arXiv subjects

Rachid Guerraoui

Publications and source records attributed to Rachid Guerraoui.

At least 73 records · Page 4Linked to original sources

arXiv2022

An Equivalence Between Data Poisoning and Byzantine Gradient Attacks

To study the resilience of distributed learning, the "Byzantine" literature considers a strong threat model where workers can report arbitrary gradients to the parameter server. Whereas this model helped obtain several fundamental results, it has sometimes been considered unrealistic, when the workers are mostly trustworthy machines. In this paper, we show a surprising equivalence between this model and data poisoning, a threat considered much more realistic. More specifically, we prove that every gradient attack can be reduced to data poisoning, in any personalized federated learning system with PAC guarantees (which we show are both desirable and realistic). This equivalence makes it possible to obtain new impossibility results on the resilience of any "robust" learning algorithm to data poisoning in highly heterogeneous applications, as corollaries of existing impossibility theorems on Byzantine machine learning. Moreover, using our equivalence, we derive a practical attack that we show (theoretically and empirically) can be very effective against classical personalized federated learning models.

cs.LG↗

arXiv2022

Byzantine Machine Learning Made Easy by Resilient Averaging of Momentums

Byzantine resilience emerged as a prominent topic within the distributed machine learning community. Essentially, the goal is to enhance distributed optimization algorithms, such as distributed SGD, in a way that guarantees convergence despite the presence of some misbehaving (a.k.a., {\em Byzantine}) workers. Although a myriad of techniques addressing the problem have been proposed, the field arguably rests on fragile foundations. These techniques are hard to prove correct and rely on assumptions that are (a) quite unrealistic, i.e., often violated in practice, and (b) heterogeneous, i.e., making it difficult to compare approaches. We present \emph{RESAM (RESilient Averaging of Momentums)}, a unified framework that makes it simple to establish optimal Byzantine resilience, relying only on standard machine learning assumptions. Our framework is mainly composed of two operators: \emph{resilient averaging} at the server and \emph{distributed momentum} at the workers. We prove a general theorem stating the convergence of distributed SGD under RESAM. Interestingly, demonstrating and comparing the convergence of many existing techniques become direct corollaries of our theorem, without resorting to stringent assumptions. We also present an empirical evaluation of the practical relevance of RESAM.

cs.LG↗

arXiv2022

Strategyproof Learning: Building Trustworthy User-Generated Datasets

We prove in this paper that, perhaps surprisingly, incentivizing data misreporting is not a fatality. By leveraging a careful design of the loss function, we propose Licchavi, a global and personalized learning framework with provable strategyproofness guarantees. Essentially, we prove that no user can gain much by replying to Licchavi's queries with answers that deviate from their true preferences. Interestingly, Licchavi also promotes the desirable "one person, one unit-force vote" fairness principle. Furthermore, our empirical evaluation of its performance showcases Licchavi's real-world applicability. We believe that our results are critical for the safety of any learning scheme that leverages user-generated data.

cs.LG↗

arXiv2021

Collaborative Learning in the Jungle (Decentralized, Byzantine, Heterogeneous, Asynchronous and Nonconvex Learning)

We study Byzantine collaborative learning, where $n$ nodes seek to collectively learn from each others' local data. The data distribution may vary from one node to another. No node is trusted, and $f < n$ nodes can behave arbitrarily. We prove that collaborative learning is equivalent to a new form of agreement, which we call averaging agreement. In this problem, nodes start each with an initial vector and seek to approximately agree on a common vector, which is close to the average of honest nodes' initial vectors. We present two asynchronous solutions to averaging agreement, each we prove optimal according to some dimension. The first, based on the minimum-diameter averaging, requires $ n \geq 6f+1$, but achieves asymptotically the best-possible averaging constant up to a multiplicative constant. The second, based on reliable broadcast and coordinate-wise trimmed mean, achieves optimal Byzantine resilience, i.e., $n \geq 3f+1$. Each of these algorithms induces an optimal Byzantine collaborative learning protocol. In particular, our equivalence yields new impossibility theorems on what any collaborative learning algorithm can achieve in adversarial and heterogeneous environments.

cs.LG↗

arXiv2021

Differential Privacy and Byzantine Resilience in SGD: Do They Add Up?

