Browser-based fog computing

The dramatic increase in the amount of data being produced and processed by connected devices has led to paradigms that seek to decentralize the traditional cloud model. In 2011 Cisco  48 introduced the vision of fog computing that combines the cloud with resources located at the edge of the network and in between. More generally, the term edge computing has been associated with the idea of adding edge-of-the network storage and computation to traditional cloud infrastructures  43.

A number of efforts in this directions focus on specific hardware, e.g. fog nodes that are responsible for connected IoT devices  49. However, many of today's applications run within web browsers or mobile phones. In this context, the recent introduction of the WebRTC API, makes it possible for browsers and smartphones to exchange directly between each other, enabling mobile, or browser-based decentralized applications.

Maygh  72, for example, uses the WebRTC API to build a decentralized Content Delivery Network that runs solely on web browsers. The fact that the application is hosted completely on a web server and downloaded with enabled websites means that webmasters can adopt the Content Delivery Network (CDN) without requiring users to install any specific software.

For us, the ability of browsers to communicate with each other using the WebRTC paradigm provides a novel playground for new programming models, and for a browser-based fog architecture combining both a centralized, cloud-based part, and a decentralized, browser-supported part.

This model offers tremendous potential by making edge-of-the-network resources available through the interconnection of web-browsers, and offers new opportunities for the protection of the personal data of end users. But consistently engineering browser-based components requires novel tools and methodologies.

In particular, WebRTC was primarily designed for exchanging media and data between two browsers in the presence of a coordinating server. Its complex mechanisms for connection establishment make many of the existing peer-to-peer protocols inefficient. To address this challenge, we plan to consider two angles of attack. First, we plan to design novel protocols that take into account the specific requirements set by this new technology. Second, we envisage to investigate variants of the current WebRTC model with cheaper connection-establishment protocols, in order to provide lower delays and bandwidth consumption in large-scale browser-based applications.

We also plan to address the trade-offs associated with hybrid browser-cloud models. For example, when should computation be delegated to browsers and when should it be executed on the cloud in order to maximize the quality of service? Or, how can a decentralized analytics algorithms operating on browser-based data complement or exploit the knowledge built by cloud-based data analytics solutions?

Emergent micro-service deployment and management

Micro-services tend to produce fine-grained applications in which many small services interact in a loosely coupled manner to produce a wide range of services within an organization. Individual services need to evolve independently of each other over time without compromising the availability of the overall application. Lightweight isolation solutions such as containers (Docker, ...), and their associated tooling ecosystem (e.g. Google's Borg  71, Kubernetes  46) have emerged to facilitate the deployment of large-scale micro-service-based applications, but only provide preliminary solutions for key concerns in these systems, which we would like to investigate and extend.

Most of today's on-line computer systems are now too large to evolve in monolithic, entirely pre-planned ways. This applies to very large data centres, for example, where the placement of virtual machines to reduce heating and power consumption can no longer be treated using top-down exhaustive optimisation approaches beyond a critical size. This is also true of social networking applications, where different mechanisms—e.g. to spread news notifications, or to recommend new contacts—must be adapted to the different sub-communities present in the system.

To cope with the inherent complexity of building complex loosely-coupled distributed systems while fostering and increasing efficiency, maintainability, and scalability, we plan to study how novel programming techniques based on declarative programming, components and epidemic protocols can help design, deploy, and maintain self-adaptive structures (e.g. placement of VM) and mechanisms (e.g. contact recommendations) that are optimized to the local context of very large distributed systems. To fulfill this vision, we plan to explore a three-pronged strategy to raise the level of programming abstraction offered to developers.

• First, we plan to explore the use of high-level domain-specific languages (DSL) to declare how large-scale topologies should be achieved, deployed, and maintained. Our vision is a declarative approach to describe how to combine, deploy and orchestrate micro-services in an abstract manner thus abstracting away developers from the underlying cloud infrastructures, and from the intricacies involved in writing low-level code to build a large-scale distributed application that scales. With this effort, we plan notably to directly support the twin properties of emergence (the adaptation “from within”) and differentiation (the possibility from parts of the system to diverge while still forming a whole). Our central objective is to search for principled programming constructs to support these two capabilities using a modular and incremental software development approach.
• On a second strand of work, we plan to investigate how unikernels enable smaller footprints, more optimization options, and faster boot times for micro-services. Isolating micro-services into VMs is not the most adequate approach as it requires the use of hypervisors, or virtual machine monitors (VMMs), to virtualize hardware resources. VMMs are well known to be heavyweight with both boot and run time overheads that may have a strong impact on performances. Unikernels seem to offer the right balance between performance and flexibility to address this challenge. One of the key underlying challenges is to compile directly the aforementioned provided DSL to a dedicated and customized machine image, ready to be deployed directly on top of a large set of bare metal servers.
• Depending on the workload it is subjected to, and the state of its execution environment (network, VMs), a large-scale distributed application may present erratic or degraded performance that is hard to anticipate and plan for. There is therefore a strong need to adapt dynamically the way resources are allocated to a running application. We would like to study how the DSL approach we envisage can be extended to enable developers to express orchestration algorithms based on machine learning algorithms.

Privacy-preserving decentralized learning

Personalization and recommendation can be seen as a specific case of general machine learning. Production-grade recommenders and personalizers typically centralize and process the available data in one location (a data-center, a cloud service). This is highly problematic, as it endangers the privacy of users, while hampering the analysis of datasets subject to privacy constraints that are held by multiple independent organizations (such as health records). A decentralized approach to machine learning appears as a promising candidate to overcome these weaknesses: if each user or participating organization keeps its data, while only exchanging gradient or model information, privacy leaks seem less likely to occur.

In some cases, decentralized learning may be achieved through relatively simple adaptations of existing centralized models, for instance by defining alternative learning models that may be more easily decentralized. But in all cases, processing growing amounts of information calls for high-performance algorithms and middleware that can handle diverse storage and computation resources, in the presence of dynamic and privacy-sensitive data. To reach this objective, we will therefore leverage our work in distributed and privacy-preserving algorithms and middleware  50, 52, 53 as well as the results of our work on large-scale hybrid architectures in Objective 1.

Personalization in user-oriented applications

As a first application perspective, we plan to design tools that exploit decentralized analytics to enhance user-centric personalized applications. As we observed above, such applications exhibit an inherent trade-off between personalization quality and privacy preservation. The most obvious goal in this direction consists in designing algorithms that can achieve high levels of personalization while protecting sensitive user information. But an equally important one consists in personalizing the trade-off itself by adapting the quality of the personalization provided to a user to his/her willingness to expose information. This, like other desirable behaviors, appears at odds with the way current systems work. For example, a user of a recommender system that does not reveal his/her profile information penalizes other users causing them to receive less accurate recommendations. We would like to mitigate this situation by means of protocols that reward users for sharing information. On the one hand, we plan to take inspiration from protocols for free-riding avoidance in peer-to-peer systems  54, 59. On the other hand, we will consider blockchains as a tool for tracking and rewarding data contributions. Ultimately, we aim at enabling users to configure the level of privacy and personalization they wish to experience.

