoint operations leaking the secret variable and how masking countermeasures can be applied. We detect leakages in the polynomial multiplication between the secret key and the ciphertext. We propose and evaluate two different masking countermeasures, based on additive and multiplicative masking. To the best of our knowledge, the multiplicative masking has not been proposed before as a countermeasure to CRYSTALS-Kyber vulnerabilities. We demonstrate their efficiency and discuss their impact in terms of performance. Our work is beneficial to assess and enhance the security of Post-Quantum Cryptography against advanced vertical side-channel analysis. (10.1007/978-3-030-90553-8_3)
  DOI : 10.1007/978-3-030-90553-8_3
• Hybrid MTJ-CMOS Integration for Sigma-Delta ADC
  
  • Wu Yu-Ang
  • Naviner Lirida
  • Cai Hao
  , 2021, pp.1-5. Previous theoretical and experimental works revealed the novel factors that Magnetic tunnel junction (MTJ) can be integrated into novel hybrid circuits except for memory applications. This paper exploits hybrid CMOS-MTJ circuit to diminish layout penalty of on-chip passive component in sigma-delta analog-to-digital converter (SD-ADC). Large poly resistance can be replaced by series connected MTJs with magneto-resistance. Simulation results show that MTJ based resistor-capacitor (RC) integrator can greatly reduce the resistance layout area by 94.52% comparing with 28-nm fully CMOS design. This novel design improves other ADC Figures of merit (FoM), including 0.06 bit Effective Numbers of Bits (ENOB) increment and stable performance under temperature variations. The design trade-off is 1.1 dB reduction of Signal-to-Noise Ratio (SNR). (10.1109/NANOARCH53687.2021.9642236)
  DOI : 10.1109/NANOARCH53687.2021.9642236
• THE WORDS REMAIN THE SAME: COVER DETECTION WITH LYRICS TRANSCRIPTION
  
  • Vaglio Andrea
  • Hennequin Romain
  • Moussallam Manuel
  • Richard Gael
  , 2021. Cover detection has gained sustained interest in the scientific community and has recently made significant progress both in terms of scalability and accuracy. However, most approaches are based on the estimation of harmonic and melodic features and neglect lyrics information although it is an important invariant across covers. In this work, we propose a novel approach leveraging lyrics without requiring access to full texts though the use of lyrics recognition on audio. Our approach relies on the fusion of a singing voice recognition framework and a more classic tonal-based cover detection method. To the best of our knowledge, this is the first time that lyrics estimation from audio has been explicitly used for cover detection. Furthermore, we exploit efficient string matching and an approximated nearest neighbors search algorithm which lead to a scalable system which is able to operate on very large databases. Extensive experiments on the largest publicly available cover detection dataset demonstrate the validity of using lyrics information for this task.
• Is There a "Language of Music-Video Clips" ? A Qualitative and Quantitative Study
  
  • Prétet Laure
  • Richard Gaël
  • Peeters Geoffroy
  , 2021. Recommending automatically a video given a music or a music given a video has become an important asset for the audiovisual industry-with user-generated or professional content. While both music and video have specific temporal organizations, most current works do not consider those and only focus on globally recommending a media. As a first step toward the improvement of these recommendation systems, we study in this paper the relationship between music and video temporal organization. We do this for the case of official music videos, with a quantitative and a qualitative approach. Our assumption is that the movement in the music are correlated to the ones in the video. To validate this, we first interview a set of internationally recognized music video experts. We then perform a largescale analysis of official music-video clips (which we manually annotated into video genres) using MIR description tools (downbeats and functional segments estimation) and Computer Vision tools (shot detection). Our study confirms that a "language of music-video clips" exists; i.e. editors favor the co-occurrence of music and video events using strategies such as anticipation. It also highlights that the amount of co-occurrence depends on the music and video genres.
• Training Deep Pitch-Class Representations With a Multi-Label CTC Loss
  
