Search

AI computing and data centers consume a large amount of freshwater, both directly for cooling and indirectly for electricity generation. While most attention has been paid to developed countries such as the U.S., this paper presents the first-of-its-kind dataset that combines nation-level weather and electricity generation data to estimate water usage efficiency for data centers in 41 African countries across five different climate regions. We also use our dataset to evaluate and estimate the water consumption of inference on two large language models (i.e., Llama-3-70B and GPT-4) in 11 selected African countries. Our findings show that writing a 10-page report using Llama-3-70B could consume about \textbf{0.7 liters} of water, while the water consumption by GPT-4 for the same task may go up to about 60 liters. For writing a medium-length email of 120-200 words, Llama-3-70B and GPT-4 could consume about \textbf{0.13 liters} and 3 liters of water, respectively. Interestingly, given the same AI model, 8 out of the 11 selected African countries consume less water than the global average, mainly because of lower water intensities for electricity generation. However, water consumption can be substantially higher in some African countries with a steppe climate than the U.S. and global averages, prompting more attention when deploying AI computing in these countries. Our dataset is publicly available on \href{https://huggingface.co/datasets/masterlion/WaterEfficientDatasetForAfricanCountries/tree/main}{Hugging Face}., Comment: Accepted by NeurIPS 2024 Workshop on Tackling Climate Change with Machine Learning

We introduce a privacy measure called statistic maximal leakage that quantifies how much a privacy mechanism leaks about a specific secret, relative to the adversary's prior information about that secret. Statistic maximal leakage is an extension of the well-known maximal leakage. Unlike maximal leakage, which protects an arbitrary, unknown secret, statistic maximal leakage protects a single, known secret. We show that statistic maximal leakage satisfies composition and post-processing properties. Additionally, we show how to efficiently compute it in the special case of deterministic data release mechanisms. We analyze two important mechanisms under statistic maximal leakage: the quantization mechanism and randomized response. We show theoretically and empirically that the quantization mechanism achieves better privacy-utility tradeoffs in the settings we study.

Automated marker makers (AMMs) are a class of decentralized exchanges that enable the automated trading of digital assets. They accept deposits of digital tokens from liquidity providers (LPs); tokens can be used by traders to execute trades, which generate fees for the investing LPs. The distinguishing feature of AMMs is that trade prices are determined algorithmically, unlike classical limit order books. Concentrated liquidity market makers (CLMMs) are a major class of AMMs that offer liquidity providers flexibility to decide not only \emph{how much} liquidity to provide, but \emph{in what ranges of prices} they want the liquidity to be used. This flexibility can complicate strategic planning, since fee rewards are shared among LPs. We formulate and analyze a game theoretic model to study the incentives of LPs in CLMMs. Our main results show that while our original formulation admits multiple Nash equilibria and has complexity quadratic in the number of price ticks in the contract, it can be reduced to a game with a unique Nash equilibrium whose complexity is only linear. We further show that the Nash equilibrium of this simplified game follows a waterfilling strategy, in which low-budget LPs use up their full budget, but rich LPs do not. Finally, by fitting our game model to real-world CLMMs, we observe that in liquidity pools with risky assets, LPs adopt investment strategies far from the Nash equilibrium. Under price uncertainty, they generally invest in fewer and wider price ranges than our analysis suggests, with lower-frequency liquidity updates. We show that across several pools, by updating their strategy to more closely match the Nash equilibrium of our game, LPs can improve their median daily returns by \$116, which corresponds to an increase of 0.009\% in median daily return on investment.

Information disclosure can compromise privacy when revealed information is correlated with private information. We consider the notion of inferential privacy, which measures privacy leakage by bounding the inferential power a Bayesian adversary can gain by observing a released signal. Our goal is to devise an inferentially-private private information structure that maximizes the informativeness of the released signal, following the Blackwell ordering principle, while adhering to inferential privacy constraints. To achieve this, we devise an efficient release mechanism that achieves the inferentially-private Blackwell optimal private information structure for the setting where the private information is binary. Additionally, we propose a programming approach to compute the optimal structure for general cases given the utility function. The design of our mechanisms builds on our geometric characterization of the Blackwell-optimal disclosure mechanisms under privacy constraints, which may be of independent interest.

Consistency models have recently been introduced to accelerate sampling from diffusion models by directly predicting the solution (i.e., data) of the probability flow ODE (PF ODE) from initial noise. However, the training of consistency models requires learning to map all intermediate points along PF ODE trajectories to their corresponding endpoints. This task is much more challenging than the ultimate objective of one-step generation, which only concerns the PF ODE's noise-to-data mapping. We empirically find that this training paradigm limits the one-step generation performance of consistency models. To address this issue, we generalize consistency training to the truncated time range, which allows the model to ignore denoising tasks at earlier time steps and focus its capacity on generation. We propose a new parameterization of the consistency function and a two-stage training procedure that prevents the truncated-time training from collapsing to a trivial solution. Experiments on CIFAR-10 and ImageNet $64\times64$ datasets show that our method achieves better one-step and two-step FIDs than the state-of-the-art consistency models such as iCT-deep, using more than 2$\times$ smaller networks. Project page: https://truncated-cm.github.io/

Data valuation is a class of techniques for quantitatively assessing the value of data for applications like pricing in data marketplaces. Existing data valuation methods define a value for a discrete dataset. However, in many use cases, users are interested in not only the value of the dataset, but that of the distribution from which the dataset was sampled. For example, consider a buyer trying to evaluate whether to purchase data from different vendors. The buyer may observe (and compare) only a small preview sample from each vendor, to decide which vendor's data distribution is most useful to the buyer and purchase. The core question is how should we compare the values of data distributions from their samples? Under a Huber characterization of the data heterogeneity across vendors, we propose a maximum mean discrepancy (MMD)-based valuation method which enables theoretically principled and actionable policies for comparing data distributions from samples. We empirically demonstrate that our method is sample-efficient and effective in identifying valuable data distributions against several existing baselines, on multiple real-world datasets (e.g., network intrusion detection, credit card fraud detection) and downstream applications (classification, regression)., Comment: Accepted to NeurIPS 2024 as a poster. Main paper with appendix (38 pages in total). Code will be released soon at https://github.com/XinyiYS/Data_Distribution_Valuation