Privacy preserving decentralized aggregation

As a second setting we would like to consider target applications running on constrained devices like in the Internet-of-Things (IoT). This setting makes it particularly important to operate on decentralized data in a light-weight privacy-preserving manner, and further highlights the synergy between this objective and Objective 1. For example, we plan to provide data subjects with the possibility to store and manage their data locally on their own devices, without having to rely on third-party managers or aggregators, but possibly storing less private information or results in the cloud. Using this strategy, we intend to design protocols that enable users themselves, or third-party companies to query distributed data in aggregate form, or to run data analytics processes on a distributed set of data repositories, thereby gathering knowledge without violating the privacy of other users. For example, we have started working on the problem of computing an aggregate function over a subset of the data in a distributed setting. This involves two major steps: selection and aggregation. With respect to selection, we envision defining a decentralized data-selection operation that can apply a selection predicate without violating privacy constraints. With respect to aggregation, we will continue our investigation of lightweight protocols that can provide privacy with limited computational complexity  44.

Spread maximization

Our goal is in particular to study spread maximization in a broader class of diffusion processes than the basic independent cascade (IC) and linear threshold (LT) models of influence  65, 63, 64 that have been studied in this context so far. This includes the randomized rumor spreading (RS) model for information dissemination  56, biased versions of the voter model  60 modelling influence, and the (graph-based) Moran processes  67 modelling the spread of mutations. We would like to consider several natural versions of the spread maximization problem, and the relationships between them. For these problems we will use the greedy algorithm and the submodularity-based analytical framework of  65, and will also explore new approaches.

Immunization optimization

Conversely we would also like to explore immunization optimization problems. Existing works on these types of problem assume a perfect-contagion model, i.e., once a node gets infected, it deterministically infects all its non-immunized neighbors. We plan to consider various diffusion processes, including the standard susceptible–infected (SI), susceptible–infected–recovered (SIR) and susceptible–infected–susceptible (SIS) epidemic models, and explore the extent to which results and techniques for the perfect-contagion model carry over to these probabilistic models. We will also investigate whether techniques for spread maximization could be applied to immunization problems.

Some immunization problems are known to be hard to approximate in general graphs, even for the perfect-contagion model, e.g., the fixed-budget version of the fire-fighter problem cannot be approximated to any $n^{1-ϵ}$ factor  45. This strand of work will consider restricted graph families, such as trees or graphs of small treewidth, for such problems. In addition, for some immunization problems, there is a large gap between the best known approximation algorithm and the best known inaproximability result, and we would like to make progress in reducing these gaps.

Randomized algorithm for mutual exclusion and coordination

To exploit the power of massive distributed applications and systems (such as those envisaged in Objectives 1 and 2) or multiple processors, algorithms must cope with the scale and asynchrony of these systems, and their inherent instability, e.g., due to node, link, or processor failures. Our goal is to explore the power and limits of randomized algorithms for large-scale networks of distributed systems, and for shared memory multi-processor systems, in effect providing fundamental building blocks to the work envisioned in Objectives 1 and 2.

For shared memory systems, randomized algorithms have notably proved extremely useful to deal with asynchrony and failures. Sometimes probabilistic algorithms provide the only solution to a problem; sometimes they are more efficient; sometimes they are simply easier to implement. We plan to devise efficient algorithms for some of the fundamental problems of shared memory computing, such as mutual exclusion, renaming, and consensus.

In particular, looking at the problem of mutual exclusion, it is desirable that mutual exclusion algorithms be abortable. This means that a process that is trying to lock the resource can abort its attempt in case it has to wait too long. Abortability is difficult to achieve for mutual exclusion algorithms. We will try to extend our algorithms for the cache-coherent (CC) and the distributed shared memory (DSM) model in order to make them abortable, while maintaining expected constant Remote Memory References (RMRs) complexity, under optimistic system assumptions. In order to achieve this, the algorithm will use strong synchronization primitives, called compare-and-swap objects. As part of our collaboration with the University of Calgary, we will work on implementing those objects from registers in such a way that they also allow aborts. Our goal is to build on existing non-abortable implementations 58. We plan then later to use these objects as building blocks in our mutual exclusion algorithm, in order to make them work even if the system does not readily provide such primitives.

We have also started working on blockchains, as these represent a new and interesting trade-off between probabilistic guarantees, scalability, and system dynamics, while revisiting some of the fundamental questions and limitations of consensus in fault-prone asynchronous systems.

Modular theory of distributed computing

Practitioners and engineers have proposed a number of reusable frameworks and services to implement specific distributed services (from Remote Procedure Calls with Java RMI or SOAP-RPC, to JGroups for group communication, and Apache Zookeeper for state machine replication). In spite of the high conceptual and practical interest of such frameworks, many of these efforts lack a sound grounding in distributed computation theory (with the notable exceptions of JGroups and Zookeeper), and often provide punctual and partial solutions for a narrow range of services. We argue that this is because we still lack a generic framework that unifies the large body of fundamental knowledge on distributed computation that has been acquired over the last 40 years.

To overcome this gap we would like to develop a systematic model of distributed computation that organizes the functionalities of a distributed computing system into reusable modular constructs assembled via well-defined mechanisms that maintain sound theoretical guarantees on the resulting system. This research vision arises from the strong belief that distributed computing is now mature enough to resolve the tension between the social needs for distributed computing systems, and the lack of a fundamentally sound and systematic way to realize these systems.

To progress on this vision, we plan in the near future to investigate, from a distributed software point of view, the impact due to failures and asynchrony on the layered architecture of distributed computing systems. A first step in this direction will address the notions of message adversaries (introduced a long time ago in  69) and process adversaries (investigated in several papers, e.g.  68, 55, 61, 62, 66). The aim of these notions is to consider failures, not as “bad events”, but as part of the normal behavior of a system. As an example, when considering round-based algorithms, a message adversary is a daemon which, at every round, is allowed to suppress some messages. The aim is then, given a problem $P$, to find the strongest adversary under which $P$ can be solved (“strongest” means here that giving more power to the adversary makes the problem impossible to solve). This work will allow us to progress in terms of general layered theory of distributed computing, and allow us to better map distributed computing models and their relations, in the steps of noticeable early efforts in this direction  68, 42.

6.1.1 KIFF

• Name:
  KIFF: An impressively fast and efficient JAVA library for KNN construction
• Keyword:
  KNN
• Functional Description:

  This package implements the KIFF algorithm reported in [1]. KIFF is a generic, fast and scalable K-Nearest-Neighbor graph construction algorithm. This algorithm connects each object to its k most similar counterparts, according to a given similarity metric. In term of comparison, this package implements also HYREC [2] and NN-DESCENT [3]. The standalone program implements cosine similarity only, however this library supports arbitrary similarity measures.

  [1] Antoine Boutet, Anne-Marie Kermarrec, Nupur Mittal, Francois Taiani. Being prepared in a sparse world: the case of KNN graph construction. ICDE 2016, Finland.