  • Weiss Christof
  • Peeters Geoffroy
  , 2021. Despite the success of end-to-end approaches, chroma (or pitch-class) features remain a useful mid-level representation of music audio recordings due to their direct interpretability. Since traditional chroma variants obtained with signal processing suffer from timbral artifacts such as overtones or vibrato, they do not directly reflect the pitch classes notated in the score. For this reason, training a chroma representation using deep learning ("deep chroma") has become an interesting strategy. Existing approaches involve the use of supervised learning with strongly aligned labels for which, however, only few datasets are available. Recently, the Connectionist Temporal Classification (CTC) loss, initially proposed for speech, has been adopted to learn monophonic (single-label) pitch-class features using weakly aligned labels based on corresponding score--audio segment pairs. To exploit this strategy for the polyphonic case, we propose the use of a multi-label variant of this CTC loss, the MCTC, and formalize this loss for the pitch-class scenario. Our experiments demonstrate that the weakly aligned approach achieves almost equivalent pitch-class estimates than training with strongly aligned annotations. We then study the sensitivity of our approach to segment duration and mismatch. Finally, we compare the learned features with other pitch-class representations and demonstrate their use for chord and local key recognition on classical music datasets.
• DARKGAN: EXPLOITING KNOWLEDGE DISTILLATION FOR COMPREHENSIBLE AUDIO SYNTHESIS WITH GANS
  
  • Nistal Hurlé Javier
  • Lattner Stefan
  • Richard Gael
  , 2021. Generative Adversarial Networks (GANs) have achieved excellent audio synthesis quality in the last years. However, making them operable with semantically meaningful controls remains an open challenge. An obvious approach is to control the GAN by conditioning it on metadata contained in audio datasets. Unfortunately, audio datasets often lack the desired annotations, especially in the musical domain. A way to circumvent this lack of annotations is to generate them, for example, with an automatic audio tagging system. The output probabilities of such systems (so-called "soft labels") carry rich information about the characteristics of the respective audios and can be used to distill the knowledge from a teacher model into a student model. In this work, we perform knowledge distillation from a large audio tagging system into an adversarial audio synthesizer that we call DarkGAN. Results show that DarkGAN can synthesize musical audio with acceptable quality and exhibits moderate attribute control even with out-of-distribution input conditioning. We release the code and provide audio examples on the accompanying website.
• Automatic Text Evaluation through the Lens of Wasserstein Barycenters
  
  • Colombo Pierre
  • Staerman Guillaume
  • Piantanida Pablo
  • Clavel Chloé
  , 2021, pp.10450–10466. A new metric BaryScore to evaluate text generation based on deep contextualized embeddings (e.g., BERT, Roberta, ELMo) is introduced. This metric is motivated by a new framework relying on optimal transport tools, i.e., Wasserstein distance and barycenter. By modelling the layer output of deep contextualized embeddings as a probability distribution rather than by a vector embedding; this framework provides a natural way to aggregate the different outputs through the Wasserstein space topology. In addition, it provides theoretical grounds to our metric and offers an alternative to available solutions (e.g., MoverScore and BertScore). Numerical evaluation is performed on four different tasks: machine translation, summarization, data2text generation and image captioning. Our results show that BaryScore outperforms other BERT based metrics and exhibits more consistent behaviour in particular for text summarization. (10.18653/v1/2021.emnlp-main.817)
  DOI : 10.18653/v1/2021.emnlp-main.817
• Few-Shot Emotion Recognition in Conversation with Sequential Prototypical Networks
  
  • Guibon Gaël
  • Labeau Matthieu
  • Flamein Hélène
  • Lefeuvre Luce
  • Clavel Chloé
  , 2021. Several recent studies on dyadic human-human interactions have been done on conversations without specific business objectives. However, many companies might benefit from studies dedicated to more precise environments such as after sales services or customer satisfaction surveys. In this work, we place ourselves in the scope of a live chat customer service in which we want to detect emotions and their evolution in the conversation flow. This context leads to multiple challenges that range from exploiting restricted, small and mostly unlabeled datasets to finding and adapting methods for such context. We tackle these challenges by using Few-Shot Learning while making the hypothesis it can serve conversational emotion classification for different languages and sparse labels. We contribute by proposing a variation of Prototypical Networks for sequence labeling in conversation that we name ProtoSeq. We test this method on two datasets with different languages: daily conversations in English and customer service chat conversations in French. When applied to emotion classification in conversations, our method proved to be competitive even when compared to other ones.
• Code-switched inspired losses for generic spoken dialog representations
  
  • Chapuis Emile
  • Colombo Pierre
  • Labeau Matthieu
  • Clavel Chloé
  Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 2021. Spoken dialog systems need to be able to handle both multiple languages and multilinguality inside a conversation (e.g in case of codeswitching). In this work, we introduce new pretraining losses tailored to learn multilingual spoken dialog representations. The goal of these losses is to expose the model to codeswitched language. To scale up training, we automatically build a pretraining corpus composed of multilingual conversations in five different languages (French, Italian, English, German and Spanish) from OpenSubtitles, a huge multilingual corpus composed of 24.3G tokens. We test the generic representations on MIAM, a new benchmark composed of five dialog act corpora on the same aforementioned languages as well as on two novel multilingual downstream tasks (i.e multilingual mask utterance retrieval and multilingual inconsistency identification). Our experiments show that our new code switched-inspired losses achieve a better performance in both monolingual and multilingual settings.
• ADQDA: A Cross-device Affinity Diagramming Tool for Fluid and Holistic Qualitative Data Analysis
  