In the international development community, the term "digital public goods" is used to describe open-source digital products (e.g., software, datasets) that aim to address the United Nations (UN) Sustainable Development Goals. DPGs are increasingly being used to deliver government services around the world (e.g., ID management, healthcare registration). Because DPGs may handle sensitive data, the UN has established user privacy as a first-order requirement for DPGs. The privacy risks of DPGs are currently managed in part by the DPG standard, which includes a prerequisite questionnaire with questions designed to evaluate a DPG's privacy posture. This study examines the effectiveness of the current DPG standard for ensuring adequate privacy protections. We present a systematic assessment of responses from DPGs regarding their protections of users' privacy. We also present in-depth case studies from three widely-used DPGs to identify privacy threats and compare this to their responses to the DPG standard. Our findings reveal limitations in the current DPG standard's evaluation approach. We conclude by presenting preliminary recommendations and suggestions for strengthening the DPG standard as it relates to privacy. Additionally, we hope this study encourages more usable privacy research on communicating privacy, not only to end users but also third-party adopters of user-facing technologies.

On-device training is currently the most common approach for training machine learning (ML) models on private, distributed user data. Despite this, on-device training has several drawbacks: (1) most user devices are too small to train large models on-device, (2) on-device training is communication- and computation-intensive, and (3) on-device training can be difficult to debug and deploy. To address these problems, we propose Private Evolution-Text (PrE-Text), a method for generating differentially private (DP) synthetic textual data. First, we show that across multiple datasets, training small models (models that fit on user devices) with PrE-Text synthetic data outperforms small models trained on-device under practical privacy regimes ($\epsilon=1.29$, $\epsilon=7.58$). We achieve these results while using 9$\times$ fewer rounds, 6$\times$ less client computation per round, and 100$\times$ less communication per round. Second, finetuning large models on PrE-Text's DP synthetic data improves large language model (LLM) performance on private data across the same range of privacy budgets. Altogether, these results suggest that training on DP synthetic data can be a better option than training a model on-device on private distributed data. Code is available at https://github.com/houcharlie/PrE-Text., Comment: ICML 2024 (Oral). Latest revision corrects a discussion on concurrent work arXiv:2403.01749. We described their work as reliant on using closed-sourced models when in reality they also evaluate and use open source models. This has been corrected in this version

Diffusion models have shown great promise for image and video generation, but sampling from state-of-the-art models requires expensive numerical integration of a generative ODE. One approach for tackling this problem is rectified flows, which iteratively learn smooth ODE paths that are less susceptible to truncation error. However, rectified flows still require a relatively large number of function evaluations (NFEs). In this work, we propose improved techniques for training rectified flows, allowing them to compete with \emph{knowledge distillation} methods even in the low NFE setting. Our main insight is that under realistic settings, a single iteration of the Reflow algorithm for training rectified flows is sufficient to learn nearly straight trajectories; hence, the current practice of using multiple Reflow iterations is unnecessary. We thus propose techniques to improve one-round training of rectified flows, including a U-shaped timestep distribution and LPIPS-Huber premetric. With these techniques, we improve the FID of the previous 2-rectified flow by up to 75\% in the 1 NFE setting on CIFAR-10. On ImageNet 64$\times$64, our improved rectified flow outperforms the state-of-the-art distillation methods such as consistency distillation and progressive distillation in both one-step and two-step settings and rivals the performance of improved consistency training (iCT) in FID. Code is available at https://github.com/sangyun884/rfpp., Comment: NeurIPS2024

Differentially private mechanisms achieving worst-case optimal error bounds (e.g., the classical Laplace mechanism) are well-studied in the literature. However, when typical data are far from the worst case, \emph{instance-specific} error bounds -- which depend on the largest value in the dataset -- are more meaningful. For example, consider the sum estimation problem, where each user has an integer $x_i$ from the domain $\{0,1,\dots,U\}$ and we wish to estimate $\sum_i x_i$. This has a worst-case optimal error of $O(U/\varepsilon)$, while recent work has shown that the clipping mechanism can achieve an instance-optimal error of $O(\max_i x_i \cdot \log\log U /\varepsilon)$. Under the shuffle model, known instance-optimal protocols are less communication-efficient. The clipping mechanism also works in the shuffle model, but requires two rounds: Round one finds the clipping threshold, and round two does the clipping and computes the noisy sum of the clipped data. In this paper, we show how these two seemingly sequential steps can be done simultaneously in one round using just $1+o(1)$ messages per user, while maintaining the instance-optimal error bound. We also extend our technique to the high-dimensional sum estimation problem and sparse vector aggregation (a.k.a. frequency estimation under user-level differential privacy).

Differentially private (DP) machine learning pipelines typically involve a two-phase process: non-private pre-training on a public dataset, followed by fine-tuning on private data using DP optimization techniques. In the DP setting, it has been observed that full fine-tuning may not always yield the best test accuracy, even for in-distribution data. This paper (1) analyzes the training dynamics of DP linear probing (LP) and full fine-tuning (FT), and (2) explores the phenomenon of sequential fine-tuning, starting with linear probing and transitioning to full fine-tuning (LP-FT), and its impact on test loss. We provide theoretical insights into the convergence of DP fine-tuning within an overparameterized neural network and establish a utility curve that determines the allocation of privacy budget between linear probing and full fine-tuning. The theoretical results are supported by empirical evaluations on various benchmarks and models. The findings reveal the complex nature of DP fine-tuning methods. These results contribute to a deeper understanding of DP machine learning and highlight the importance of considering the allocation of privacy budget in the fine-tuning process.