• Contact:
  Antoine Boutet
• Partner:
  LIRIS

6.1.2 Dietcoin

• Keywords:
  Blockchain, Bitcoin, Smartphone
• Functional Description:
  dietcoin-lib is a Java library implementing the Dietcoin protocol which is an extension of the Bitcoin protocol with the goal of enhancing Bitcoin user security on resource-constrained devices such as smartphones. This piece of software is based on the bitcoinj library, which is the most popular Java implementation of Bitcoin notably used for Android applications. dietcoin-lib also enables a block-by-block replay of the existing Bitcoin blockchain to simulate a Dietcoin client downloading the blockchain from a Dietcoin server, both ends using dietcoin-lib, with the goal of evaluation the protocole on a large real dataset.
• URL:
  https://­gitlab.­inria.­fr/­wide/­dietcoin
• Publications:
  hal-02315154, tel-01964628, hal-01743995
• Contact:
  Pierre-Louis Roman
• Participants:
  Pierre-Louis Roman, Davide Frey, François Taiani, Marc Makkes, François Taiani

6.1.3 Basalt

• Keywords:
  Peer-sampling, Blockchain
• Functional Description:
  A number of novel blockchain and cryptocurrency implementations rely on random peer sampling. But existing protocols remain vulnerable to Sybil attacks. BASALT is a peer-sampling protocol that addresses this limitation by leveraging three main components. First it employs a novel sampling approach, termed stubborn chaotic search, that exploits ranking functions to define a dynamic target random graph (i.e. a set of v target neighbors for each node) that cannot be controlled by Byzantine nodes. Second, it adopts a hit-counter mechanism that favors the exploration of new peers even in the presence of Byzantine nodes that flood the network with their identities. Finally, it incorporates hierarchical ranking functions that ensure that nodes sample their peers from a variety of address prefixes. The first two mechanisms ensure that the number of Byzantine nodes in a node’s view cannot be increased arbitrarily by attackers. This offers protection from general Byzantine behaviors including those resulting from botnet attacks, as defined above. The third mechanism ensures that nodes sample their peers from a variety of address prefixes, thereby countering institutional attacks where the attacker controls a limited number of entire address prefixes.
• Contact:
  Alex Auvolat-Bernstein

6.1.4 Killerdroid-packer

• Name:
  Practical implémentation of dissimulation techniques for Android application
• Keywords:
  Malware, Android, Machine learning, Cybersecurity
• Functional Description:
  Killerdroid-packer allows to intricate a Android guest application in a Android host application. A common use case is to hide a malicious code inside a benign one to evade typical detection systems. Killerdroid-packer has been proven effective and scalable. Applications generated with the tool can evade state-of-the art detection systems (Drebin, Mamadroid ...) and can be used to forge large scale datasets of malicious applications.
• Publication:
  hal-03146161
• Contact:
  David Bromberg
• Participants:
  David Bromberg, Matthieu Simonin, Louison Gitzinger
• Partner:
  DGA-MI

6.1.5 python-andromak

• Name:
  Andromak: A library to assess Android detector quality
• Keywords:
  Android, Anomaly detection, Cybersecurity, Distribution, Malware, Machine learning
• Functional Description:

  State of the Art detectors are numerous, but actually comparing them in a thorough way is tedious. Andromak aims to make a step toward easier and sound comparisons between the top existing detectors.

  Python-andromak is a library to

  (a) scale your training (features extraction can transparently leverage your laptop of a large scale grid) (b) evaluate the quality of you model against several input datasets and scenario (c) analyse the performance results (domain-specific metrics and statistical tests are shipped)

• Contact:
  David Bromberg

7.1.1 Byzantine-Tolerant Causal Broadcast

Participants: Davide Frey, Alex Auvolat, Michel Raynal, Francois Taiani.

With this work, published on Journal of Theoretical Computer Science (TCS) 13 we were the first to give a formal definition of a causal broadcast abstraction in the presence of Byzantine processes, in the form of two equivalent characterizations, and to present a simple causal broadcast algorithm that implements it. The main difficulty in the design and the proof of this algorithm comes from the very nature of Byzantine faults: Byzantine processes may have arbitrary behavior, and the algorithm must ensure that correct processes (i) maintain a coherent view of causality and (ii) are never prevented from communicating between themselves. To this end, our algorithm is built modularly, namely it works on top of any Byzantine-tolerant reliable broadcast algorithm. Due to this modularity, the proposed algorithm is easy to understand and inherits the computability assumptions (notably the maximal number of processes that may be Byzantine) and the message/time complexities of the underlying reliable broadcast on top of which it is built.

7.1.2 Byzantine-Tolerant Reliable Broadcast in the Presence of Silent Churn

Participants: Timothe Albouy, Davide Frey, Michel Raynal, Francois Taiani.

7.1.3 On the Versatility of Bracha’s Byzantine Reliable Broadcast Algorithm

Participants: Michel Raynal.

G. Bracha presented in 1987 a simple and efficient reliable broadcast algorithm for $n$-process asynchronous message-passing systems, which tolerates up to $t<n/3$ Byzantine processes. Following an idea recently introduced by Hirt, Kastrato and Liu-Zhang (OPODIS 2020), instead of considering the upper bound on the number of Byzantine processes, we consider in 20 two types of Byzantine behavior: the ones that can prevent the safety property from being satisfied, and the ones that can prevent the liveness property from being satisfied (a Byzantine process can exhibit only one or both types of failures). This Byzantine differentiated failure model is captured by two associated upper bounds denoted $t_s$ (for safety) and $t_{\ell }$ for liveness). We show that only the threshold values used in the predicates of Bracha’s algorithm must be modified to obtain an algorithm that works with this differentiated Byzantine failure model.

7.1.4 Consensus-Free Ledgers When Operations of Distinct Processes are Commutative

Participants: Davide Frey, Lucie Guillou, Michel Raynal, Francois Taiani.

Considering asynchronous message-passing systems in which any number of processes may crash, we explored the construction of ledger objects where the append operations issued from distinct processes commute. In a very interesting way, it appears that the implementation of such ledgers does not need consensus, which makes them both attractive and efficient. We also proposed a formalization based on Mazurkiewicz's traces. We published this work at PACT2021 28.

7.1.5 The graph neural networking challenge: A worldwide competition for education in AI/ML for networks

Participants: Loïck Bonniot, Francois Taiani.

During the last decade, Machine Learning (ML) has increasingly become a hot topic in the field of Computer Networks and is expected to be gradually adopted for a plethora of control, monitoring and management tasks in real-world deployments. This poses the need to count on new generations of students, researchers and practitioners with a solid background in ML applied to networks. During 2020, the International Telecommunication Union (ITU) has organized the "ITU AI/ML in 5G challenge", an open global competition that has introduced to a broad audience some of the current main challenges in ML for networks. This large-scale initiative has gathered 23 different challenges proposed by network operators, equipment manufacturers and academia, and has attracted a total of 1300+ participants from 60+ countries. This work 21 describes our experience organizing one of the proposed challenges: the "Graph Neural Networking Challenge 2020". We describe the problem presented to participants, the tools and resources provided, some organization aspects and participation statistics, an outline of the top-3 awarded solutions, and a summary with some lessons learned during all this journey. As a result, this challenge leaves a curated set of educational resources openly available to anyone interested in the topic.

This is a joint work with Christoph Neumann and François Schnitzler from InterDigital.

7.1.6 Towards Internet-Scale Convolutional Root-Cause Analysis with DiagNet

Participants: Loïck Bonniot, Francois Taiani.

Diagnosing problems in Internet-scale services remains particularly difficult and costly for both content providers and ISPs. Because the Internet is decentralized, the cause of such problems might lie anywhere between a user’s device and the datacenters hosting the service. Further, the set of possible problems and causes is not known in advance, making it impossible in practice to train a classifier with all combinations of problems, causes and locations. In this work 23, we explore how machine learning techniques can be used for Internet-scale root cause analysis based on measurements taken from end-user devices. Using convolutional neural networks, we show how to build generic models that (i) are agnostic to the underlying network topology, (ii) do not require to define the full set of possible causes during training, and (iii) can be quickly adapted to diagnose new services. We evaluate our proposal, DiagNet, on a geodistributed multi-cloud deployment of online services, using a combination of fault injection and emulated clients running within automated browsers. Our experiments demonstrate the promising capabilities of our technique, delivering a recall of 73.9%, including on causes that were unknown at training time.

