Publications

2018

• Network Coding with Random Packet-Index Assignment for Data Collection Networks
  
  • Adjih Cédric
  • Kieffer Michel
  • Greco Claudio
  , 2018. This paper considers data collection using a network of uncoordinated, heterogeneous, and possibly mobile devices. Using medium and short-range radio technologies, multi-hop communication is required to deliver data to some sink. While numerous techniques from managed networks can be adapted, one of the most efficient (from the energy and spectrum use perspective) is network coding (NC). NC is well suited to networks with mobility and unreliability, however, practical NC requires a precise identification of individual packets that have been mixed together. In a purely decentralized system, this requires either conveying identifiers in headers along with coded information as in COPE, or integrating a more complex protocol in order to efficiently identify the sources (participants) and their payloads. A novel solution, Network Coding with Random Packet Index Assignment (NeCoRPIA), is presented where packet indices in NC headers are selected in a decentralized way, by choosing them randomly. Traditional network decoding can be applied when all original packets have different indices. When this is not the case, \emph{i.e.}, in case of collisions of indices, a specific decoding algorithm is proposed. A theoretical analysis of its performance in terms of complexity and decoding error probability is described. Simulation results match well the theoretical results. Comparisons of NeCoRPIA header lengths with those of a COPE-based NC protocol are also provided.
• Protecting data confidentiality combining data fragmentation, encryption, and dispersal over a distributed environment.
  
  • Kapusta Katarzyna
  , 2018. This thesis dissertation revisits state-of-the-art fragmentation techniques making them faster and cost-efficient. The main focus is put on increasing data confidentiality without deteriorating the processing performance. The ultimate goal is to provide a user with a set of fast fragmentation methods that could be directly applied inside an industrial context to reinforce the confidentiality of the stored data and/or accelerate the fragmentation processing. First, a rich survey on fragmentation as a way of preserving data confidentiality is presented. Second, the family of all-or-nothing transforms is extended with three new proposals. They all aim at protecting encrypted and fragmented data against the exposure of the encryption key but are designed to be employed in three different contexts: for data fragmentation in a multi-cloud environment, a distributed storage system, and an environment composed of one storage provider and one private device. Third, a way of accelerating fragmentation is presented that achieves better performance than data encryption using the most common symmetric-key encryption algorithm. Fourth, a lightweight fragmentation scheme based on data encoding, permuting, and dispersing is introduced. It totally gets rid of data encryption allowing the fragmentation to be performed even faster; up to twice as fast as data encryption. Finally, fragmentation inside sensor networks is revisited, particularly in the Unattended Wireless Sensor Networks. The main focus in this case is put not solely on the fragmentation performance, but also on the reduction of storage and transmission costs by using data aggregation.
• Protecting data confidentiality combining data fragmentation, encryption, and dispersal over a distributed environment
  
  • Kapusta Katarzyna
  , 2018. This thesis dissertation revisits state-of-the-art fragmentation techniques making them faster and cost-efficient. The main focus is put on increasing data confidentiality without deteriorating the processing performance. The ultimate goal is to provide a user with a set of fast fragmentation methods that could be directly applied inside an industrial context to reinforce the confidentiality of the stored data and/or accelerate the fragmentation processing. First, a rich survey on fragmentation as a way of preserving data confidentiality is presented. Second, the family of all-or-nothing transforms is extended with three new proposals. They all aim at protecting encrypted and fragmented data against the exposure of the encryption key but are designed to be employed in three different contexts: for data fragmentation in a multi-cloud environment, a distributed storage system, and an environment composed of one storage provider and one private device. Third, a way of accelerating fragmentation is presented that achieves better performance than data encryption using the most common symmetric-key encryption algorithm. Fourth, a lightweight fragmentation scheme based on data encoding, permuting, and dispersing is introduced. It totally gets rid of data encryption allowing the fragmentation to be performed even faster; up to twice as fast as data encryption. Finally, fragmentation inside sensor networks is revisited, particularly in the Unattended Wireless Sensor Networks. The main focus in this case is put not solely on the fragmentation performance, but also on the reduction of storage and transmission costs by using data aggregation.
• Modeling the speed-accuracy tradeoff using the tools of information theory
  