Differential privacy (DP) is increasingly used to protect the release of hierarchical, tabular population data, such as census data. A common approach for implementing DP in this setting is to release noisy responses to a predefined set of queries. For example, this is the approach of the TopDown algorithm used by the US Census Bureau. Such methods have an important shortcoming: they cannot answer queries for which they were not optimized. An appealing alternative is to generate DP synthetic data, which is drawn from some generating distribution. Like the TopDown method, synthetic data can also be optimized to answer specific queries, while also allowing the data user to later submit arbitrary queries over the synthetic population data. To our knowledge, there has not been a head-to-head empirical comparison of these approaches. This study conducts such a comparison between the TopDown algorithm and private synthetic data generation to determine how accuracy is affected by query complexity, in-distribution vs. out-of-distribution queries, and privacy guarantees. Our results show that for in-distribution queries, the TopDown algorithm achieves significantly better privacy-fidelity tradeoffs than any of the synthetic data methods we evaluated; for instance, in our experiments, TopDown achieved at least $20\times$ lower error on counting queries than the leading synthetic data method at the same privacy budget. Our findings suggest guidelines for practitioners and the synthetic data research community., Comment: The 5th AAAI Workshop on Privacy-Preserving Artificial Intelligence

Long-term time series forecasting (LTSF) aims to predict future values of a time series given the past values. The current state-of-the-art (SOTA) on this problem is attained in some cases by linear-centric models, which primarily feature a linear mapping layer. However, due to their inherent simplicity, they are not able to adapt their prediction rules to periodic changes in time series patterns. To address this challenge, we propose a Mixture-of-Experts-style augmentation for linear-centric models and propose Mixture-of-Linear-Experts (MoLE). Instead of training a single model, MoLE trains multiple linear-centric models (i.e., experts) and a router model that weighs and mixes their outputs. While the entire framework is trained end-to-end, each expert learns to specialize in a specific temporal pattern, and the router model learns to compose the experts adaptively. Experiments show that MoLE reduces forecasting error of linear-centric models, including DLinear, RLinear, and RMLP, in over 78% of the datasets and settings we evaluated. By using MoLE existing linear-centric models can achieve SOTA LTSF results in 68% of the experiments that PatchTST reports and we compare to, whereas existing single-head linear-centric models achieve SOTA results in only 25% of cases.

Mobile money services in Sub-Saharan Africa (SSA) have increased access to financial services. To ensure proper identification of users, countries have put in place Know-Your-Customer (KYC) measures such as SIM registration using an official identification. However, half of the 850 million people without IDs globally live in SSA, and the use of SIM cards registered in another person's name (third-party SIM) is prevalent. In this study, we explore challenges that contribute to and arise from the use of third-party SIM cards. We interviewed 36 participants in Kenya and Tanzania. Our results highlight great strides in ID accessibility, but also highlight numerous institutional and social factors that contribute to the use of third-party SIM cards. While privacy concerns contribute to the use of third-party SIM cards, third-party SIM card users are exposed to significant security and privacy risks, including scams, financial loss, and wrongful arrest.

Digital financial services have catalyzed financial inclusion in Africa. Commonly implemented as a mobile wallet service referred to as mobile money (MoMo), the technology provides enormous benefits to its users, some of whom have long been unbanked. While the benefits of mobile money services have largely been documented, the challenges that arise -- especially in the interactions between human stakeholders -- remain relatively unexplored. In this study, we investigate the practices of mobile money users in their interactions with mobile money agents. We conduct 72 structured interviews in Kenya and Tanzania (n=36 per country). The results show that users and agents design workarounds in response to limitations and challenges that users face within the ecosystem. These include advances or loans from agents, relying on the user-agent relationships in place of legal identification requirements, and altering the intended transaction execution to improve convenience. Overall, the workarounds modify one or more of what we see as the core components of mobile money: the user, the agent, and the transaction itself. The workarounds pose new risks and challenges for users and the overall ecosystem. The results suggest a need for rethinking privacy and security of various components of the ecosystem, as well as policy and regulatory controls to safeguard interactions while ensuring the usability of mobile money., Comment: To be published in IEEE Symposium on Security and Privacy 2024

On tabular data, a significant body of literature has shown that current deep learning (DL) models perform at best similarly to Gradient Boosted Decision Trees (GBDTs), while significantly underperforming them on outlier data. However, these works often study idealized problem settings which may fail to capture complexities of real-world scenarios. We identify a natural tabular data setting where DL models can outperform GBDTs: tabular Learning-to-Rank (LTR) under label scarcity. Tabular LTR applications, including search and recommendation, often have an abundance of unlabeled data, and scarce labeled data. We show that DL rankers can utilize unsupervised pretraining to exploit this unlabeled data. In extensive experiments over both public and proprietary datasets, we show that pretrained DL rankers consistently outperform GBDT rankers on ranking metrics -- sometimes by as much as 38% -- both overall and on outliers., Comment: ICML-MFPL 2023 Workshop Oral, SPIGM@ICML2024

Crash fault tolerant (CFT) consensus algorithms are commonly used in scenarios where system components are trusted -- e.g., enterprise settings and government infrastructure. However, CFT consensus can be broken by even a single corrupt node. A desirable property in the face of such potential Byzantine faults is \emph{accountability}: if a corrupt node breaks protocol and affects consensus safety, it should be possible to identify the culpable components with cryptographic integrity from the node states. Today, the best-known protocol for providing accountability to CFT protocols is called PeerReview; it essentially records a signed transcript of all messages sent during the CFT protocol. Because PeerReview is agnostic to the underlying CFT protocol, it incurs high communication and storage overhead. We propose CFT-Forensics, an accountability framework for CFT protocols. We show that for a special family of \emph{forensics-compliant} CFT protocols (which includes widely-used CFT protocols like Raft and multi-Paxos), CFT-Forensics gives provable accountability guarantees. Under realistic deployment settings, we show theoretically that CFT-Forensics operates at a fraction of the cost of PeerReview. We subsequently instantiate CFT-Forensics for Raft, and implement Raft-Forensics as an extension to the popular nuRaft library. In extensive experiments, we demonstrate that Raft-Forensics adds low overhead to vanilla Raft. With 256 byte messages, Raft-Forensics achieves a peak throughput 87.8\% of vanilla Raft at 46\% higher latency ($+44$ ms). We finally integrate Raft-Forensics into the open-source central bank digital currency OpenCBDC, and show that in wide-area network experiments, Raft-Forensics achieves 97.8\% of the throughput of Raft, with 14.5\% higher latency ($+326$ ms).