This paper addresses the problem of combining Byzantine resilience with privacy in machine learning (ML). Specifically, we study if a distributed implementation of the renowned Stochastic Gradient Descent (SGD) learning algorithm is feasible with both differential privacy (DP) and $(α,f)$-Byzantine resilience. To the best of our knowledge, this is the first work to tackle this problem from a theoretical point of view. A key finding of our analyses is that the classical approaches to these two (seemingly) orthogonal issues are incompatible. More precisely, we show that a direct composition of these techniques makes the guarantees of the resulting SGD algorithm depend unfavourably upon the number of parameters of the ML model, making the training of large models practically infeasible. We validate our theoretical results through numerical experiments on publicly-available datasets; showing that it is impractical to ensure DP and Byzantine resilience simultaneously.

cs.LG↗

arXiv2021

Velos: One-sided Paxos for RDMA applications

Modern data centers are becoming increasingly equipped with RDMA-capable NICs. These devices enable distributed systems to rely on algorithms designed for shared memory. RDMA allows consensus to terminate within a few microsecond in failure-free scenarios, yet, RDMA-optimized algorithms still use expensive two-sided operations in case of failure. In this work, we present a new leader-based consensus algorithm that relies solely on one-sided RDMA verbs. Our algorithm is based on Paxos, it decides in a single one-sided RDMA operation in the common case, and changes leader also in a single one-sided RDMA operation in case of failure. We implement our algorithm in the form of an SMR system named Velos, and we evaluated our system against the state-of-the-art competitor Mu. Compared to Mu, our solution adds a small overhead of approximately 0.6 microseconds in failure-free executions and shines during failover periods during which it is 13 times faster in changing leader.

cs.DC↗

arXiv2021

Differentially Private Stochastic Coordinate Descent

In this paper we tackle the challenge of making the stochastic coordinate descent algorithm differentially private. Compared to the classical gradient descent algorithm where updates operate on a single model vector and controlled noise addition to this vector suffices to hide critical information about individuals, stochastic coordinate descent crucially relies on keeping auxiliary information in memory during training. This auxiliary information provides an additional privacy leak and poses the major challenge addressed in this work. Driven by the insight that under independent noise addition, the consistency of the auxiliary information holds in expectation, we present DP-SCD, the first differentially private stochastic coordinate descent algorithm. We analyze our new method theoretically and argue that decoupling and parallelizing coordinate updates is essential for its utility. On the empirical side we demonstrate competitive performance against the popular stochastic gradient descent alternative (DP-SGD) while requiring significantly less tuning.

cs.LG↗

arXiv2021

The Impact of RDMA on Agreement

Remote Direct Memory Access (RDMA) is becoming widely available in data centers. This technology allows a process to directly read and write the memory of a remote host, with a mechanism to control access permissions. In this paper, we study the fundamental power of these capabilities. We consider the well-known problem of achieving consensus despite failures, and find that RDMA can improve the inherent trade-off in distributed computing between failure resilience and performance. Specifically, we show that RDMA allows algorithms that simultaneously achieve high resilience and high performance, while traditional algorithms had to choose one or another. With Byzantine failures, we give an algorithm that only requires $n \geq 2f_P + 1$ processes (where $f_P$ is the maximum number of faulty processes) and decides in two (network) delays in common executions. With crash failures, we give an algorithm that only requires $n \geq f_P + 1$ processes and also decides in two delays. Both algorithms tolerate a minority of memory failures inherent to RDMA, and they provide safety in asynchronous systems and liveness with standard additional assumptions.