  • Gori Julien
  , 2018. Fitts’ law, which relates movement time MTin a pointing task to the target’s dimensions D and Wis usually expressed by mimicking Shannon’s capacityformula MT = a + b log 2 (1 + D/W). Yet, the currentlyreceived analysis is incomplete and unsatisfactory: itstems from a vague analogy and there is no explicitcommunication model for pointing.I first develop a transmission model for pointing taskswhere the index of difficulty ID = log 2 (1 + D/W) isthe expression of both a source entropy and a chan-nel capacity, thereby reconciling Shannon’s informa-tion theory with Fitts’ law. This model is then levera-ged to analyze pointing data gathered from controlledexperiments but also from field studies.I then develop a second model which builds on thevariability of human movements and accounts for thetremendous diversity displayed by movement control:with of without feedback, intermittent or continuous.From a chronometry of the positional variance, eva-luated from a set of trajectories, it is observed thatmovement can be separated into two phases: a firstwhere the variance increases over time and wheremost of the distance to the target is covered, follo-wed by a second phase where the variance decreasesuntil it satisfies accuracy constraints. During this se-cond phase, the problem of aiming can be reduced toa Shannon-like communication problem where infor-mation is transmitted from a “source” (variance at theend of the first phase), to a “destination” (the limb ex-tremity) over a “channel” perturbed by Gaussian noisewith a feedback link. I show that the optimal solution tothis transmission problem amounts to a scheme firstsuggested by Elias. I show that the variance can de-crease at best exponentially during the second phase,and that this result induces Fitts’ law.
• Multimodal Music Recording Remastering
  
  • Cantisani Giorgia
  • Essid Slim
  • Richard Gael
  , 2018. Our aim is to develop methods for a user-centered remastering of music performance recordings for giving the user an interactive multimedia experience. The idea is to guide audio source separation/enhancement using the user's attention as a high-level control/feedback to select which, for him/her, is the desired source to enhance. In the case of music performances, the source to enhance is represented by a particular instrument in the ensemble, thus we have a polyphonic music source separation problem.
• Background reconstruction from multiple images
  
  • Yang Xiaoyi
  , 2018. The general topic of this thesis is to reconstruct the background scene from a burst of images in presence of masks. We focus on the background detection methods as well as on solutions to geometric and chromatic distortions introduced during ph-otography. A series of process is proposed, which con­sists of geometric alignment, chromatic adjustment, image fusion and defect correction.We consider the case where the background scene is a flat surface. The geometric align­ment between a reference image and any other images in the sequence, depends on the computation of a homography followed by a bilinear interpolation.The chromatic adjustment aims to attach a similar contrast to the scene in different im­ages. We propose to model the chromatic mapping between images with linear approximations whose parameters are decided by matched pixels of SIFT .These two steps are followed by a discus­sion on image fusion. Several methods have been compared.The first proposition is a generation of typical median filter to the vector range. It is robust when more than half of the images convey the background information. Besides, we design an original algorithm based on the notion of clique. It serves to distinguish the biggest cloud of pixels in RGB space. This approach is highly reliable even when the background pixels are the minority.During the implementation, we notice that some fusion results bear blur-like defects due to the existence of geometric alignment errors. We provide therefore a combination method as a complementary step to ameli-orate the fusion results. It is based on a com-parison between the fusion image and other aligned images after applying a Gaussian filter. The output is a mosaic of patches with clear details issued from the aligned images which are the most similar to their related fusion patches.The performance of our methods is evaluated by a data set containing extensive images of different qualities. Experiments confirm the reliability and robustness of our design under a variety of photography conditions.
• Knowledge Management for Democratic Governance of Socio-Technical Systems
  