We study a setting where a data holder wishes to share data with a receiver, without revealing certain summary statistics of the data distribution (e.g., mean, standard deviation). It achieves this by passing the data through a randomization mechanism. We propose summary statistic privacy, a metric for quantifying the privacy risk of such a mechanism based on the worst-case probability of an adversary guessing the distributional secret within some threshold. Defining distortion as a worst-case Wasserstein-1 distance between the real and released data, we prove lower bounds on the tradeoff between privacy and distortion. We then propose a class of quantization mechanisms that can be adapted to different data distributions. We show that the quantization mechanism's privacy-distortion tradeoff matches our lower bounds under certain regimes, up to small constant factors. Finally, we demonstrate on real-world datasets that the proposed quantization mechanisms achieve better privacy-distortion tradeoffs than alternative privacy mechanisms.

In Federated Learning (FL), accessing private client data incurs communication and privacy costs. As a result, FL deployments commonly prefinetune pretrained foundation models on a (large, possibly public) dataset that is held by the central server; they then FL-finetune the model on a private, federated dataset held by clients. Evaluating prefinetuning dataset quality reliably and privately is therefore of high importance. To this end, we propose FreD (Federated Private Fr\'echet Distance) -- a privately computed distance between a prefinetuning dataset and federated datasets. Intuitively, it privately computes and compares a Fr\'echet distance between embeddings generated by a large language model on both the central (public) dataset and the federated private client data. To make this computation privacy-preserving, we use distributed, differentially-private mean and covariance estimators. We show empirically that FreD accurately predicts the best prefinetuning dataset at minimal privacy cost. Altogether, using FreD we demonstrate a proof-of-concept for a new approach in private FL training: (1) customize a prefinetuning dataset to better match user data (2) prefinetune (3) perform FL-finetuning.

There are a number of forums where people participate under pseudonyms. One example is peer review, where the identity of reviewers for any paper is confidential. When participating in these forums, people frequently engage in "batching": executing multiple related tasks (e.g., commenting on multiple papers) at nearly the same time. Our empirical analysis shows that batching is common in two applications we consider $\unicode{x2013}$ peer review and Wikipedia edits. In this paper, we identify and address the risk of deanonymization arising from linking batched tasks. To protect against linkage attacks, we take the approach of adding delay to the posting time of batched tasks. We first show that under some natural assumptions, no delay mechanism can provide a meaningful differential privacy guarantee. We therefore propose a "one-sided" formulation of differential privacy for protecting against linkage attacks. We design a mechanism that adds zero-inflated uniform delay to events and show it can preserve privacy. We prove that this noise distribution is in fact optimal in minimizing expected delay among mechanisms adding independent noise to each event, thereby establishing the Pareto frontier of the trade-off between the expected delay for batched and unbatched events. Finally, we conduct a series of experiments on Wikipedia and Bitcoin data that corroborate the practical utility of our algorithm in obfuscating batching without introducing onerous delay to a system.

The privacy implications of generative adversarial networks (GANs) are a topic of great interest, leading to several recent algorithms for training GANs with privacy guarantees. By drawing connections to the generalization properties of GANs, we prove that under some assumptions, GAN-generated samples inherently satisfy some (weak) privacy guarantees. First, we show that if a GAN is trained on m samples and used to generate n samples, the generated samples are (epsilon, delta)-differentially-private for (epsilon, delta) pairs where delta scales as O(n/m). We show that under some special conditions, this upper bound is tight. Next, we study the robustness of GAN-generated samples to membership inference attacks. We model membership inference as a hypothesis test in which the adversary must determine whether a given sample was drawn from the training dataset or from the underlying data distribution. We show that this adversary can achieve an area under the ROC curve that scales no better than O(m^{-1/4})., Comment: AISTATS 2021

Backdoor attacks are dangerous and difficult to prevent in federated learning (FL), where training data is sourced from untrusted clients over long periods of time. These difficulties arise because: (a) defenders in FL do not have access to raw training data, and (b) a new phenomenon we identify called backdoor leakage causes models trained continuously to eventually suffer from backdoors due to cumulative errors in defense mechanisms. We propose shadow learning, a framework for defending against backdoor attacks in the FL setting under long-range training. Shadow learning trains two models in parallel: a backbone model and a shadow model. The backbone is trained without any defense mechanism to obtain good performance on the main task. The shadow model combines filtering of malicious clients with early-stopping to control the attack success rate even as the data distribution changes. We theoretically motivate our design and show experimentally that our framework significantly improves upon existing defenses against backdoor attacks.

Most permissionless blockchain networks run on peer-to-peer (P2P) networks, which offer flexibility and decentralization at the expense of performance (e.g., network latency). Historically, this tradeoff has not been a bottleneck for most blockchains. However, an emerging host of blockchain-based applications (e.g., decentralized finance) are increasingly sensitive to latency; users who can reduce their network latency relative to other users can accrue (sometimes significant) financial gains. In this work, we initiate the study of strategic latency reduction in blockchain P2P networks. We first define two classes of latency that are of interest in blockchain applications. We then show empirically that a strategic agent who controls only their local peering decisions can manipulate both types of latency, achieving 60\% of the global latency gains provided by the centralized, paid service bloXroute, or, in targeted scenarios, comparable gains. Finally, we show that our results are not due to the poor design of existing P2P networks. Under a simple network model, we theoretically prove that an adversary can always manipulate the P2P network's latency to their advantage, provided the network experiences sufficient peer churn and transaction activity.