This is a joint work with Christoph Neumann and François Schnitzler from InterDigital.

7.1.7 Tell me when you are sleepy and what may wake you up!

Participants: Barbe Thystere Mvondo Djob.

Nowadays, there is a shift in the deployment model of applications in the Cloud. Applications are now deployed as a set of several small units communicating with each other, thus an obvious similarity with the microservice model. Moreover, each unit – a microservice, may be implemented as a virtual machine, container, function, etc., spanning the different cloud service models IaaS, PaaS, FaaS, etc. A microservice is instantiated upon the reception of a request (e.g., an http packet or a trigger), and a rack-level or data-center-level scheduler decides the placement for such unit of execution considering for example data locality and load balancing. With such a configuration, it is common to encounter scenarios where different units, as well as multiple instances of the same unit, may be running on a single server at the same time.

When multiple microservices are running on the same server not necessarily all of them are doing actual processing, some may be busy-waiting – i.e., waiting for events (or requests) sent by other units. However, these "idle" units are consuming CPU time which could be used by other running units or cloud utility functions on the server (e.g., monitoring daemons). In a controlled experiment, we observe that units can spend up to 20% - 55% of their CPU time waiting, thus a great amount of CPU time is wasted; these values significantly grow when overcommitting CPU resources (i.e., CPU reservations of units exceed server CPU capacity), where we observe up to 69% - 75%. This is a result of the lack of information/context about what is running in each unit from the server CPU scheduler perspective. In 35, we first provide evidence of the problem and discuss several research questions. Then, we propose a handful of solutions worth exploring that consists in revisiting hypervisor and host OS scheduler designs to reduce the CPU time wasted on idle units. Our proposal leverages the concepts of informed scheduling, and monitoring for internal and external events. Based on the aforementioned solutions, we propose our initial implementation on Linux/KVM.

7.2.1 SurFree: A fast surrogate-free black-box attack

Participants: Erwan Le Merrer.

Machine learning classifiers are critically prone to evasion attacks. Adversarial examples are slightly modified inputs that are then misclassified, while remaining perceptively close to their originals. Last couple of years have witnessed a striking decrease in the amount of queries a black box attack submits to the target classifier, in order to forge adversarials. This particularly concerns the black-box score-based setup, where the attacker has access to top predicted probabilites: the amount of queries went from to millions of to less than a thousand. In 34 we presents SurFree, a geometrical approach that achieves a similar drastic reduction in the amount of queries in the hardest setup: black box decision-based attacks (only the top-1 label is available). We first highlight that the most recent attacks in that setup, HSJA, QEBA and GeoDA all perform costly gradient surrogate estimations. SurFree proposes to bypass these, by instead focusing on careful trials along diverse directions, guided by precise indications of geometrical properties of the classifier decision boundaries. We motivate this geometric approach before performing a head-to-head comparison with previous attacks with the amount of queries as a first class citizen. We exhibit a faster distortion decay under low query amounts (few hundreds to a thousand), while remaining competitive at higher query budgets.

Join work with Teddy Furon (Inria) and Thibault Maho (Inria).

7.2.2 RoBIC: A benchmark suite for assessing classifiers robustness

Participants: Erwan Le Merrer.

Many defenses have emerged with the development of adversarial attacks. Models must be objectively evaluated accordingly. In 33 we systematically tackle this concern by proposing a new parameter-free benchmark we coin RoBIC. RoBIC fairly evaluates the robustness of image classifiers using a new half-distortion measure. It gauges the robustness of the network against white and black box attacks, independently of its accuracy. RoBIC is faster than the other available benchmarks. We present the significant differences in the robustness of 16 recent models as assessed by RoBIC. We make this benchmark publicly available for use and contribution at https://gitlab.inria.fr/tmaho/robustness_benchmark.

Join work with Benoît Bonnet (Inria), Teddy Furon (Inria) and Thibault Maho (Inria).

7.2.3 Setting the Record Straighter on Shadow Banning

Participants: Erwan Le Merrer.

"Shadow banning" consists for an online social network in limiting the visibility of some of its users, without them being aware of it. Twitter declares that it does not use such a practice, sometimes arguing about the occurrence of "bugs" to justify restrictions on some users. Our study in 32 is the first to address the plausibility of shadow banning on a major online platform, by adopting both a statistical and a graph topological approach. We first conduct an extensive data collection and analysis campaign, gathering occurrences of visibility limitations on user profiles (we crawl more than 2.5 millions of them). In such a black-box observation setup, we highlight the salient user profile features that may explain a banning practice (using machine learning predictors. We then pose two hypotheses for the phenomenon: i) limitations are bugs, as claimed by Twitter, and ii) shadow banning propagates as an epidemic on user interaction ego-graphs. We show that hypothesis i) is statistically unlikely with regards to the data we collected. We then show some interesting correlation with hypothesis ii), suggesting that the interaction topology is a good indicator of the presence of groups of shadow banned users on the service.

Join work with Benoît Morgan (IRIT-ACADIE) and Gilles Trédan (LAAS/CNRS).

7.2.4 zoNNscan: A boundary-entropy index for zone inspection of neural models

Participants: Erwan Le Merrer.

The training of deep neural network classifiers results in decision boundaries whose geometry is still not well understood. This is in direct relation with classification problems such as so-called corner case inputs. We introduce zoNNscan in 31, an index that is intended to inform on the boundary uncertainty (in terms of the presence of other classes) around one given input datapoint. It is based on confidence entropy, and is implemented through Monte Carlo sampling in the multidimensional ball surrounding that input. We detail the zoNNscan index, give an algorithm for approximating it, and finally illustrate its benefits on three applications.

Join work with Adel Jaouen (Technicolor).

7.2.5 Simple, Efficient and Convenient Decentralized Multi-Task Learning for Neural Networks

Participants: Amaury Bouchra-Pilet, Davide Frey, Francois Taiani.

7.2.6 AUCCCR: Agent Utility Centered Clustering for Cooperation Recommendation

Participants: Amaury Bouchra-Pilet, Davide Frey, Francois Taiani.

Providing recommendation to agents (e.g. people or organizations) regarding whom they should collaborate with in order to reach some objective is a recurring problem in a wide range of domains. It can be useful for instance in the context of collaborative machine learning, grouped purchases, and group holidays. This problem has been modeled by hedonic games, but this generic formulation cannot easily be used to provide efficient algorithmic solutions. In this work, we defined a class of hedonic games that allows us to provide an algorithmic solution to a wide class of collaboration recommendation problems by means of a clustering algorithm. We evaluated our algorithm, theoretically and experimentally, and show that it performs better than other well-known clustering algorithms in this context. We published this work at NETYS 2021 24.

7.2.7 Cluster-and-Conquer: When Randomness Meets Graph Locality

Participants: George Giakkoupis, Francois Taiani.