cs.DC↗

arXiv2021

Fast and Robust Distributed Learning in High Dimension

Could a gradient aggregation rule (GAR) for distributed machine learning be both robust and fast? This paper answers by the affirmative through multi-Bulyan. Given $n$ workers, $f$ of which are arbitrary malicious (Byzantine) and $m=n-f$ are not, we prove that multi-Bulyan can ensure a strong form of Byzantine resilience, as well as an ${\frac{m}{n}}$ slowdown, compared to averaging, the fastest (but non Byzantine resilient) rule for distributed machine learning. When $m \approx n$ (almost all workers are correct), multi-Bulyan reaches the speed of averaging. We also prove that multi-Bulyan's cost in local computation is $O(d)$ (like averaging), an important feature for ML where $d$ commonly reaches $10^9$, while robust alternatives have at least quadratic cost in $d$. Our theoretical findings are complemented with an experimental evaluation which, in addition to supporting the linear $O(d)$ complexity argument, conveys the fact that multi-Bulyan's parallelisability further adds to its efficiency.

cs.DC↗

arXiv2020

Garfield: System Support for Byzantine Machine Learning

We present Garfield, a library to transparently make machine learning (ML) applications, initially built with popular (but fragile) frameworks, e.g., TensorFlow and PyTorch, Byzantine-resilient. Garfield relies on a novel object-oriented design, reducing the coding effort, and addressing the vulnerability of the shared-graph architecture followed by classical ML frameworks. Garfield encompasses various communication patterns and supports computations on CPUs and GPUs, allowing addressing the general question of the very practical cost of Byzantine resilience in SGD-based ML applications. We report on the usage of Garfield on three main ML architectures: (a) a single server with multiple workers, (b) several servers and workers, and (c) peer-to-peer settings. Using Garfield, we highlight several interesting facts about the cost of Byzantine resilience. In particular, (a) Byzantine resilience, unlike crash resilience, induces an accuracy loss, (b) the throughput overhead comes more from communication than from robust aggregation, and (c) tolerating Byzantine servers costs more than tolerating Byzantine workers.

cs.LG↗

arXiv2020

FLeet: Online Federated Learning via Staleness Awareness and Performance Prediction

Federated Learning (FL) is very appealing for its privacy benefits: essentially, a global model is trained with updates computed on mobile devices while keeping the data of users local. Standard FL infrastructures are however designed to have no energy or performance impact on mobile devices, and are therefore not suitable for applications that require frequent (online) model updates, such as news recommenders. This paper presents FLeet, the first Online FL system, acting as a middleware between the Android OS and the machine learning application. FLeet combines the privacy of Standard FL with the precision of online learning thanks to two core components: (i) I-Prof, a new lightweight profiler that predicts and controls the impact of learning tasks on mobile devices, and (ii) AdaSGD, a new adaptive learning algorithm that is resilient to delayed updates. Our extensive evaluation shows that Online FL, as implemented by FLeet, can deliver a 2.3x quality boost compared to Standard FL, while only consuming 0.036% of the battery per day. I-Prof can accurately control the impact of learning tasks by improving the prediction accuracy up to 3.6x (computation time) and up to 19x (energy). AdaSGD outperforms alternative FL approaches by 18.4% in terms of convergence speed on heterogeneous data.

cs.LG↗

arXiv2020

Dynamic Byzantine Reliable Broadcast [Technical Report]

Reliable broadcast is a communication primitive guaranteeing, intuitively, that all processes in a distributed system deliver the same set of messages. The reason why this primitive is appealing is twofold: (i) we can implement it deterministically in a completely asynchronous environment, unlike stronger primitives like consensus and total-order broadcast, and yet (ii) reliable broadcast is powerful enough to implement important applications like payment systems. The problem we tackle in this paper is that of dynamic reliable broadcast, i.e., enabling processes to join or leave the system. This property is desirable for long-lived applications (aiming to be highly available), yet has been precluded in previous asynchronous reliable broadcast protocols. We study this property in a general adversarial (i.e., Byzantine) environment. We introduce the first specification of a dynamic Byzantine reliable broadcast (DBRB) primitive that is amenable to an asynchronous implementation. We then present an algorithm implementing this specification in an asynchronous network. Our DBRB algorithm ensures that if any correct process in the system broadcasts a message, then every correct process delivers that message unless it leaves the system. Moreover, if a correct process delivers a message, then every correct process that has not expressed its will to leave the system delivers that message. We assume that more than $2/3$ of processes in the system are correct at all times, which is tight in our context. We also show that if only one process in the system can fail---and it can fail only by crashing---then it is impossible to implement a stronger primitive, ensuring that if any correct process in the system broadcasts or delivers a message, then every correct process in the system delivers that message---including those that leave.