  • Pitt Jeremy
  • Diaconescu Ada
  • Ober Josiah
  , 2018, 11300, pp.38-61. The Digital Transformation (DX) is a broad term describing the changes and innovations brought about by the introduction of information and communication technologies into all aspects of society. One such innovation is to empower bottom-up, self-governing socio-technical systems for a range of applications. Such systems can be based on Ostrom’s design principles for self-governing institutions for sustainable common-pool resource management. However, two of these principles, both focussing on self-determination, are vulnerable to distortion: either from within, as a narrow clique take control and run the system in their own, rather than the collective, interest; or from without, as an external authority constrains opportunities for self-organisation. In this chapter, we propose that one approach to maintaining ‘good’, ‘democratic’ self-governance is to appeal to the transparent and inclusive knowledge management processes that were critical to the successful and sustained period of classical Athenian democracy, and reproduce those in computational form. We review a number of emerging technologies which could provide the building blocks for democratic self-governance in socio-technical systems. However, the reproduction of analogue social processes in digital form is not seamless and not without impact on, or consequences for, society, and we also consider a number of open issues which could disrupt this proposal. We conclude with the observation that ‘democracy’ is not an end-state, and emphasise that self-governing socio-technical systems need responsible design and deployment of technologies that allow for continuous re-design and self-organisation. (10.1007/978-3-030-05333-8_4)
  DOI : 10.1007/978-3-030-05333-8_4
• Towards a better formalisation of the side-channel threat
  
  • Cherisey Eloi De
  , 2018. In the field of the security of the embeded systems, it is necessary to know and understandthe possible physical attacks that could break the security of cryptographic components. Sincethe current algorithms such as Advanced Encryption Standard (AES) are very resilient agaisntdifferential and linear cryptanalysis, other methods are used to recover the secrets of thesecomponents. Indeed, the secret key used to encrypt data leaks during the computation of thealgorithm, and it is possible to measure this leakage and exploit it. This technique to recoverthe secret key is called side-channel analysis.The main target of this Ph. D. manuscript is to increase and consolidate the knowledge onthe side-channel threat. To do so, we apply some information theoretic results to side-channelanalysis. The main objective is show how a side-channel leaking model can be seen as acommunication channel.We first show that the security of a chip is dependant to the signal-to-noise ratio (SNR) ofthe leakage. This result is very usefull since it is a genereic result independant from the attack.When a designer builds a chip, he might not be able to know in advance how his embededsystem will be attacked, maybe several years later. The tools that we provide in this manuscriptwill help designers to estimated the level of fiability of their chips.
• Contributions to handwriting recognition using deep neural networks and quantum computation
  
  • Cîrstea Bogdan-Ionut
  , 2018. In this thesis, we provide several contributions from the fields of deep learning and quantum computation to handwriting recognition. We begin by integrating some of the more recent deep learning techniques (such as dropout, batch normalization and different activation functions) into convolutional neural networks and show improved performance on the well-known MNIST dataset. We then propose Tied Spatial Transformer Networks (TSTNs), a variant of Spatial Transformer Networks (STNs) with shared weights, as well as different training variants of the TSTN. We show improved performance on a distorted variant of the MNIST dataset. In another work, we compare the performance of Associative Long Short-Term Memory (ALSTM), a recently introduced recurrent neural network (RNN) architecture, against Long Short-Term Memory (LSTM), on the Arabic handwriting recognition IFN-ENIT dataset. Finally, we propose a neural network architecture, which we name a hybrid classical-quantum neural network, which can integrate and take advantage of quantum computing. While our simulations are performed using classical computation (on a GPU), our results on the Fashion-MNIST dataset suggest that exponential improvements in computational requirements might be achievable, especially for recurrent neural networks trained for sequence classification.
• Contributions to handwriting recognition using deep neural networks and quantum computing
  