We study the problem of learning generative adversarial networks (GANs) for a rare class of an unlabeled dataset subject to a labeling budget. This problem is motivated from practical applications in domains including security (e.g., synthesizing packets for DNS amplification attacks), systems and networking (e.g., synthesizing workloads that trigger high resource usage), and machine learning (e.g., generating images from a rare class). Existing approaches are unsuitable, either requiring fully-labeled datasets or sacrificing the fidelity of the rare class for that of the common classes. We propose RareGAN, a novel synthesis of three key ideas: (1) extending conditional GANs to use labelled and unlabelled data for better generalization; (2) an active learning approach that requests the most useful labels; and (3) a weighted loss function to favor learning the rare class. We show that RareGAN achieves a better fidelity-diversity tradeoff on the rare class than prior work across different applications, budgets, rare class fractions, GAN losses, and architectures., Comment: Published in AAAI 2022

Private set intersection (PSI) allows two mutually untrusting parties to compute an intersection of their sets, without revealing information about items that are not in the intersection. This work introduces a PSI variant called distance-aware PSI (DA-PSI) for sets whose elements lie in a metric space. DA-PSI returns pairs of items that are within a specified distance threshold of each other. This paper puts forward DA-PSI constructions for two metric spaces: (i) Minkowski distance of order 1 over the set of integers (i.e., for integers $a$ and $b$, their distance is $|a-b|$); and (ii) Hamming distance over the set of binary strings of length $\ell$. In the Minkowski DA-PSI protocol, the communication complexity scales logarithmically in the distance threshold and linearly in the set size. In the Hamming DA-PSI protocol, the communication volume scales quadratically in the distance threshold and is independent of the dimensionality of string length $\ell$. Experimental results with real applications confirm that DA-PSI provides more effective matching at lower cost than naive solutions.

Vector mean estimation is a central primitive in federated analytics. In vector mean estimation, each user $i \in [n]$ holds a real-valued vector $v_i\in [-1, 1]^d$, and a server wants to estimate the mean of all $n$ vectors. Not only so, we would like to protect each individual user's privacy. In this paper, we consider the $k$-sparse version of the vector mean estimation problem, that is, suppose that each user's vector has at most $k$ non-zero coordinates in its $d$-dimensional vector, and moreover, $k \ll d$. In practice, since the universe size $d$ can be very large (e.g., the space of all possible URLs), we would like the per-user communication to be succinct, i.e., independent of or (poly-)logarithmic in the universe size. In this paper, we are the first to show matching upper- and lower-bounds for the $k$-sparse vector mean estimation problem under local differential privacy. Specifically, we construct new mechanisms that achieve asymptotically optimal error as well as succinct communication, either under user-level-LDP or event-level-LDP. We implement our algorithms and evaluate them on synthetic as well as real-world datasets. Our experiments show that we can often achieve one or two orders of magnitude reduction in error in comparison with prior works under typical choices of parameters, while incurring insignificant communication cost.

Federated learning (FL) aims to minimize the communication complexity of training a model over heterogeneous data distributed across many clients. A common approach is local methods, where clients take multiple optimization steps over local data before communicating with the server (e.g., FedAvg). Local methods can exploit similarity between clients' data. However, in existing analyses, this comes at the cost of slow convergence in terms of the dependence on the number of communication rounds R. On the other hand, global methods, where clients simply return a gradient vector in each round (e.g., SGD), converge faster in terms of R but fail to exploit the similarity between clients even when clients are homogeneous. We propose FedChain, an algorithmic framework that combines the strengths of local methods and global methods to achieve fast convergence in terms of R while leveraging the similarity between clients. Using FedChain, we instantiate algorithms that improve upon previously known rates in the general convex and PL settings, and are near-optimal (via an algorithm-independent lower bound that we show) for problems that satisfy strong convexity. Empirical results support this theoretical gain over existing methods., Comment: abstract typo correction

Abusive language is a massive problem in online social platforms. Existing abusive language detection techniques are particularly ill-suited to comments containing heterogeneous abusive language patterns, i.e., both abusive and non-abusive parts. This is due in part to the lack of datasets that explicitly annotate heterogeneity in abusive language. We tackle this challenge by providing an annotated dataset of abusive language in over 11,000 comments from YouTube. We account for heterogeneity in this dataset by separately annotating both the comment as a whole and the individual sentences that comprise each comment. We then propose an algorithm that uses a supervised attention mechanism to detect and categorize abusive content using multi-task learning. We empirically demonstrate the challenges of using traditional techniques on heterogeneous content and the comparative gains in performance of the proposed approach over state-of-the-art methods., Comment: AAAI 2021 (AI for Social Impact track)

Fringe groups and organizations have a long history of using euphemisms--ordinary-sounding words with a secret meaning--to conceal what they are discussing. Nowadays, one common use of euphemisms is to evade content moderation policies enforced by social media platforms. Existing tools for enforcing policy automatically rely on keyword searches for words on a "ban list", but these are notoriously imprecise: even when limited to swearwords, they can still cause embarrassing false positives. When a commonly used ordinary word acquires a euphemistic meaning, adding it to a keyword-based ban list is hopeless: consider "pot" (storage container or marijuana?) or "heater" (household appliance or firearm?) The current generation of social media companies instead hire staff to check posts manually, but this is expensive, inhumane, and not much more effective. It is usually apparent to a human moderator that a word is being used euphemistically, but they may not know what the secret meaning is, and therefore whether the message violates policy. Also, when a euphemism is banned, the group that used it need only invent another one, leaving moderators one step behind. This paper will demonstrate unsupervised algorithms that, by analyzing words in their sentence-level context, can both detect words being used euphemistically, and identify the secret meaning of each word. Compared to the existing state of the art, which uses context-free word embeddings, our algorithm for detecting euphemisms achieves 30-400% higher detection accuracies of unlabeled euphemisms in a text corpus. Our algorithm for revealing euphemistic meanings of words is the first of its kind, as far as we are aware. In the arms race between content moderators and policy evaders, our algorithms may help shift the balance in the direction of the moderators., Comment: 18 pages, 5 figures, 10 tables, 42nd IEEE Symposium on Security & Privacy (2021)

Credit networks rely on decentralized, pairwise trust relationships (channels) to exchange money or goods. Credit networks arise naturally in many financial systems, including the recent construct of payment channel networks in blockchain systems. An important performance metric for these networks is their transaction throughput. However, predicting the throughput of a credit network is nontrivial. Unlike traditional communication channels, credit channels can become imbalanced; they are unable to support more transactions in a given direction once the credit limit has been reached. This potential for imbalance creates a complex dependency between a network's throughput and its topology, path choices, and the credit balances (state) on every channel. Even worse, certain combinations of these factors can lead the credit network to deadlocked states where no transactions can make progress. In this paper, we study the relationship between the throughput of a credit network and its topology and credit state. We show that the presence of deadlocks completely characterizes a network's throughput sensitivity to different credit states. Although we show that identifying deadlocks in an arbitrary topology is NP-hard, we propose a peeling algorithm inspired by decoding algorithms for erasure codes that upper bounds the severity of the deadlock. We use the peeling algorithm as a tool to compare the performance of different topologies as well as to aid in the synthesis of topologies robust to deadlocks.