cs.DC↗

arXiv2020

Microsecond Consensus for Microsecond Applications

We consider the problem of making apps fault-tolerant through replication, when apps operate at the microsecond scale, as in finance, embedded computing, and microservices apps. These apps need a replication scheme that also operates at the microsecond scale, otherwise replication becomes a burden. We propose Mu, a system that takes less than 1.3 microseconds to replicate a (small) request in memory, and less than a millisecond to fail-over the system - this cuts the replication and fail-over latencies of the prior systems by at least 61% and 90%. Mu implements bona fide state machine replication/consensus (SMR) with strong consistency for a generic app, but it really shines on microsecond apps, where even the smallest overhead is significant. To provide this performance, Mu introduces a new SMR protocol that carefully leverages RDMA. Roughly, in Mu a leader replicates a request by simply writing it directly to the log of other replicas using RDMA, without any additional communication. Doing so, however, introduces the challenge of handling concurrent leaders, changing leaders, garbage collecting the logs, and more - challenges that we address in this paper through a judicious combination of RDMA permissions and distributed algorithmic design. We implemented Mu and used it to replicate several systems: a financial exchange app called Liquibook, Redis, Memcached, and HERD. Our evaluation shows that Mu incurs a small replication latency, in some cases being the only viable replication system that incurs an acceptable overhead.

cs.DC↗

arXiv2020

Efficient Multi-word Compare and Swap

Atomic lock-free multi-word compare-and-swap (MCAS) is a powerful tool for designing concurrent algorithms. Yet, its widespread usage has been limited because lock-free implementations of MCAS make heavy use of expensive compare-and-swap (CAS) instructions. Existing MCAS implementations indeed use at least 2k+1 CASes per k-CAS. This leads to the natural desire to minimize the number of CASes required to implement MCAS. We first prove in this paper that it is impossible to "pack" the information required to perform a k-word CAS (k-CAS) in less than k locations to be CASed. Then we present the first algorithm that requires k+1 CASes per call to k-CAS in the common uncontended case. We implement our algorithm and show that it outperforms a state-of-the-art baseline in a variety of benchmarks in most considered workloads. We also present a durably linearizable (persistent memory friendly) version of our MCAS algorithm using only 2 persistence fences per call, while still only requiring k+1 CASes per k-CAS.

cs.DC↗

arXiv2020

Who started this rumor? Quantifying the natural differential privacy guarantees of gossip protocols

Gossip protocols are widely used to disseminate information in massive peer-to-peer networks. These protocols are often claimed to guarantee privacy because of the uncertainty they introduce on the node that started the dissemination. But is that claim really true? Can the source of a gossip safely hide in the crowd? This paper examines, for the first time, gossip protocols through a rigorous mathematical framework based on differential privacy to determine the extent to which the source of a gossip can be traceable. Considering the case of a complete graph in which a subset of the nodes are curious, we study a family of gossip protocols parameterized by a ``muting'' parameter $s$: nodes stop emitting after each communication with a fixed probability $1-s$. We first prove that the standard push protocol, corresponding to the case $s=1$, does not satisfy differential privacy for large graphs. In contrast, the protocol with $s=0$ achieves optimal privacy guarantees but at the cost of a drastic increase in the spreading time compared to standard push, revealing an interesting tension between privacy and spreading time. Yet, surprisingly, we show that some choices of the muting parameter $s$ lead to protocols that achieve an optimal order of magnitude in both privacy and speed. We also confirm empirically that, with appropriate choices of $s$, we indeed obtain protocols that are very robust against concrete source location attacks while spreading the information almost as fast as the standard (and non-private) push protocol.