K-Nearest-Neighbors (KNN) graphs are central to many emblematic data mining and machine-learning applications. Some of the most efficient KNN graph algorithms are incremental and local: they start from a random graph, which they incrementally improve by traversing neighbors-of-neighbors links. Unfortunately, the initial random graph exhibits a poor graph locality, leading to many unnecessary similarity computations. In this work 30, we remove this drawback with Cluster-and-Conquer (C2 for short). Cluster-and-Conquer boosts the starting configuration of greedy algorithms thanks to a novel lightweight clustering mechanism, dubbed FastRandomHash. FastRandomHash leverages randomness and recursion to pre-cluster similar nodes at a very low cost. Our extensive evaluation on real datasets shows that Cluster-and-Conquer significantly outperforms existing approaches, including LSH, yielding speedups of up to ×4.42 and even improving the KNN quality.

This is a joint work with Anne-Marie Kermarrec from EPFL and Olivier Ruas from the SPIRALS team at Inria Lille.

7.3.1 Self-stabilizing clock synchronization with 1-bit messages

Participants: George Giakkoupis.

In 22 we study the fundamental problem of distributed clock synchronization in a basic probabilistic communication setting. We consider a synchronous fully-connected network of $n$ agents, where each agent has a local clock, that is, a counter increasing by one modulo $T$ in each round. The clocks have arbitrary values initially, and they must all indicate the same time eventually. We assume a pull communication model, where in every round each agent receives an $\ell$-bit message from a random agent. We devise several fast synchronization algorithms that use small messages and are self-stabilizing, that is, the complete initial state of each agent (not just its clock value) can be arbitrary.

We first provide a surprising algorithm for synchronizing a binary clock ($T=2$) using 1-bit messages ($\ell =1$). This is a variant of the voter model and converges in $O(logn)$ rounds w.h.p., unlike the voter model which needs polynomial time. Next we present an elegant extension of our algorithm that synchronizes a modulo $T=4$ clock, with $\ell =1$, in $O(logn)$ rounds. Using these two algorithms, we refine an algorithm of Boczkowski et al. (SODA'17), that synchronizes a modulo $T$ clock in polylogarithmic time (in $n$ and $T$). The original algorithm uses $\ell =3$ bit messages, and each agent receives messages from two agents per round. Our algorithm reduces the message size to $\ell =2$, and the number of messages received to one per round, without increasing the running time. Finally, we present two algorithms that simulate our last algorithm achieving $\ell <2$, without hurting the asymptotic running time. The first algorithm uses a message space of size 3, i.e., $\ell ={log}_2\left(3\right)$. The second requires a rough upper bound on $logn$, and uses just 1-bit messages. More generally, our constructions can simulate any self-stabilizing algorithm that requires a shared clock, without increasing the message size and by only increasing the running time by a constant factor and a polylogarithmic term.

Join work with Paul Bastide (ENS Rennes and WIDE Inria) and Hayk Saribekyan (U. Cambridge, UK).

7.3.2 Search via parallel Lévy walks

Participants: George Giakkoupis.

Motivated by the Lévy foraging hypothesis – the premise that various animal species have adapted to follow Lévy walks to optimize their search efficiency – we study in 26 the parallel hitting time of Lévy walks on the infinite two-dimensional grid. We consider $k$ independent discrete-time Lévy walks, with the same exponent $\alpha \in (1,\infty )$, that start from the same node, and analyze the number of steps until the first walk visits a given target at distance $\ell$. We show that for any choice of $k$ and $\ell$ from a large range, there is a unique optimal exponent ${\alpha }_{k,\ell }\in (2,3)$, for which the hitting time is $\tilde{O}({\ell }^2/k)$ w.h.p., while modifying the exponent by any constant term $ϵ>0$ increases the hitting time by a factor polynomial in $\ell$, or the walks fail to hit the target almost surely. Based on that, we propose a surprisingly simple and effective parallel search strategy, for the setting where $k$ and $\ell$ are unknown: The exponent of each Lévy walk is just chosen independently and uniformly at random from the interval $(2,3)$. This strategy achieves optimal search time (modulo polylogarithmic factors) among all possible algorithms (even centralized ones that know $k$). Our results should be contrasted with a line of previous work showing that the exponent $\alpha =2$ is optimal for various search problems. In our setting of $k$ parallel walks, we show that the optimal exponent depends on $k$ and $\ell$, and that randomizing the choice of the exponents works simultaneously for all $k$ and $\ell$.

Join work with Andrea Clementi (U. Rome “Tor Vergata”), Francesco d'Amore (CNRS Sophia Antipolis), and Emanuele Natale (CNRS Sophia Antipolis).

7.4.1 Efficient randomized DCAS

Participants: George Giakkoupis.

Double Compare-And-Swap (DCAS) is a tremendously useful synchronization primitive, which is also notoriously difficult to implement efficiently from objects that are provided by hardware. In 29 we present a randomized implementation of DCAS with $O(logn)$ expected amortized step complexity against the oblivious adversary, where $n$ is the number of processes in the system. This is the only algorithm to-date that achieves sub-linear step complexity. We achieve that by first implementing two novel algorithms as building blocks. One is a mechanism that allows processes to repeatedly agree on a random value among multiple proposed ones, and the other one is a restricted bipartite version of DCAS.

Join work with Mehrdad Jafari Giv (U. Calgary) and Philipp Woelfel (U. Calgary).

7.4.2 On the weakest information on failures to solve mutual exclusion and consensus in asynchronous crash-prone read/write systems

Participants: Michel Raynal.

Mutual exclusion and consensus are among the most important coordination problems encountered in asynchronous concurrent systems, whether processes communicate using read/write registers or message passing. Unfortunately, neither can be solved in crash-prone systems, as soon as even a single process may crash. Hence, an important question: which is the weakest information on failures that must be given to the processes so that these problems can be solved whatever the number of crashes. This approach to circumvent impossibility results is known under the name failure detectors. Considering mutual exclusion and consensus in a crash-prone asynchronous system where the processes communicate through read/write registers, we answer in 15 the previous question by presenting two failure detectors. The first, called Quasi-Perfect ($QP$) allows mutual exclusion to be solved in the presence of any number of process crashes. The second, called ${\Omega }^{*}$, allows consensus to be solved in the general model where not all but an a priori unknown subset of processes participate in consensus. In addition to algorithms solving each of the previous problems with the help of the associated failure detector, the article shows that $QP$ and ${\Omega }^{*}$ provide the weakest information on failures needed to solve mutual exclusion and participant-restricted consensus respectively.

Join work with Carole Delporte-Gallet and Hugues Fauconnier from Inria Paris.

7.4.3 Set-constrained delivery broadcast: A communication abstraction for read/write implementable distributed objects

Participants: Michel Raynal.

In 18 we introduce a new communication abstraction, called Set-Constrained Delivery Broadcast (SCD-broadcast), whose aim is to provide its users with an appropriate abstraction level when they have to implement objects or distributed tasks in an asynchronous message-passing system prone to process crash failures. This abstraction allows each process to broadcast messages and deliver a sequence of sets of messages in such a way that, if a process delivers a set of messages including a message m and later delivers a set of messages including a message, no process delivers first a set of messages including and later a set of message including m. After having presented an algorithm implementing SCD-broadcast, we investigate its programming power and its computability limits. On the “power” side we present SCD-broadcast-based algorithms, which are both simple and efficient, building objects (such as snapshot and conflict-free replicated data types), and distributed tasks. On the “computability limits” side we show that SCD-broadcast and read/write registers are computationally equivalent.

Join work with Damien Imbs (U. Aix-Marseille), Achour Mostéfaoui (U. Nantes), and Matthieu Perrin (U. Nantes).