  • Cirstea Bogdan-Ionut
  , 2018. In this thesis, we provide several contributions from the fields of deep learning and quantum computation to handwriting recognition. We begin by integrating some of the more recent deep learning techniques (such as dropout, batch normalization and different activation functions) into convolutional neural networks and show improved performance on the well-known MNIST dataset. We then propose Tied Spatial Transformer Networks (TSTNs), a variant of Spatial Transformer Networks (STNs) with shared weights, as well as different training variants of the TSTN. We show improved performance on a distorted variant of the MNIST dataset. In another work, we compare the performance of Associative Long Short-Term Memory (ALSTM), a recently introduced recurrent neural network (RNN) architecture, against Long Short-Term Memory (LSTM), on the Arabic handwriting recognition IFN-ENIT dataset. Finally, we propose a neural network architecture, which we name a hybrid classical-quantum neural network, which can integrate and take advantage of quantum computing. While our simulations are performed using classical computation (on a GPU), our results on the Fashion-MNIST dataset suggest that exponential improvements in computational requirements might be achievable, especially for recurrent neural networks trained for sequence classification.
• Advanced information extraction by example
  
  • Er Ngurah Agus Sanjaya
  , 2018. Searching for information on the Web is generally achieved by constructing a query from a set of keywords and firing it to a search engine. This traditional method requires the user to have a relatively good knowledge of the domain of the targeted information to come up with the correct keywords. The search results, in the form of Web pages, are ranked based on the relevancy of each Web page to the given keywords. For the same set of keywords, the Web pages returned by the search engine would be ranked differently depending on the user. Moreover, finding specific information such as a country and its capital city would require the user to browse through all the documents and reading its content manually. This is not only time consuming but also requires a great deal of effort. We address in this thesis an alternative method of searching for information, i.e. by giving examples of the information in question. First, we try to improve the accuracy of the search by example systems by expanding the given examples syntactically. Next, we use truth discovery paradigm to rank the returned query results. Finally, we investigate the possibility of expanding the examples semantically through labelling each group of elements of the examples.
• Minimum complexity principle for knowledge transfer in artificial learning
  
  • Murena Pierre-Alexandre
  , 2018. Classical learning methods are often based on a simple but restrictive assumption: The present and future data are generated according to the same distributions. This hypothesis is particularly convenient when it comes to developing theoretical guarantees that the learning is accurate. However, it is not realistic from the point of view of applicative domains that have emerged in the last years.In this thesis, we focus on four distinct problems in artificial intelligence, that have mainly one common point: All of them imply knowledge transfer from one domain to the other. The first problem is analogical reasoning and concerns statements of the form "A is to B as C is to D". The second one is transfer learning and involves classification problem in situations where the training data and test data do not have the same distribution (nor even belong to the same space). The third one is data stream mining, ie. managing data that arrive one by one in a continuous and high-frequency stream with changes in the distributions. The last one is collaborative clustering and focuses on exchange of information between clustering algorithms to improve the quality of their predictions.The main contribution of this thesis is to present a general framework to deal with these transfer problems. This framework is based on the notion of Kolmogorov complexity, which measures the inner information of an object. This tool is particularly adapted to the problem of transfer, since it does not rely on probability distributions while being able to model the changes in the distributions.Apart from this modeling effort, we propose, in this thesis, various discussions on aspects and applications of the different problems of interest. These discussions all concern the possibility of transfer in multiple domains and are not based on complexity only.
• Bootstrap and uniform bounds for Harris Markov chains
  
  • Ciolek Gabriela
  , 2018. This thesis concentrates on some extensions of empirical processes theory when the data are Markovian. More specifically, we focus on some developments of bootstrap, robustness and statistical learning theory in a Harris recurrent framework. Our approach relies on the regenerative methods that boil down to division of sample paths of the regenerative Markov chain under study into independent and identically distributed (i.i.d.) blocks of observations. These regeneration blocks correspond to path segments between random times of visits to a well-chosen set (the atom) forming a renewal sequence. In the first part of the thesis we derive uniform bootstrap central limit theorems for Harris recurrent Markov chains over uniformly bounded classes of functions. We show that the result can be generalized also to the unbounded case. We use the aforementioned results to obtain uniform bootstrap central limit theorems for Fr´echet differentiable functionals of Harris Markov chains. Propelledby vast applications, we discuss how to extend some concepts of robustness from the i.i.d. framework to a Markovian setting. In particular, we consider the case when the data are Piecewise-determinic Markov processes. Next, we propose the residual and wild bootstrap procedures for periodically autoregressive processes and show their consistency. In the second part of the thesis we establish maximal versions of Bernstein, Hoeffding and polynomial tail type concentration inequalities. We obtain the inequalities as a function of covering numbers and moments of time returns and blocks. Finally, we use those tail inequalities toderive generalization bounds for minimum volume set estimation for regenerative Markov chains.
• Subpopulation treatment effect Modeling : machine learning approaches to model treatment effect heterogeneity
  