  • Liu Jiali
  • Eagan James R
  Proceedings of the ACM on Human-Computer Interaction, Association for Computing Machinery (ACM), 2021, 5 (ISS), pp.19. Affinity diagramming is widely applied to analyze qualitative data such as interview transcripts. It involves multiple analytic processes and is often performed collaboratively. Drawing on interviews with three practitioners and upon our own experience, we show how practitioners combine multiple analytic processes and adopt different artifacts to help them analyze their data. Current tools, however, fail to adequately support mixing analytic processes, devices, and collaboration styles. We present a vision and prototype ADQDA, a cross-device, collaborative affinity diagramming tool for qualitative data analysis, implemented using distributed web technologies. We show how this approach enables analysts to appropriate available pertinent digital devices as they fluidly migrate between analytic phases or adopt different methods and representations, all while preserving consistent analysis artifacts. We validate this approach through a set of application scenarios that explore how it enables new ways of analyzing qualitative data that better align with identified analytic practices. CCS Concepts: • Human-centered computing → Interactive systems and tools; • Information systems → Collaborative and social computing systems and tools. (10.1145/3488534)
  DOI : 10.1145/3488534
• Checking SysML Models Against Safety and Security Properties
  
  • de Saqui-Sannes Pierre
  • Apvrille Ludovic
  • Vingerhoeds Rob
  Journal of Aerospace Information Systems, American Institute of Aeronautics and Astronautics, 2021, pp.1 - 13. Systems engineering, or engineering in general, has long been relying on document-centric approaches. Switching to model-based systems engineering, or MBSE for short, has extensively been discussed over the past three decades. Since about two decades, MBSE has been commonly associated with the modeling language SysML (Systems Modeling Language), which offers a standardized notation, not a methodology of using it. SysML needs therefore to be associated with a methodology supported by tools. In this paper, a methodology supported by the free and open-source software TTool is associated with SysML. This paper focuses discussion on methodological issues, leading the authors to share their experience in real-time systems modeling. Modeling with SysML is more than just drawing the different diagrams. Associated tools offer possibilities to analyze SysML models for specific properties. In this paper, verification addresses both safety and security properties. The TTool model checker inputs the SysML model enriched with safety properties to be verified and outputs an yes/no answer for each property. Security verification checks SysML models against confidentiality, integrity, and authenticity properties. As an illustration of the proposed approach, an aircraft cockpit door control system is modeled in SysML and verified against safety and security properties. (10.2514/1.i010950)
  DOI : 10.2514/1.i010950
• Improving the performance of bagging ensembles for data streams through mini-batching
  
  • Cassales Guilherme Weigert
  • Gomes Heitor Murilo
  • Bifet Albert
  • Pfahringer Bernhard
  • Senger Hermes
  Information Sciences, Elsevier, 2021, 580, pp.260--282. Often, machine learning applications have to cope with dynamic environments where data are collected in the form of continuous data streams with potentially infinite length and transient behavior. Compared to traditional (batch) data mining, stream processing algorithms have additional requirements regarding computational resources and adaptability to data evolution. They must process instances incrementally because the data’s continuous flow prohibits storing data for multiple passes. Ensemble learning achieved remarkable predictive performance in this scenario. Implemented as a set of (several) individual classifiers, ensembles are naturally amendable for task parallelism. However, the incremental learning and dynamic data structures used to capture the concept drift increase the cache misses and hinder the benefit of parallelism. This paper proposes a mini-batching strategy that can improve memory access locality and performance of several ensemble algorithms for stream mining in multi-core environments. With the aid of a formal framework, we demonstrate that mini-batching can significantly decrease the reuse distance (and the number of cache misses). Experiments on six different state-of-the-art ensemble algorithms applying four benchmark datasets with varied characteristics show speedups of up to 5X on 8-core processors. These benefits come at the expense of a small reduction in predictive performance. (10.1016/J.INS.2021.08.085)
  DOI : 10.1016/J.INS.2021.08.085
• La récurrence chocolat
  
  • Zayana Karim
  • Boyer Ivan
  • Rabiet Victor
  CultureMath, ENS, 2021. Mathématiques et récurrence au service d'une petite expérience gourmande... Enjoy!
• Quantization-aware Processing for Massive MIMO Uplink Could RAN
  