7.5.1 Coronasurveys

Participants: Davide Frey.

CIFRE InterDigital: Distributed troubleshooting of edge-compute functions (2018-2021)

Participants: Loïck Bonniot, François Taïani.

This project seeks to explore how recent generations of end-user gateways (or more generally end-user devices) could implement an edge-compute paradigm powered by user-side micro-services. Our vision is that the devices distributed among the homes of end-users will expose (as a service) their computing power and their ability to quickly deploy compute functions in an execution environment. In order for service and application providers to actually use the system and deploy applications, the system must however ensure an appropriate level of reliability, while simultaneously requiring a very low level of maintenance in order to address the typical size and economics of gateway deployments (at least a few tens of million units). Providing a good level of reliability in such a large system at a reasonable cost is unfortunately difficult. To address this challenge, we aim in this thesis to exploit the natural distribution of such large-scale user-side device deployments to quickly pinpoint problems and troubleshoot applications experiencing performance degradations.

This project is in collaboration with Christoph Neumann from InterDigital.

9.1.1 Associate Teams in the framework of an Inria International Lab or in the framework of an Inria International Program

9.2.1 FP7 & H2020 projects

ANR Project PAMELA (2016-2022)

Participants: Davide Frey, George Giakkoupis, François Taïani.

PAMELA is a collaborative ANR project involving Inria/IRISA, Inria Lille (MAGNET team), UMPC, Mediego and Snips. The project aims at developing machine learning theories and algorithms in order to learn local and personalized models from data distributed over networked infrastructures. This project seeks to provide fundamental answers to modern information systems built by interconnecting many personal devices holding private user data in the search of personalized suggestions and recommendations. A significant asset of the project is the quality of its industrial partners, Snips and Mediego, who bring in their expertise in privacy protection and distributed computing as well as use cases and datasets.

ANR Project OBrowser (2016-2021)

Participants: David Bromberg, Davide Frey, François Taïani.

OBrowser is a collaborative ANR project involving Inria, the University of Nantes, the University of South Brittany, and Orange. The project emerges from the vision of designing and deploying distributed applications on millions of machines using web-enabled technologies without relying on a cloud or a central authority. OBrowser proposes to build collaborative applications through a decentralized execution environment composed of users' browsers that autonomously manages issues such as communication, naming, heterogeneity, and scalability.

ANR Project DESCARTES (2016-2021)

Participants: George Giakkoupis, Michel Raynal, François Taïani.

DESCARTES is a collaborative ANR project involving Inria/IRISA, Labri (U. Bordeaux), IRIF (U. Paris Diderot), Inria Paris (GANG Team), Vérimag (Grenoble), LIF (Marseilles), and LS2N (former LINA, Nantes). The DESCARTES project aims at bridging the lack of a generic theoretical framework in order to unify the large body of fundamental knowledge on distributed computation that has been acquired over the last 40 years. In particular, the project's objective is to develop a systematic model of distributed computation that organizes the functionalities of a distributed computing system into reusable modular constructs assembled via well-defined mechanisms that maintain sound theoretical guarantees on the resulting system.

Collaboration with the Conseil Supérieur de l'Audiovisuel (CSA)

Participants: Erwan Le Merrer.

Collaborating with the CSA (the French regulator of audio-visual content) on the YouTube recommender system, in order to assess the presence or not of the so called rabbit-hole phenomenon.

ANR Project ByBloS (2021-2025)

Participants: George Giakkoupis, Michel Raynal, Davide Frey, David Bromberg, François Taïani, Timothé Albouy.

Partners: IRISA (coordinator, U. Rennes I) in Rennes, LIRIS (INSA Lyon) in Lyon, and LS2N (Université de Nantes) in Nantes. Budget: 252 220€

PriCLeSS (Labex CominLabs and Brittany Region) (2021-2024)

Participants: Michel Raynal, Davide Frey, François Taïani, Arthur Rauch.

10.1.1 Scientific events: organisation

10.1.2 Scientific events: selection

10.1.3 Invited talks

Participants: George Giakkoupis, Barbe Thystere Mvondo Djob.

• Barbe Thystere Mvondo Djob. Cloud Computing (virtualisation) Evoluer avec les nouveaux paradigmes. 4th edition of the Computer Science Research Days (JRI), Virtual, Burkina Faso, Nov 11 2021.
• Barbe Thystere Mvondo Djob. Virtualisation et Cloud Computing : Des technologies utiles pour l'enseignement et la recherche. RECIF SEMINAR, Virtual, University of, Oct 22 2021.
• Barbe Thystere Mvondo Djob. Heterogeneity across system layers. DiverSE Team, Rennes, Oct 14 2021.
• George Giakkoupis. Self-stabilizing clock synchronization with 1-bit messages. Algorithms, Dynamics, and Information Flow in Networks (ADYN-Seminar), Virtual, Germany, Mar 1 2021.
• George Giakkoupis. An optimal leader election population protocol. 2nd Workshop on Distributed Algorithms for Low-Functional Robots (WDALFR), Virtual, Nara, Japan, Jan 8 2021.

10.1.4 Scientific expertise

Participants: Davide Frey, Francois Taiani.

• Francois Taiani served as scientific expert for the call for proposals organized by IDEX Université Grenoble Alpes.
• Davide Frey served as scientific expert for CIFRE PhD Grants.
• Davide Frey served as scientific expert for the call for proposal in the context of the Flemish action plan on Cybersecurity organized by Flanders Innovation & Entrepreneurship (VLAIO).

10.1.5 Research administration

Participants: Erwan Le Merrer, Francois Taiani.

• Erwan Le Merrer serves on the scientific board of Inria's REGALIA since 2021.
• Erwan Le Merrer serves on the administrative board of the Société Informatique de France since 2019.
• Francois Taiani has served as Career Advice Person, (Référent conseil-parcours professionel chercheurs) for IRISA/Inria Rennes Bretagne Atlantique since 2019.

10.2.1 Teaching

• Master: Barbe Thystere Mvondo Djob, Network and Security for IOT, 30h, ESIR M1, Rennes, France
• Master: Barbe Thystere Mvondo Djob, Cloud Computing for IOT, 26h, ESIR M2, Rennes, France
• Bachelor: George Giakkoupis, Distributed Algorithms, 9h, L3 parcours SI, ISTIC, ENS Rennes, France.
• Master: George Giakkoupis, Distributed Systems (Systèmes Répartis), 9h, ENSAI Rennes, France.
• Master: Erwan Le Merrer, Network Science, 12h, ESIR Rennes, France.
• Engineering School: Francois Taiani, Synchronization and Parallel Programming, 30h, 2nd year of Engineering School (M1), ESIR / U. Rennes I, France.
• Engineering School: Francois Taiani, Distributed Systems, 12h, 3rd year of Engineering School (M2), ESIR / U. Rennes I, France.
• Engineering School: François Taiani, Introduction to Operating Systems, 24h, 1st year of Engineering School (L3), ESIR / U. Rennes I, France.
• Master: Davide Frey, Scalable Distributed Systems, 10h, M1, EIT/ICT Labs Master School, U. Rennes I, France.
• ENS L3 : Davide Frey, Distributed Algorithms, 12h, ENS Rennes, France.
• Master: Davide Frey, Cloud Computing, 6h, M2-MIAGE, U. Rennes I, France.
• Master: Davide Frey, Distributed Systems/Systèmes Répartis, 12h, ENSAI, France.
• Master: Davide Frey, Apprentice Tutoring, 16h ETD, M2 Alternance U. Rennes I, France.