  • Shaar Atef
  , 2018. Subpopulation treatment efect modeling (STEM) is a machine learning techniquethat is used to choose the optimal treatment (i.e., stimulus) for each subgroup. Acritical challenge facing the STEM is information uncertainty. Data uncertaintyexists due to the fundamental problem of causal inference, i.e., only a subset oftreatments' responses are observed. In machine learning domain, specific binningtechniques are applied to bypass the problem of uncertainty. However, one drawbackof current STEM binning approaches is the poor handling of continuous, ordered,and time-series data variables, leading to unreliable and non-interpretable results.In this thesis, first, we all the gaps in the literature and propose a detailed studyof current techniques. Second, we solve STEM shortcomings regarding uncertaintyin the data by proposing subpopulation treatment effect sliding trees. Third, wepropose the subpopulation treatment effect neighborhood random forests to minimizethe effect of noise in data. Fourth, we address the problem of disturbance in databy proposing the balanced reflective uplift modeling technique. We evaluate theperformance of the proposed solutions using simulated and real datasets, and weshow how our approaches outperform other methods in terms of Qini and Spearman'srank correlated coefficient.
• Confused yet successful: Theoretical computation of distinguishers for monobit leakages in terms of confusion coefficient and SNR
  
  • Cherisey Eloi De
  • Guilley Sylvain
  • Rioul Olivier
  , 2018, 11449. Many side-channel distinguishers (such as DPA/DoM, CPA, Euclidean Distance, KSA, MIA, etc.) have been devised and studied to extract keys from cryptographic devices. Each has pros and cons and find applications in various contexts. These distinguishers have been described theoretically in order to determine which distinguisher is best for a given context, enabling an unambiguous characterization in terms of success rate or number of traces required to extract the secret key. In this paper, we show that in the case of monobit leakages, the the- oretical expression of all distinguishers depend only on two parameters: the confusion coefficient and the signal-to-noise ratio. We provide closed- form expressions and leverage them to compare the distinguishers in terms of convergence speed for distinguishing between key candidates. This study contrasts with previous works where only the asymptotic behavior was determined—when the number of traces tends to infinity, or when the signal-to-noise ratio tends to zero. (10.1007/978-3-030-14234-6_28)
  DOI : 10.1007/978-3-030-14234-6_28
• Matrix entropy-power inequality via normal transport
  
  • Rioul Olivier
  • Zamir Ram
  , 2018.
• Techniques d'interaction exploitant la mémoire pour faciliter l'activation de commandes
  
  • Fruchard Bruno
  , 2018. Pour contrôler un système interactif, un utilisateur doit habituellement sélectionner des commandes en parcourant des listes et des menus hiérarchiques. Pour les sélectionner plus rapidement, il peut effectuer des raccourcis gestuels. Cependant, pour être efficace, il doit mémoriser ces raccourcis, une tâche difficile s’il doit activer un grand nombre de commandes. Nous étudions dans une première partie les avantages des gestes positionnels (pointage) et directionnels (Marking menus) pour la mémorisation de commandes, ainsi que l’utilisation du corps de l’utilisateur comme surface d’interaction et l’impact de deux types d’aides sémantiques (histoires, images) sur l’efficacité à mémoriser. Nous montrons que les gestes positionnels permettent d’apprendre plus rapidement et plus facilement, et que suggérer aux utilisateurs de créer des histoires liées aux commandes améliore considérablement leurs taux de rappel. Dans une deuxième partie, nous présentons des gestes bi-positionnels qui permettent l’activation d’un grand nombre de commandes. Nous montrons leur efficacité à l’aide de deux contextes d’interaction : le pavé tactile d’un ordinateur portable (MarkPad) et une montre intelligente (SCM).
• Cognitive management of self organized radio networks of fifth generation
  