We consider strongly convex-concave minimax problems in the federated setting, where the communication constraint is the main bottleneck. When clients are arbitrarily heterogeneous, a simple Minibatch Mirror-prox achieves the best performance. As the clients become more homogeneous, using multiple local gradient updates at the clients significantly improves upon Minibatch Mirror-prox by communicating less frequently. Our goal is to design an algorithm that can harness the benefit of similarity in the clients while recovering the Minibatch Mirror-prox performance under arbitrary heterogeneity (up to log factors). We give the first federated minimax optimization algorithm that achieves this goal. The main idea is to combine (i) SCAFFOLD (an algorithm that performs variance reduction across clients for convex optimization) to erase the worst-case dependency on heterogeneity and (ii) Catalyst (a framework for acceleration based on modifying the objective) to accelerate convergence without amplifying client drift. We prove that this algorithm achieves our goal, and include experiments to validate the theory.

Spectral normalization (SN) is a widely-used technique for improving the stability and sample quality of Generative Adversarial Networks (GANs). However, there is currently limited understanding of why SN is effective. In this work, we show that SN controls two important failure modes of GAN training: exploding and vanishing gradients. Our proofs illustrate a (perhaps unintentional) connection with the successful LeCun initialization. This connection helps to explain why the most popular implementation of SN for GANs requires no hyper-parameter tuning, whereas stricter implementations of SN have poor empirical performance out-of-the-box. Unlike LeCun initialization which only controls gradient vanishing at the beginning of training, SN preserves this property throughout training. Building on this theoretical understanding, we propose a new spectral normalization technique: Bidirectional Scaled Spectral Normalization (BSSN), which incorporates insights from later improvements to LeCun initialization: Xavier initialization and Kaiming initialization. Theoretically, we show that BSSN gives better gradient control than SN. Empirically, we demonstrate that it outperforms SN in sample quality and training stability on several benchmark datasets., Comment: 54 pages, 74 figures

Incentive mechanisms are central to the functionality of permissionless blockchains: they incentivize participants to run and secure the underlying consensus protocol. Designing incentive-compatible incentive mechanisms is notoriously challenging, however. As a result, most public blockchains today use incentive mechanisms whose security properties are poorly understood and largely untested. In this work, we propose SquirRL, a framework for using deep reinforcement learning to analyze attacks on blockchain incentive mechanisms. We demonstrate SquirRL's power by first recovering known attacks: (1) the optimal selfish mining attack in Bitcoin [52], and (2) the Nash equilibrium in block withholding attacks [16]. We also use SquirRL to obtain several novel empirical results. First, we discover a counterintuitive flaw in the widely used rushing adversary model when applied to multi-agent Markov games with incomplete information. Second, we demonstrate that the optimal selfish mining strategy identified in [52] is actually not a Nash equilibrium in the multi-agent selfish mining setting. In fact, our results suggest (but do not prove) that when more than two competing agents engage in selfish mining, there is no profitable Nash equilibrium. This is consistent with the lack of observed selfish mining in the wild. Third, we find a novel attack on a simplified version of Ethereum's finalization mechanism, Casper the Friendly Finality Gadget (FFG) that allows a strategic agent to amplify her rewards by up to 30%. Notably, [10] show that honest voting is a Nash equilibrium in Casper FFG: our attack shows that when Casper FFG is composed with selfish mining, this is no longer the case. Altogether, our experiments demonstrate SquirRL's flexibility and promise as a framework for studying attack settings that have thus far eluded theoretical and empirical understanding.

Limited data access is a longstanding barrier to data-driven research and development in the networked systems community. In this work, we explore if and how generative adversarial networks (GANs) can be used to incentivize data sharing by enabling a generic framework for sharing synthetic datasets with minimal expert knowledge. As a specific target, our focus in this paper is on time series datasets with metadata (e.g., packet loss rate measurements with corresponding ISPs). We identify key challenges of existing GAN approaches for such workloads with respect to fidelity (e.g., long-term dependencies, complex multidimensional relationships, mode collapse) and privacy (i.e., existing guarantees are poorly understood and can sacrifice fidelity). To improve fidelity, we design a custom workflow called DoppelGANger (DG) and demonstrate that across diverse real-world datasets (e.g., bandwidth measurements, cluster requests, web sessions) and use cases (e.g., structural characterization, predictive modeling, algorithm comparison), DG achieves up to 43% better fidelity than baseline models. Although we do not resolve the privacy problem in this work, we identify fundamental challenges with both classical notions of privacy and recent advances to improve the privacy properties of GANs, and suggest a potential roadmap for addressing these challenges. By shedding light on the promise and challenges, we hope our work can rekindle the conversation on workflows for data sharing., Comment: Published in IMC 2020. 20 pages, 26 figures

Bitcoin is the first fully-decentralized permissionless blockchain protocol to achieve a high level of security, but at the expense of poor throughput and latency. Scaling the performance of Bitcoin has a been a major recent direction of research. One successful direction of work has involved replacing proof of work (PoW) by proof of stake (PoS). Proposals to scale the performance in the PoW setting itself have focused mostly on parallelizing the mining process, scaling throughput; the few proposals to improve latency have either sacrificed throughput or the latency guarantees involve large constants rendering it practically useless. Our first contribution is to design a new PoW blockchain Prism++ that has provably low latency and high throughput; the design retains the parallel-chain approach espoused in Prism but invents a new confirmation rule to infer the permanency of a block by combining information across the parallel chains. We show security at the level of Bitcoin with very small confirmation latency (a small constant factor of block interarrival time). A key aspect to scaling the performance is to use a large number of parallel chains, which puts significant strain on the system. Our second contribution is the design and evaluation of a practical system to efficiently manage the memory, computation, and I/O imperatives of a large number of parallel chains. Our implementation of Prism++ achieves a throughput of over 80,000 transactions per second and confirmation latency of tens of seconds on networks of up to 900 EC2 Virtual Machines.