10.2.2 Supervision

• PhD: Loïck Bonniot, Distributed Troubleshooting of Edge-Compute Functions, U. Rennes I, supervised by François Taïani and Christoph Neumann (Interdigital, CIFRE), defended on Jun 10 2021 37.
• PhD: Hayk Saribekyan, Information Dissemination via Random Walks, University of Cambridge, UK, Sep 24 2021, supervised by George Giakkoupis and Thomas Sauerwald (U. Cambridge)  70.
• PhD: Amaury Bouchra Pilet, Robust and Lightweight Overlay Management for Decentralized Learning, U. Rennes I, supervised by Francois Taiani and Davide Frey, defended on Nov 10 2021.
• PhD: Alex Auvolat, Towards probabilistic decentralized systematic design for large-scale privacy-preserving collaborative systems, U. Rennes I, supervised by François Taïani and David Bromberg, defended on Dec 2 2021.
• PhD in progress: Thibault Maho, On the security of neural networks, since 2020, supervised by Erwan Le Merrer and Teddy Furon (Linkmedia team).
• PhD in progress: Timothé Albouy, Towards Lightweight Scalable and Open Byzantine-Fault-Tolerant Distributed Objects, U. Rennes I, supervised by Francois Taiani and Davide Frey, started on Oct 18 2021.
• PhD in progress: Mathieu Gestin, Private Authenticated Storage for Online Services, Inria Rennes, supervised by Davide Frey, started on Oct 1 2021.
• PhD in progress: Arthur Rauch, Frugal and Legal for Future Blockchain, Inria Rennes, supervised by Emmanuelle Anceaume and Davide Frey, started on Oct 1 2021.

10.2.3 Juries

• Francois Taiani was chairman for the PhD thesis of Loic Guegan: Scalable end-to-end models for the time and energy performance of Fog infrastructures, ENS Rennes, Jan 29 2021.
• Francois Taiani was external reviewer for the PhD thesis of Marinus Abraham De Vos: Decentralization and Disintermediation in Blockchain-based Marketplaces, TU Delft (NL), Jun 16 2021.
• Francois Taiani was external reviewer for the PhD thesis of Stefanos Peros: Software Services to Simplify Time Management for IoT Application Development, KU Leuven (Belgium), Sep 10 2021.
• Francois Taiani was chairman for the PhD thesis of Grégoire Bonnin: Structures des données pour environnements distribués à grande échelle, University of Nantes, Nov 17 2021.
• Francois Taiani was chairman for the PhD thesis of Hamidreza Arkian: Resource Management for Data Stream Processing in Geo-Distributed Environments, U. Rennes I, Dec 7 2021.
• Francois Taiani was reviewer for the HDR thesis of Claudia-Lavinia Ignat: Large-scale trustworthy distributed collaboration, Université de Lorraine, Apr 23 2021.
• Davide Frey is an external reviewer for the PhD thesis of Hayman Salih MOHAMMED, Università di Roma, La Sapienza.

10.3.1 Articles and contents

• Erwan Le Merrer wrote an article in Interstices: "Le problème du videur : la crédibilité des explications de l’IA en question" 41.

10.3.2 Education

• Michel Raynal wrote an article entitled “Distributed Computability: A Few Results Masters Students Should Know” that appeared to the ACM SIGACT News 19.

International journals

• 13 articleA.Alex Auvolat, D.Davide Frey, M.Michel Raynal and F.François Taïani. Byzantine-Tolerant Causal Broadcast.Theoretical Computer Science885September 2021, 55-68
  HALDOIback to text
• 14 articleR.-A.Ronan-Alexandre Cherrueau, M.Marie Delavergne, A.Alexandre Van Kempen, A.Adrien Lebre, D.Dimitri Pertin, J.Javier Rojas Balderrama, A.Anthony Simonet and M.Matthieu Simonin. EnosLib: A Library for Experiment-Driven Research in Distributed Computing.IEEE Transactions on Parallel and Distributed SystemsSeptember 2021, 1-15
  HALDOI
• 15 articleC.Carole Delporte-Gallet, H.Hugues Fauconnier and M.Michel Raynal. On the weakest information on failures to solve mutual exclusion and consensus in asynchronous crash-prone read/write systems.Journal of Parallel and Distributed Computing153July 2021, 110-118
  HALDOIback to text
• 16 articleD.Davide Frey, C.Carlos Baquero, P.Paolo Casari, A.Antonio Fernandez Anta, A.Amanda García-García, A.Augusto Garcia-Agundez, C.Chryssis Georgiou, B.Benjamin Girault, A.Antonio Ortega, M.Mathieu Goessens, H.Harold Hernández-Roig, N.Nicolas Nicolaou, E.Efstathios Stavrakis, O.Oluwasegun Ojo, J.Julian Roberts and I.Ignacio Sanchez. The CoronaSurveys System for COVID-19 Incidence Data Collection and Processing.Frontiers in Computer Science3June 2021
  HALDOIback to text
• 17 articleD.Davide Frey, A.Augusto Garcia-Agundez, O.Oluwasegun Ojo, H.Harold Hernández-Roig, C.Carlos Baquero, C.Chryssis Georgiou, M.Mathieu Goessens, R.Rosa Lillo, R.Raquel Menezes, N.Nicolas Nicolaou, A.Antonio Ortega, E.Efstathios Stavrakis and A.Antonio Fernandez Anta. Estimating the COVID-19 Prevalence in Spain With Indirect Reporting via Open Surveys.Frontiers in Public Health9April 2021
  HALDOIback to text
• 18 articleD.Damien Imbs, A.Achour Mostéfaoui, M.Matthieu Perrin and M.Michel Raynal. Set-constrained delivery broadcast: A communication abstraction for read/write implementable distributed objects.Theoretical Computer Science8862021, 49-68
  HALDOIback to text
• 19 articleM.Michel Raynal. Distributed Computability: A Few Results Masters Students Should Know.ACM SIGACT News522June 2021, 92-110
  HALDOIback to text
• 20 articleM.Michel Raynal. On the Versatility of Bracha’s Byzantine Reliable Broadcast Algorithm.Parallel Processing Letters3103September 2021, 1-7
  HALDOIback to text
• 21 articleJ.José Suárez-Varela, M.Miquel Ferriol-Galmés, A.Albert López, P.Paul Almasan, G.Guillermo Bernárdez, D.David Pujol-Perich, K.Krzysztof Rusek, L.Loïck Bonniot, C.Christoph Neumann, F.François Schnitzler, F.François Taïani, M.Martin Happ, C.Christian Maier, J. L.Jia Lei Du, M.Matthias Herlich, P.Peter Dorfinger, N. V.Nick Vincent Hainke, S.Stefan Venz, J.Johannes Wegener, H.Henrike Wissing, B.Bo Wu, S.Shihan Xiao, P.Pere Barlet-Ros and A.Albert Cabellos-Aparicio. The graph neural networking challenge: a worldwide competition for education in AI/ML for networks.Computer Communication Review513July 2021, 9-16
  HALDOIback to text