  • Daher Tony
  , 2018. The pressure on operators to improve the network management efficiency is constantly growing for many reasons: the user traffic that is increasing very fast, higher end users expectations, emerging services with very specific requirements. Self-Organizing Networks (SON) concept was introduced by the 3rd Generation Partnership Project as a promising solution to simplify the operation and management of complex networks. Many SON modules are already being deployed in today’s networks. Such networks are known as SON enabled networks, and they have proved to be useful in reducing the complexity of network management. However, SON enabled networks are still far from realizing a network that is autonomous and self-managed as a whole. In fact, the behavior of the SON functions depends on the parameters of their algorithm, as well as on the network environment where it is deployed. Besides, SON objectives and actions might be conflicting with each other, leading to incompatible parameter tuning in the network. Each SON function hence still needs to be itself manually configured, depending on the network environment and the objectives of the operator. In this thesis, we propose an approach for an integrated SON management system through a Cognitive Policy Based SON Management (C-PBSM) approach, based on Reinforcement Learning (RL). The C-PBSM translates autonomously high level operator objectives, formulated as target Key Performance Indicators (KPIs), into configurations of the SON functions. Furthermore, through its cognitive capabilities, the C-PBSM is able to build its knowledge by interacting with the real network. It is also capable of adapting with the environment changes. We investigate different RL approaches, we analyze the convergence time and the scalability and propose adapted solutions. We tackle the problem of non-stationarity in the network, notably the traffic variations, as well as the different contexts present in a network. We propose as well an approach for transfer learning and collaborative learning. Practical aspects of deploying RL agents in real networks are also investigated under Software Defined Network (SDN) architecture.
• Scheduling Multi-Periodic Mixed-Criticality DAGs on Multi-Core Architectures
  
  • Medina Roberto
  • Borde Etienne
  • Pautet Laurent
  , 2018. Thanks to Mixed-Criticality (MC) scheduling, high and low-criticality tasks can share the same execution platform, improving considerably the usage of computation resources. Even if the execution platform is shared with low-criticality tasks, deadlines of high-criticality tasks must be respected. This is usually enforced thanks to operational modes of the system: if necessary, a high-criticality execution mode allocates more time to high-criticality tasks at the expense of low-criticality tasks' execution. Nonetheless, most MC scheduling policies in the literature have only considered independent task sets. For safety-critical real-time systems, this is a strong limitation: models used to describe reactive safety-critical software often consider dependencies among tasks or jobs. In this paper, we define a meta-heuristic to schedule multi-processor systems composed of multi-periodic Directed Acyclic Graphs of MC tasks. This meta-heuristic computes the scheduling of the system in the high-criticality mode first. The computation of the low-criticality scheduling respects a condition on high-criticality tasks' jobs, ensuring that high-criticality tasks never miss their deadlines. Two implementations of this meta-heuristic are presented. In high-criticality mode, high-criticality tasks are scheduled as late as possible. Then two global scheduling tables are produced, one per criticality mode. Experimental results demonstrate our method outperforms approaches of the literature in terms of acceptance rate for randomly generated systems. (10.1109/RTSS.2018.00042)
  DOI : 10.1109/RTSS.2018.00042
• Shedding the Shackles of Time-Division Multiplexing
  