A blockchain is a database of sequential events that is maintained by a distributed group of nodes. A key consensus problem in blockchains is that of determining the next block (data element) in the sequence. Many blockchains address this by electing a new node to propose each new block. The new block is (typically) appended to the tip of the proposer's local blockchain, and subsequently broadcast to the rest of the network. Without network delay (or adversarial behavior), this procedure would give a perfect chain, since each proposer would have the same view of the blockchain. A major challenge in practice is forking. Due to network delays, a proposer may not yet have the most recent block, and may, therefore, create a side chain that branches from the middle of the main chain. Forking reduces throughput, since only one a single main chain can survive, and all other blocks are discarded. We propose a new P2P protocol for blockchains called Barracuda, in which each proposer, prior to proposing a block, polls $\ell$ other nodes for their local blocktree information. Under a stochastic network model, we prove that this lightweight primitive improves throughput as if the entire network were a factor of $\ell$ faster. We provide guidelines on how to implement Barracuda in practice, guaranteeing robustness against several real-world factors., Comment: ACM Mobihoc 2019, Best paper award

Payment channel networks (PCNs) are viewed as one of the most promising scalability solutions for cryptocurrencies today. Roughly, PCNs are networks where each node represents a user and each directed, weighted edge represents funds escrowed on a blockchain; these funds can be transacted only between the endpoints of the edge. Users efficiently transmit funds from node A to B by relaying them over a path connecting A to B, as long as each edge in the path contains enough balance (escrowed funds) to support the transaction. Whenever a transaction succeeds, the edge weights are updated accordingly. However, in deployed PCNs, channel balances (i.e., edge weights) are not revealed to users for privacy reasons; users know only the initial weights at time 0. Hence, when routing transactions, users first guess a path, then check if it supports the transaction. This guess-and-check process dramatically reduces the success rate of transactions. At the other extreme, knowing full channel balances can give substantial improvements in success rate at the expense of privacy. In this work, we study whether a network can reveal noisy channel balances to trade off privacy for utility. We show fundamental limits on such a tradeoff, and propose noise mechanisms that achieve the fundamental limit for a general class of graph topologies. Our results suggest that in practice, PCNs should operate either in the low-privacy or low-utility regime; it is not possible to get large gains in utility by giving up a little privacy, or large gains in privacy by sacrificing a little utility., Comment: This revision corrects an error in Thm. 3.2, which previously under-estimated the converse bound by a factor of 2/n

Disentangled generative models map a latent code vector to a target space, while enforcing that a subset of the learned latent codes are interpretable and associated with distinct properties of the target distribution. Recent advances have been dominated by Variational AutoEncoder (VAE)-based methods, while training disentangled generative adversarial networks (GANs) remains challenging. In this work, we show that the dominant challenges facing disentangled GANs can be mitigated through the use of self-supervision. We make two main contributions: first, we design a novel approach for training disentangled GANs with self-supervision. We propose contrastive regularizer, which is inspired by a natural notion of disentanglement: latent traversal. This achieves higher disentanglement scores than state-of-the-art VAE- and GAN-based approaches. Second, we propose an unsupervised model selection scheme called ModelCentrality, which uses generated synthetic samples to compute the medoid (multi-dimensional generalization of median) of a collection of models. The current common practice of hyper-parameter tuning requires using ground-truths samples, each labelled with known perfect disentangled latent codes. As real datasets are not equipped with such labels, we propose an unsupervised model selection scheme and show that it finds a model close to the best one, for both VAEs and GANs. Combining contrastive regularization with ModelCentrality, we improve upon the state-of-the-art disentanglement scores significantly, without accessing the supervised data., Comment: Published in ICML 2020. 45 pages, 52 figures, a new unsupervised model selection scheme (ModelCentrality) is introduced in this version

Motivated by applications in blockchains and sensor networks, we consider a model of $n$ nodes trying to reach consensus on their majority bit. Each node $i$ is assigned a bit at time zero, and is a finite automaton with $m$ bits of memory (i.e., $2^m$ states) and a Poisson clock. When the clock of $i$ rings, $i$ can choose to communicate, and is then matched to a uniformly chosen node $j$. The nodes $j$ and $i$ may update their states based on the state of the other node. Previous work has focused on minimizing the time to consensus and the probability of error, while our goal is minimizing the number of communications. We show that when $m>3 \log\log\log(n)$, consensus can be reached at linear communication cost, but this is impossible if $m<\log\log\log(n)$. We also study a synchronous variant of the model, where our upper and lower bounds on $m$ for achieving linear communication cost are $2\log\log\log(n)$ and $\log\log\log(n)$, respectively. A key step is to distinguish when nodes can become aware of knowing the majority bit and stop communicating. We show that this is impossible if their memory is too low., Comment: 62 pages, 5 figures

Transaction throughput, confirmation latency and confirmation reliability are fundamental performance measures of any blockchain system in addition to its security. In a decentralized setting, these measures are limited by two underlying physical network attributes: communication capacity and speed-of-light propagation delay. Existing systems operate far away from these physical limits. In this work we introduce Prism, a new proof-of-work blockchain protocol, which can achieve 1) security against up to 50% adversarial hashing power; 2) optimal throughput up to the capacity C of the network; 3) confirmation latency for honest transactions proportional to the propagation delay D, with confirmation error probability exponentially small in CD ; 4) eventual total ordering of all transactions. Our approach to the design of this protocol is based on deconstructing the blockchain into its basic functionalities and systematically scaling up these functionalities to approach their physical limits., Comment: Computer and Communications Security, 2019