cs.DC↗

arXiv2020

Fast Machine Learning with Byzantine Workers and Servers

Machine Learning (ML) solutions are nowadays distributed and are prone to various types of component failures, which can be encompassed in so-called Byzantine behavior. This paper introduces LiuBei, a Byzantine-resilient ML algorithm that does not trust any individual component in the network (neither workers nor servers), nor does it induce additional communication rounds (on average), compared to standard non-Byzantine resilient algorithms. LiuBei builds upon gradient aggregation rules (GARs) to tolerate a minority of Byzantine workers. Besides, LiuBei replicates the parameter server on multiple machines instead of trusting it. We introduce a novel filtering mechanism that enables workers to filter out replies from Byzantine server replicas without requiring communication with all servers. Such a filtering mechanism is based on network synchrony, Lipschitz continuity of the loss function, and the GAR used to aggregate workers' gradients. We also introduce a protocol, scatter/gather, to bound drifts between models on correct servers with a small number of communication messages. We theoretically prove that LiuBei achieves Byzantine resilience to both servers and workers and guarantees convergence. We build LiuBei using TensorFlow, and we show that LiuBei tolerates Byzantine behavior with an accuracy loss of around 5% and around 24% convergence overhead compared to vanilla TensorFlow. We moreover show that the throughput gain of LiuBei compared to another state-of-the-art Byzantine-resilient ML algorithm (that assumes network asynchrony) is 70%.

cs.LG↗

arXiv2020

Host-Pathongen Co-evolution Inspired Algorithm Enables Robust GAN Training

Generative adversarial networks (GANs) are pairs of artificial neural networks that are trained one against each other. The outputs from a generator are mixed with the real-world inputs to the discriminator and both networks are trained until an equilibrium is reached, where the discriminator cannot distinguish generated inputs from real ones. Since their introduction, GANs have allowed for the generation of impressive imitations of real-life films, images and texts, whose fakeness is barely noticeable to humans. Despite their impressive performance, training GANs remains to this day more of an art than a reliable procedure, in a large part due to training process stability. Generators are susceptible to mode dropping and convergence to random patterns, which have to be mitigated by computationally expensive multiple restarts. Curiously, GANs bear an uncanny similarity to a co-evolution of a pathogen and its host's immune system in biology. In a biological context, the majority of potential pathogens indeed never make it and are kept at bay by the hots' immune system. Yet some are efficient enough to present a risk of a serious condition and recurrent infections. Here, we explore that similarity to propose a more robust algorithm for GANs training. We empirically show the increased stability and a better ability to generate high-quality images while using less computational power.

cs.NE↗

arXiv2020

Genuinely Distributed Byzantine Machine Learning

Machine Learning (ML) solutions are nowadays distributed, according to the so-called server/worker architecture. One server holds the model parameters while several workers train the model. Clearly, such architecture is prone to various types of component failures, which can be all encompassed within the spectrum of a Byzantine behavior. Several approaches have been proposed recently to tolerate Byzantine workers. Yet all require trusting a central parameter server. We initiate in this paper the study of the ``general'' Byzantine-resilient distributed machine learning problem where no individual component is trusted. We show that this problem can be solved in an asynchronous system, despite the presence of $\frac{1}{3}$ Byzantine parameter servers and $\frac{1}{3}$ Byzantine workers (which is optimal). We present a new algorithm, ByzSGD, which solves the general Byzantine-resilient distributed machine learning problem by relying on three major schemes. The first, Scatter/Gather, is a communication scheme whose goal is to bound the maximum drift among models on correct servers. The second, Distributed Median Contraction (DMC), leverages the geometric properties of the median in high dimensional spaces to bring parameters within the correct servers back close to each other, ensuring learning convergence. The third, Minimum-Diameter Averaging (MDA), is a statistically-robust gradient aggregation rule whose goal is to tolerate Byzantine workers. MDA requires loose bound on the variance of non-Byzantine gradient estimates, compared to existing alternatives (e.g., Krum). Interestingly, ByzSGD ensures Byzantine resilience without adding communication rounds (on a normal path), compared to vanilla non-Byzantine alternatives. ByzSGD requires, however, a larger number of messages which, we show, can be reduced if we assume synchrony.

cs.DC↗