  • Hebbache Farouk
  • Jan Mathieu
  • Brandner Florian
  • Pautet Laurent
  , 2018, pp.456-468. Multi-core architectures pose many challenges in real-time systems, which arise from contention between concurrent accesses to shared memory. Among the available memory arbitration policies, Time Division Multiplexing (TDM) ensures a predictable behavior by bounding access latencies and guaranteed bandwidth to tasks independently from the other tasks. To do so, TDM guarantees exclusive access to the shared memory in a fixed time window. TDM, however, provides a low resource utilization as it is non-work-conserving. Besides, it is very inefficient for resources having highly variable latencies, such as sharing the access to a DRAM memory. The constant length of a TDM slot is, hence, highly pessimistic and causes an underutilization of the memory. To address these limitations, we present dynamic arbitration schemes that are based on TDM. However, instead of arbitrating at the level of TDM slots, our approach operates at the granularity of clock cycles by exploiting slack time accumulated from preceding requests. This allows the arbiter to reorder memory requests, exploit the actual access latencies of requests, and thus improve memory utilization. We demonstrate that our policies are analyzable as they preserve the guarantees of TDM in the worst case, while our experiments show an improved memory utilization on average. (10.1109/RTSS.2018.00059)
  DOI : 10.1109/RTSS.2018.00059
• Generic Architecture for Lightweight Block Ciphers: A First Step Towards Agile Implementation of Multiple Ciphers
  
  • Tehrani Etienne
  • Danger Jean-Luc
  • Graba Tarik
  , 2019, LNCS-11469, pp.28-43. Lightweight cryptography is at the heart of today’s security needs for embedded systems. The standardised cryptographic algorithms, such as the Advanced Encryption Standard (AES), hardly fits the resource restrictions of those small and pervasive devices. From this observation a plethora of Lightweight Block Ciphers have been proposed. Every algorithm has its own advantages in terms of security, complexity, latency, performances. This paper presents first a classification of some popular Substitution-Permutation-Networks (SPN) class of lightweight ciphers according to their architecture and features which share many common operators. From this last point, we studied a round-based generic hardware architecture that allows a security architect to dynamically change the lightweight cryptographic algorithms to be executed. The results of the ASIC implementation show that the configuration part of the proposed flexible architecture adds significant complexity. If compared with the parallel implementation of several algorithms, the complexity ratio becomes interesting when the number of algorithms (or the level of agility) increases. For instance, if we consider 6 SPN ciphers, the configurable architecture provides a complexity reduction of 62.5%, whereas there is no reduction with 4 algorithms. (10.1007/978-3-030-20074-9_4)
  DOI : 10.1007/978-3-030-20074-9_4
• Towards end-to-end privacy for publish/subscribe architectures in the Internet of Things
  
  • Coroller Stevan
  • Chabridon Sophie
  • Laurent Maryline
  • Conan Denis
  • Leneutre Jean
  , 2018, pp.35 - 40. The Internet of Things paradigm lacks end-to-end privacy solutions to consider its full adoption in real life scenarios in the near future. The recent enactment of the EU General Data Protection Regulation (GDPR) indeed emphasises the need for stronger security and privacy measures for personal data processing and free movement, including consent management and accountability by the data controller and processor. In this paper, we suggest an architecture to enforce end-to-end data usage control in Distributed Event-Based Systems (DEBS), from data producers to consumer services, taking into account some of the GDPR requirements concerning consent management and data processing transparency. Our architecture proposal is based on UCON ABC usage control models, which we overlap with a distributed hash table overlay for scalability and fault-tolerance concerns, and across and within systems data usage control. Our proposal highlights the benefits of combining both DEBS and end-user usage control architectures. To complete our approach, we quickly survey existing encryption models that ensure data confidentiality in topic-based Publish/Subscribe systems and highlight the remaining obstacles to transpose them to content-based DEBS with an overlay of brokers (10.1145/3286719.3286727)
  DOI : 10.1145/3286719.3286727
• Prediction-Based Intrusion Detection System for In-Vehicle Networks Using Supervised Learning and Outlier-Detection
  