  • Askri Aymen
  • Zhang Chao
  • Rekaya-Ben Othman Ghaya
  IEEE Communications Letters, Institute of Electrical and Electronics Engineers, 2021.
• Improving Multimodal fusion via Mutual Dependency Maximisation
  
  • Colombo Pierre
  • Chapuis Emile
  • Labeau Matthieu
  • Clavel Chloé
  , 2021, pp.231-245. Multimodal sentiment analysis is a trending area of research, and multimodal fusion is one of its most active topic. Acknowledging humans communicate through a variety of channels (i.e visual, acoustic, linguistic), multimodal systems aim at integrating different unimodal representations into a synthetic one. So far, a consequent effort has been made on developing complex architectures allowing the fusion of these modalities. However, such systems are mainly trained by minimising simple losses such as L1 or cross-entropy. In this work, we investigate unexplored penalties and propose a set of new objectives that measure the dependency between modalities. We demonstrate that our new penalties lead to a consistent improvement (up to 4.3 on accuracy) across a large variety of state-of-the-art models on two well-known sentiment analysis datasets: CMU-MOSI and CMU-MOSEI. Our method not only achieves a new SOTA on both datasets but also produces representations that are more robust to modality drops. Finally, a by-product of our methods includes a statistical network which can be used to interpret the high dimensional representations learnt by the model. (10.18653/v1/2021.emnlp-main.21)
  DOI : 10.18653/v1/2021.emnlp-main.21
• Experimental estimation of power loss amplitudes in optical fibers
  
  • Ciblat Philippe
  IEEE Photonics Technology Letters, Institute of Electrical and Electronics Engineers, 2021. We propose a method to estimate the amplitude of an unexpected power loss which, leveraging on a calibration, enables the real-time monitoring of a network link. It is based on an existing fiber-longitudinal power profile evaluation technique. The reliability of the method is assessed experimentally. When the anomaly is located at 0 km from the beginning of the span, the estimation bias is smaller than 0.2 dB for losses up to 10 dB. When the anomaly is located at 25 km from the beginning of the span, the same estimation bias is observed but for losses up to 5 dB. In both cases, the standard deviation of the estimation is smaller than 0.2 dB. (10.1109/LPT.2021.3115627)
  DOI : 10.1109/LPT.2021.3115627
• On constructions of weightwise perfectly balanced Boolean functions
  
  • Mesnager Sihem
  • Su Sihong
  Cryptography and Communications - Discrete Structures, Boolean Functions and Sequences, Springer, 2021, 13 (6), pp.951-979. (10.1007/s12095-021-00481-3)
  DOI : 10.1007/s12095-021-00481-3
• L'évolution de la biodiversité génétique : le principe de Hardy-Weinberg
  
  • Zayana Karim
  • Rabiet Victor
  CultureMath, ENS, 2021. Ce texte revient sur le principe d'équilibre de Hardy-Weinberg : en l'absence de forces évolutives et pour peu que le hasard joue correctement son rôle, la structure génotypique d'une grande population de même espèce se stabilise dès la première descendance. Nous en réaliserons la démonstration formelle en l'accordant avec le langage des probabilités vu au lycée. Nous raisonnerons avec un nombre p d'allèles quelconque plutôt que de nous limiter au seul cas où p=2. Plus réaliste, ce choix nous permettra, en outre, de suivre plus aisément le fil de l'exposé. Interviennent en particulier la notion de probabilité conditionnelle, le concept d'indépendance et la formule des probabilités totales.
• Multi-resource scheduling for FPGA systems
  
  • Pacalet Renaud
  • Bertolino Matteo
  • Apvrille Ludovic
  • Enrici Andrea
  Microprocessors and Microsystems: Embedded Hardware Design, Elsevier, 2021, 87, pp.104373. In modern cloud data centers, reconfigurable devices (FPGAs) are used as an alternative to Graphics Processing Units to accelerate data-intensive computations (e.g., machine learning, image and signal processing). Currently, FPGAs are configured to execute fixed workloads, repeatedly over long periods of time. This conflicts with the needs, proper to cloud computing, to flexibly allocate different workloads and to offer the use of physical devices to multiple users. This raises the need for novel, efficient FPGA scheduling algorithms that can decide execution orders close to the optimum in a short time. In this context, we propose a novel scheduling heuristic where groups of tasks that execute together are interposed by hardware reconfigurations. Our contribution is based on gathering tasks around a high-latency task that hides the latency of tasks, within the same group, that run in parallel and have shorter latencies. We evaluated our solution on a benchmark of 37500 random workloads, synthesized from realistic designs (i.e., topology, resource occupancy). For this testbench, on average, our heuristic produces optimum makespan solutions in 47.4% of the cases. It produces acceptable solutions for moderately constrained systems (i.e., the deadline falls within 10% of the optimum makespan) in 90.1% of the cases. (10.1016/j.micpro.2021.104373)
  DOI : 10.1016/j.micpro.2021.104373
• On Correlation Immune Boolean Functions With Minimum Hamming Weight Power of 2
  