Proof-of-stake (PoS) is a promising approach for designing efficient blockchains, where block proposers are randomly chosen with probability proportional to their stake. A primary concern with PoS systems is the "rich getting richer" phenomenon, whereby wealthier nodes are more likely to get elected, and hence reap the block reward, making them even wealthier. In this paper, we introduce the notion of equitability, which quantifies how much a proposer can amplify her stake compared to her initial investment. Even with everyone following protocol (i.e., honest behavior), we show that existing methods of allocating block rewards lead to poor equitability, as does initializing systems with small stake pools and/or large rewards relative to the stake pool. We identify a \emph{geometric} reward function, which we prove is maximally equitable over all choices of reward functions under honest behavior and bound the deviation for strategic actions; the proofs involve the study of optimization problems and stochastic dominances of Polya urn processes, and are of independent mathematical interest. These results allow us to provide a systematic framework to choose the parameters of a practical incentive system for PoS cryptocurrencies., Comment: 26 pages

Despite growing adoption of cryptocurrencies, making fast payments at scale remains a challenge. Payment channel networks (PCNs) such as the Lightning Network have emerged as a viable scaling solution. However, completing payments on PCNs is challenging: payments must be routed on paths with sufficient funds. As payments flow over a single channel (link) in the same direction, the channel eventually becomes depleted and cannot support further payments in that direction; hence, naive routing schemes like shortest-path routing can deplete key payment channels and paralyze the system. Today's PCNs also route payments atomically, worsening the problem. In this paper, we present Spider, a routing solution that "packetizes" transactions and uses a multi-path transport protocol to achieve high-throughput routing in PCNs. Packetization allows Spider to complete even large transactions on low-capacity payment channels over time, while the multi-path congestion control protocol ensures balanced utilization of channels and fairness across flows. Extensive simulations comparing Spider with state-of-the-art approaches shows that Spider requires less than 25% of the funds to successfully route over 95% of transactions on balanced traffic demands, and offloads 4x more transactions onto the PCN on imbalanced demands.

Recent work has demonstrated significant anonymity vulnerabilities in Bitcoin's networking stack. In particular, the current mechanism for broadcasting Bitcoin transactions allows third-party observers to link transactions to the IP addresses that originated them. This lays the groundwork for low-cost, large-scale deanonymization attacks. In this work, we present Dandelion++, a first-principles defense against large-scale deanonymization attacks with near-optimal information-theoretic guarantees. Dandelion++ builds upon a recent proposal called Dandelion that exhibited similar goals. However, in this paper, we highlight simplifying assumptions made in Dandelion, and show how they can lead to serious deanonymization attacks when violated. In contrast, Dandelion++ defends against stronger adversaries that are allowed to disobey protocol. Dandelion++ is lightweight, scalable, and completely interoperable with the existing Bitcoin network. We evaluate it through experiments on Bitcoin's mainnet (i.e., the live Bitcoin network) to demonstrate its interoperability and low broadcast latency overhead.

Generative adversarial networks (GANs) are innovative techniques for learning generative models of complex data distributions from samples. Despite remarkable recent improvements in generating realistic images, one of their major shortcomings is the fact that in practice, they tend to produce samples with little diversity, even when trained on diverse datasets. This phenomenon, known as mode collapse, has been the main focus of several recent advances in GANs. Yet there is little understanding of why mode collapse happens and why existing approaches are able to mitigate mode collapse. We propose a principled approach to handling mode collapse, which we call packing. The main idea is to modify the discriminator to make decisions based on multiple samples from the same class, either real or artificially generated. We borrow analysis tools from binary hypothesis testing---in particular the seminal result of Blackwell [Bla53]---to prove a fundamental connection between packing and mode collapse. We show that packing naturally penalizes generators with mode collapse, thereby favoring generator distributions with less mode collapse during the training process. Numerical experiments on benchmark datasets suggests that packing provides significant improvements in practice as well., Comment: 49 pages, 24 figures

Bitcoin is a popular alternative to fiat money, widely used for its perceived anonymity properties. However, recent attacks on Bitcoin's peer-to-peer (P2P) network demonstrated that its gossip-based flooding protocols, which are used to ensure global network consistency, may enable user deanonymization---the linkage of a user's IP address with her pseudonym in the Bitcoin network. In 2015, the Bitcoin community responded to these attacks by changing the network's flooding mechanism to a different protocol, known as diffusion. However, no systematic justification was provided for the change, and it is unclear if diffusion actually improves the system's anonymity. In this paper, we model the Bitcoin networking stack and analyze its anonymity properties, both pre- and post-2015. In doing so, we consider new adversarial models and spreading mechanisms that have not been previously studied in the source-finding literature. We theoretically prove that Bitcoin's networking protocols (both pre- and post-2015) offer poor anonymity properties on networks with a regular-tree topology. We validate this claim in simulation on a 2015 snapshot of the real Bitcoin P2P network topology.

Bitcoin and other cryptocurrencies have surged in popularity over the last decade. Although Bitcoin does not claim to provide anonymity for its users, it enjoys a public perception of being a `privacy-preserving' financial system. In reality, cryptocurrencies publish users' entire transaction histories in plaintext, albeit under a pseudonym; this is required for transaction validation. Therefore, if a user's pseudonym can be linked to their human identity, the privacy fallout can be significant. Recently, researchers have demonstrated deanonymization attacks that exploit weaknesses in the Bitcoin network's peer-to-peer (P2P) networking protocols. In particular, the P2P network currently forwards content in a structured way that allows observers to deanonymize users. In this work, we redesign the P2P network from first principles with the goal of providing strong, provable anonymity guarantees. We propose a simple networking policy called Dandelion, which achieves nearly-optimal anonymity guarantees at minimal cost to the network's utility. We also provide a practical implementation of Dandelion.

In recent years governments have shown themselves willing to impose blackouts to shut off key communication infrastructure during times of civil strife, and to surveil citizen communications whenever possible. However, it is exactly during such strife that citizens need reliable and anonymous communications the most. In this paper, we present Rangzen, a system for anonymous broadcast messaging during network blackouts. Rangzen is distinctive in both aim and design. Our aim is to provide an anonymous, one-to-many messaging layer that requires only users' smartphones and can withstand network-level attacks. Our design is a delay-tolerant mesh network which deprioritizes adversarial messages by means of a social graph while preserving user anonymity. We built a complete implementation that runs on Android smartphones, present benchmarks of its performance and battery usage, and present simulation results suggesting Rangzen's efficacy at scale.