International peer-reviewed conferences

• 22 inproceedingsP.Paul Bastide, G.George Giakkoupis and H.Hayk Saribekyan. Self-Stabilizing Clock Synchronization with 1-bit Messages.SODA 2021 - ACM-SIAM Symposium on Discrete AlgorithmsAlexandria, VA, United StatesACMJanuary 2021, 1-27
  HALDOIback to text
• 23 inproceedingsL.Loïck Bonniot, C.Christoph Neumann and F.François Taïani. Towards Internet-Scale Convolutional Root-Cause Analysis with DiagNet.IPDPS 2021 - The 35th IEEE International Parallel and Distributed Processing SymposiumPortland / Virtual, United StatesMay 2021
  HALDOIback to text
• 24 inproceedingsA.Amaury Bouchra Pilet, D.Davide Frey and F.François Taïani. AUCCCR: Agent Utility Centered Clustering for Cooperation Recommendation.NETYS 2021 - 9th International Conference on NETworked sYStems12754AUCCCR: Agent Utility Centered Clustering for Cooperation RecommendationMarrakech, MoroccoSpringerMay 2021
  HALDOIback to text
• 25 inproceedingsA.Amaury Bouchra Pilet, D.Davide Frey and F.François Taïani. Simple, Efficient and Convenient Decentralized Multi-Task Learning for Neural Networks.IDA 2021 - 19th Symposium on Intelligent Data Analysis12695Lecture Notes in Computer SciencePorto, PortugalSpringerApril 2021
  HALDOIback to text
• 26 inproceedingsA.Andrea Clementi, F.Francesco d'Amore, G.George Giakkoupis and E.Emanuele Natale. Search via Parallel Lévy Walks on ${}^2$.PODC 2021 - ACM Symposium on Principles of Distributed ComputingSalerno, ItalyInria & Université Cote d'Azur, CNRS, I3S, Sophia Antipolis, France; Università degli Studi di Roma "Tor Vergata"; Univ Rennes, Inria, CNRS, IRISA, FranceApril 2020, 81–91
  HALDOIback to text
• 27 inproceedingsD.Davide Frey, T.Timothé Albouy, M.Michel Raynal and F.François Taïani. Byzantine-Tolerant Reliable Broadcast in the Presence of Silent Churn.Stabilization, Safety, and Security of Distributed SystemsSSS 2021 - 23rd International Symposium on Stabilizing, Safety, and Security of Distributed Systems13046Lecture Notes in Computer ScienceVirtual, FranceSpringer International PublishingNovember 2021, 21-33
  HALDOIback to text
• 28 inproceedingsD.Davide Frey, L.Lucie Guillou, M.Michel Raynal and F.François Taïani. Consensus-Free Ledgers When Operations of Distinct Processes are Commutative.PaCT 2021 - 16th International Conference on Parallel Computing Technologies12942LNCSKaliningrad, RussiaSpringerSeptember 2021, 359-370
  HALDOIback to text
• 29 inproceedingsG.George Giakkoupis, M.Mehrdad Jafari Giv and P.Philipp Woelfel. Efficient Randomized DCAS.STOC 2021 - 53rd Annual ACM SIGACT Symposium on Theory of ComputingRome (Virtual), ItalyACMJune 2021, 1-64
  HALDOIback to text
• 30 inproceedingsG.George Giakkoupis, A.-M.Anne-Marie Kermarrec, O.Olivier Ruas and F.François Taïani. Cluster-and-Conquer: When Randomness Meets Graph Locality.ICDE 2021 - IEEE 37th International Conference on Data EngineeringChania, GreeceIEEEApril 2021, 2027-2032
  HALDOIback to text
• 31 inproceedingsA.Adel Jaouen and E.Erwan Le Merrer. zoNNscan: a boundary-entropy index for zone inspection of neural models.MCS 2020 - Monte Carlo Search workshopMonte Carlos Search (MCS) workshop at IJCAIVirtual, JapanJanuary 2021, 1-8
  HALback to text
• 32 inproceedingsE.Erwan Le Merrer, B.Benoît Morgan and G.Gilles Trédan. Setting the Record Straighter on Shadow Banning.INFOCOM 2021 - IEEE International Conference on Computer CommunicationsVirtual, CanadaIEEEMay 2021, 1-10
  HALDOIback to text
• 33 inproceedingsT.Thibault Maho, B.Benoît Bonnet, T.Teddy Furon and E.Erwan Le Merrer. RoBIC: A benchmark suite for assessing classifiers robustness.ICIP 2021 - IEEE International Conference on Image ProcessingAnchorage, Alaska, United StatesIEEESeptember 2021, 1-5
  HALDOIback to text
• 34 inproceedingsT.Thibault Maho, T.Teddy Furon and E. L.Erwan Le Merrer. SurFree: a fast surrogate-free black-box attack.CVPR 2021 - Conference on Computer Vision and Pattern RecognitionProc. of {IEEE} Conference on Computer Vision and Pattern Recognition, {CVPR}Virtual, FranceJune 2021, 10430--10439
  HALback to text
• 35 inproceedingsD.Djob Mvondo, A.Antonio Barbalace, A.Alain Tchana and G.Gilles Muller. Tell me when you are sleepy and what may wake you up!SoCC '21: ACM Symposium on Cloud ComputingSeattle WA USA, United StatesACMNovember 2021, 562-569
  HALDOIback to text

Conferences without proceedings

• 36 inproceedingsA.Antonin Voyez, T.Tristan Allard, G.Gildas Avoine, P.Pierre Cauchois, E.Elisa Fromont and M.Matthieu Simonin. Attaque par inférence d'appartenance sur des séries temporelles agrégées en utilisant la programmation par contraintes.BDA 2021 - 37ème Conférence sur la Gestion de Données – Principes, Technologies et ApplicationsParis, FranceOctober 2021, 1
  HAL

Doctoral dissertations and habilitation theses

• 37 thesisL.Loïck Bonniot. Computer network modeling and root cause analysis with statistical learning.Université Rennes 1June 2021
  HALback to text
• 38 thesisQ.Quentin Dufour. High-throughput real-time onion networks to protect everyone’s privacy.Université Rennes 1February 2021
  HAL

Reports & preprints

• 39 miscA.Alex Auvolat, Y.-D.Yérom-David Bromberg, D.Davide Frey and F.François Taïani. $BASALT$: A Rock-Solid Foundation for Epidemic Consensus Algorithms in Very Large, Very Open Networks.February 2021
  HAL
• 40 miscO.Oluwasegun Ojo, A.Augusto García-Agundez, B.Benjamin Girault, H.Harold Hernández, E.Elisa Cabana, A.Amanda García-García, P.Payman Arabshahi, C.Carlos Baquero, P.Paolo Casari, E. J.Ednaldo José Ferreira, D.Davide Frey, C.Chryssis Georgiou, M.Mathieu Goessens, A.Anna Ishchenko, E.Ernesto Jiménez, O.Oleksiy Kebkal, R.Rosa Lillo, R.Raquel Menezes, N.Nicolas Nicolaou, A.Antonio Ortega, P.Paul Patras, J. C.Julian C Roberts, E.Efstathios Stavrakis, Y.Yuichi Tanaka and A. F.Antonio Fernández Anta. CoronaSurveys: Using Surveys with Indirect Reporting to Estimate the Incidence and Evolution of Epidemics.February 2021
  HAL

Scientific popularization

• 41 articleE.Erwan Le Merrer and G.Gilles Trédan. Le problème du videur : la crédibilité des explications de l’IA en question.IntersticesJuly 2021
  HALback to text