  • Mesnager Sihem
  • Su Sihong
  IEEE Transactions on Information Theory, Institute of Electrical and Electronics Engineers, 2021, 67 (11), pp.7501-7517. (10.1109/TIT.2021.3109946)
  DOI : 10.1109/TIT.2021.3109946
• SOME/IP Intrusion Detection using Deep Learning-based Sequential Models in Automotive Ethernet Networks
  
  • Alkhatib Natasha
  • Danger Jean-Luc
  • Ghauch Hadi
  , 2021. Intrusion Detection Systems are widely used to detect cyberattacks, especially on protocols vulnerable to hacking attacks such as SOME/IP. In this paper, we present a deep learning-based sequential model for offline intrusion detection on SOME/IP application layer protocol. To assess our intrusion detection system, we have generated and labeled a dataset 1 with several classes representing realistic intrusions, and a normal class-a significant contribution due to the absence of such publicly available datasets. Furthermore, we also propose a recurrent neural network (RNN), as an instance of deep learningbased sequential model, that we apply to our generated dataset. The numerical results show that RNN excel at predicting invehicle intrusions, with F1 Scores and AUC values greater than 0.8 depending on each intrusion type.
• Extreme Value Theory and Machine Learning
  
  • Sabourin Anne
  , 2021.
• Ultra-low Power Access Strategy for Process-Voltage-Temperature Aware STT-MRAM
  
  • Zhang You-You
  • Naviner Lirida
  • Cai Hao
  , 2021, pp.1-4. With the development of circuit integration, low power consumption design has become the design challenge of on-chip memory. This work focuses on ultra-low power access strategy for STT-MRAM. A high margin voltage sensing amplifier (VSA) is implemented based on the bit-line (BL) parasitic capacitance, whereas a pulse-detect write self-termination is included for MRAM writing. Simulation is performed based on 28-nm CMOS and 40-nm CD magnetic tunnel junction (MTJ). Monte Carlo simulation show that the proposed sensing circuit achieves a reading yield of over 98% as well as 38% energy saving compared to previous work. Meanwhile, the self-termination scheme achieves an energy saving for more than 80%. These MRAM access strategies is well-adapted to process-voltage-temperature (PVT) variations including Tunnel Magneto Resistance (TMR) (20%-200%), temperature (0℃-120℃) and supply voltage (0.6V-1.8V). (10.1109/ASICON52560.2021.9620529)
  DOI : 10.1109/ASICON52560.2021.9620529
• Network-wide intrusion detection through statistical analysis of event logs : an interaction-centric approach
  
  • Larroche Corentin
  , 2021. Event logs are structured records of all kinds of activities taking place in a computer network. In particular, malicious actions taken by intruders are likely to leave a trace in the logs, making this data source useful for security monitoring and intrusion detection. However, the considerable volume of real-world event logs makes them difficult to analyze. This limitation has motivated a fair amount of research on malicious behavior detection through statistical methods. This thesis addresses some of the challenges that currently hinder the use of this approach in realistic settings. First of all, building an abstract representation of the data is nontrivial: event logs are complex and multi-faceted, making it difficult to capture all the relevant information they contain in a simple mathematical object. We take an interaction-centric approach to event log representation, motivated by the intuition that malicious events can often be seen as unexpected interactions between entities (users, hosts, etc.). While this representation preserves critical information, it also makes statistical modelling difficult. We thus build an ad hoc model and design a suitable inference procedure, using elements of latent space modelling, Bayesian filtering and multi-task learning.Another key challenge in event log analysis is that benign events account for a vast majority of the data, including a lot of unusual albeit legitimate events. Detecting individually anomalous events is thus not enough, and we also deal with spotting clusters of potentially malicious events. To that end, we leverage the concept of event graph and recast event-wise anomaly scores as a noisy graph-structured signal. This allows us to use graph signal processing tools to improve anomaly scores provided by statistical models.Finally, we propose scalable methods for anomalous cluster detection in node-valued signals defined over large graphs.