  • Elaabid Moulay Abdelaziz
  • Karray Khaled
  • Danger Jean-Luc
  • Guilley Sylvain
  , 2019, LNCS-11469, pp.109-128. Modern connected vehicles are composed of multiple electronic control units (ECUs) holding sensors, actuators but also wired and wireless connection interfaces, all communicating over shared internal communication buses. The cyber-physical architecture based on this ECU network has been proven vulnerable to multiple types of attacks leveraging remote, direct and indirect physical access. Attacks initiated from these access vectors go through the internal communication buses and spread over the whole network of ECUs. For this reason it is important to detect, and if possible to mitigate, attacks on the internal buses of the vehicle.In this article, a novel intrusion detection system is developed to monitor vehicle state from information collected on internal buses. Based on supervised machine learning techniques, a normal behavior is learned and used as a reference to detect deviations. The principle is to learn how to predict the next state of the vehicle based on information and sensor values sent over communication buses. Experimental validation is conducted using data collected from different drivers. Results show that the approach is able to learn the nominal behavior with high accuracy for a single driver as well as for a set of different drivers. Results also demonstrate its ability to predict attacks with low false negative rate. This motivates the approach to be used for indirect and remote attacks intrusion detection as well as for safety purposes to detect sensor failures, lost connection with the sensor, etc. (10.1007/978-3-030-20074-9_9)
  DOI : 10.1007/978-3-030-20074-9_9
• A Sketch-Based Naive Bayes Algorithms for Evolving Data Streams
  
  • Bahri Maroua
  • Maniu Silviu
  • Bifet Albert
  , 2018, pp.604-613. A well-known learning task in big data stream mining is classification. Extensively studied in the offline setting, in the streaming setting - where data are evolving and even infinite - it is still a challenge. In the offline setting, training needs to store all the data in memory for the learning task; yet, in the streaming setting, this is impossible to do due to the massive amount of data that is generated in real-time. To cope with these resource issues, this paper proposes and analyzes several evolving naive Bayes classification algorithms, based on the well-known count-min sketch, in order to minimize the space needed to store the training data. The proposed algorithms also adapt concept drift approaches, such as ADWIN, to deal with the fact that streaming data may be evolving and change over time. However, handling sparse, very high-dimensional data in such framework is highly challenging. Therefore, we include the hashing trick, a technique for dimensionality reduction, to compress that down to a lower dimensional space, which leads to a large memory saving.We give a theoretical analysis which demonstrates that our proposed algorithms provide a similar accuracy quality to the classical big data stream mining algorithms using a reasonable amount of resources. We validate these theoretical results by an extensive evaluation on both synthetic and real-world datasets. (10.1109/BigData.2018.8622178)
  DOI : 10.1109/BigData.2018.8622178
• Learning Fast and Slow: A Unified Batch/Stream Framework
  
  • Montiel Jacob
  • Bifet Albert
  • Losing Viktor
  • Read Jesse
  • Abdessalem Talel
  , 2018, pp.1065-1072. Data ubiquity highlights the need of efficient and adaptable data-driven solutions. In this paper, we present FAST AND SLOW LEARNING (FSL), a novel unified framework that sheds light on the symbiosis between batch and stream learning. FSL works by employing Fast (stream) and Slow (batch) Learners, emulating the mechanisms used by humans to make decisions. We showcase the applicability of FSL on the task of classification by introducing the FAST AND SLOW CLASSIFIER (FSC). A Fast Learner provides predictions on the spot, continuously updating its model and adapting to changes in the data. On the other hand, the Slow Learner provides predictions considering a wider spectrum of seen data, requiring more time and data to create complex models. Once that enough data has been collected, FSC trains the Slow Learner and starts tracking the performance of both learners. A drift detection mechanism triggers the creation of new Slow models when the current Slow model becomes obsolete. FSC selects between Fast and Slow Learners according to their performance on new incoming data. Test results on real and synthetic data show that FSC effectively drives the positive interaction of stream and batch models for learning from evolving data streams. (10.1109/BigData.2018.8622222)
  DOI : 10.1109/BigData.2018.8622222