Lili Su

We consider the problem of distributed statistical machine learning in adversarial settings, where some unknown and time-varying subset of working machines may be compromised and behave arbitrarily to prevent an accurate model from being learned. This setting captures the potential adversarial attacks faced by Federated Learning -- a modern machine learning paradigm that is proposed by Google researchers and has been intensively studied for ensuring user privacy. Formally, we focus on a distributed system consisting of a parameter server and $m$ working machines. Each working machine keeps $N/m$ data samples, where $N$ is the total number of samples. The goal is to collectively learn the underlying true model parameter of dimension $d$. In classical batch gradient descent methods, the gradients reported to the server by the working machines are aggregated via simple averaging, which is vulnerable to a single Byzantine failure. In this paper, we propose a Byzantine gradient descent method based on the geometric median of means of the gradients. We show that our method can tolerate $q \le (m-1)/2$ Byzantine failures, and the parameter estimate converges in $O(\log N)$ rounds with an estimation error of $\sqrt{d(2q+1)/N}$, hence approaching the optimal error rate $\sqrt{d/N}$ in the centralized and failure-free setting. The total computational complexity of our algorithm is of $O((Nd/m) \log N)$ at each working machine and $O(md + kd \log^3 N)$ at the central server, and the total communication cost is of $O(m d \log N)$. We further provide an application of our general results to the linear regression problem. A key challenge arises in the above problem is that Byzantine failures create arbitrary and unspecified dependency among the iterations and the aggregated gradients. We prove that the aggregated gradient converges uniformly to the true gradient function.

We study fault-tolerant distributed optimization of a sum of convex (cost) functions with real-valued scalar input/output in the presence of crash faults or Byzantine faults. In particular, the goal is to optimize a global cost function $\frac{1}{n}\sum_{i\in \mathcal{V}} h_i(x)$, where $\mathcal{V}=\{1, \ldots, n\}$ is the collection of agents, and $h_i(x)$ is agent $i$'s local cost function, which is initially known only to agent $i$. Since the above global cost function cannot be optimized exactly in presence of crash faults or Byzantine faults, we define two weaker versions of the problem for crash faults and Byzantine faults, respectively. When some agents may crash, the goal for the weaker problem is to generate an output that is an optimum of a function formed as $$C(\sum_{i\in \mathcal{N}} h_i(x)+\sum_{i\in \mathcal{F}} \alpha_i h_i(x)),$$ where $\mathcal{N}$ is the set of non-faulty agents, $\mathcal{F}$ is the set of faulty agents (crashed agents), $0\le \alpha_i\le 1$ for each $i\in \mathcal{F}$ and $C$ is a normalization constant such that $C(|\mathcal{N}|+\sum_{i\in \mathcal{F}} \alpha_i)=1$. We present an iterative algorithm in which each agent only needs to perform local computation, and send one message per iteration. When some agents may be Byzantine, the system cannot take full advantage of the data kept by non-faulty agents. The goal for the associated weaker problem is to generate an output that is an optimum of a function formed as $$\sum_{i\in \mathcal{N}}\alpha_i h_i(x),$$ such that $\alpha_i\geq 0$ for each $i\in \mathcal{N}$ and $\sum_{i\in \mathcal{N}}\alpha_i=1$. We present an iterative algorithm, where only local computation is needed and only one message per agent is sent in each iteration, that ensures that at least $|\mathcal{N}|-f$ agents have weights ($\alpha_i$'s) that are lower bounded by $\frac{1}{2(|\mathcal{N}|-f)}$.

We study Byzantine fault-tolerant distributed optimization of a sum of convex (cost) functions with real-valued scalar input/ouput. In particular, the goal is to optimize a global cost function $\frac{1}{|\mathcal{N}|}\sum_{i\in \mathcal{N}} h_i(x)$, where $\mathcal{N}$ is the set of non-faulty agents, and $h_i(x)$ is agent $i$'s local cost function, which is initially known only to agent $i$. In general, when some of the agents may be Byzantine faulty, the above goal is unachievable, because the identity of the faulty agents is not necessarily known to the non-faulty agents, and the faulty agents may behave arbitrarily. Since the above global cost function cannot be optimized exactly in presence of Byzantine agents, we define a weaker version of the problem. The goal for the weaker problem is to generate an output that is an optimum of a function formed as a convex combination of local cost functions of the non-faulty agents. More precisely, for some choice of weights $\alpha_i$ for $i\in \mathcal{N}$ such that $\alpha_i\geq 0$ and $\sum_{i\in \mathcal{N}}\alpha_i=1$, the output must be an optimum of the cost function $\sum_{i\in \mathcal{N}} \alpha_ih_i(x)$. Ideally, we would like $\alpha_i=\frac{1}{|\mathcal{N}|}$ for all $i\in \mathcal{N}$ -- however, this cannot be guaranteed due to the presence of faulty agents. In fact, we show that the maximum achievable number of nonzero weights ($\alpha_i$'s) is $|\mathcal{N}|-f$, where $f$ is the upper bound on the number of Byzantine agents. In addition, we present algorithms that ensure that at least $|\mathcal{N}|-f$ agents have weights that are bounded away from 0. We also propose a low-complexity suboptimal algorithm, which ensures that at least $\lceil \frac{n}{2}\rceil-\phi$ agents have weights that are bounded away from 0, where $n$ is the total number of agents, and $\phi$ ($\phi\le f$) is the actual number of Byzantine agents.

In Part I of this report, we introduced a Byzantine fault-tolerant distributed optimization problem whose goal is to optimize a sum of convex (cost) functions with real-valued scalar input/ouput. In this second part, we introduce a condition-based variant of the original problem over arbitrary directed graphs. Specifically, for a given collection of $k$ input functions $h_1(x), \ldots, h_k(x)$, we consider the scenario when the local cost function stored at agent $j$, denoted by $g_j(x)$, is formed as a convex combination of the $k$ input functions $h_1(x), \ldots, h_k(x)$. The goal of this condition-based problem is to generate an output that is an optimum of $\frac{1}{k}\sum_{i=1}^k h_i(x)$. Depending on the availability of side information at each agent, two slightly different variants are considered. We show that for a given graph, the problem can indeed be solved despite the presence of faulty agents. In particular, even in the absence of side information at each agent, when adequate redundancy is available in the optima of input functions, a distributed algorithm is proposed in which each agent carries minimal state across iterations.