Ari Juels

Ari Juels, Farinaz Koushanfar

We propose protected pipelines or props for short, a new approach for authenticated, privacy-preserving access to deep-web data for machine learning (ML). By permitting secure use of vast sources of deep-web data, props address the systemic bottleneck of limited high-quality training data in ML development. Props also enable privacy-preserving and trustworthy forms of inference, allowing for safe use of sensitive data in ML applications. Props are practically realizable today by leveraging privacy-preserving oracle systems initially developed for blockchain applications.

Kushal Babel, Philip Daian, Mahimna Kelkar et al.

We introduce the Clockwork Finance Framework (CFF), a general purpose, formal verification framework for mechanized reasoning about the economic security properties of composed decentralized-finance (DeFi) smart contracts. CFF features three key properties. It is contract complete, meaning that it can model any smart contract platform and all its contracts--Turing complete or otherwise. It does so with asymptotically constant model overhead. It is also attack-exhaustive by construction, meaning that it can automatically and mechanically extract all possible economic attacks on users' cryptocurrency across modeled contracts. Thanks to these properties, CFF can support multiple goals: economic security analysis of contracts by developers, analysis of DeFi trading risks by users, fees UX, and optimization of arbitrage opportunities by bots or miners. Because CFF offers composability, it can support these goals with reasoning over any desired set of potentially interacting smart contract models. We instantiate CFF as an executable model for Ethereum contracts that incorporates a state-of-the-art deductive verifier. Building on previous work, we introduce extractable value (EV), a new formal notion of economic security in composed DeFi contracts that is both a basis for CFF and of general interest. We construct modular, human-readable, composable CFF models of four popular, deployed DeFi protocols in Ethereum: Uniswap, Uniswap V2, Sushiswap, and MakerDAO, representing a combined 24 billion USD in value as of March 2022. We use these models along with some other common models such as flash loans, airdrops and voting to show experimentally that CFF is practical and can drive useful, data-based EV-based insights from real world transaction activity. Without any explicitly programmed attack strategies, CFF uncovers on average an expected $56 million of EV per month in the recent past.

Michael Mirkin, Yan Ji, Jonathan Pang et al.

Proof-of-work (PoW) cryptocurrency blockchains like Bitcoin secure vast amounts of money. Their operators, called miners, expend resources to generate blocks and receive monetary rewards for their effort. Blockchains are, in principle, attractive targets for Denial-of-Service (DoS) attacks: There is fierce competition among coins, as well as potential gains from short selling. Classical DoS attacks, however, typically target a few servers and cannot scale to systems with many nodes. There have been no successful DoS attacks to date against prominent cryptocurrencies. We present Blockchain DoS (BDoS), the first incentive-based DoS attack that targets PoW cryptocurrencies. Unlike classical DoS, BDoS targets the system's mechanism design: It exploits the reward mechanism to discourage miner participation. Previous DoS attacks against PoW blockchains require an adversary's mining power to match that of all other miners. In contrast, BDoS can cause a blockchain to grind to a halt with significantly fewer resources, e.g., 21% as of March 2020 in Bitcoin, according to our empirical study. We find that Bitcoin's vulnerability to BDoS increases rapidly as the mining industry matures and profitability drops. BDoS differs from known attacks like Selfish Mining in its aim not to increase an adversary's revenue, but to disrupt the system. Although it bears some algorithmic similarity to those attacks, it introduces a new adversarial model, goals, algorithm, and game-theoretic analysis. Beyond its direct implications for operational blockchains, BDoS introduces the novel idea that an adversary can manipulate miners' incentives by proving the existence of blocks without actually publishing them.

Charlie Hou, Mingxun Zhou, Yan Ji et al.

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.

Fan Zhang, Sai Krishna Deepak Maram, Harjasleen Malvai et al.

Thanks to the widespread deployment of TLS, users can access private data over channels with end-to-end confidentiality and integrity. What they cannot do, however, is prove to third parties the {\em provenance} of such data, i.e., that it genuinely came from a particular website. Existing approaches either introduce undesirable trust assumptions or require server-side modifications. As a result, the value of users' private data is locked up in its point of origin. Users cannot export their data with preserved integrity to other applications without help and permission from the current data holder. We propose DECO (short for \underline{dec}entralized \underline{o}racle) to address the above problems. DECO allows users to prove that a piece of data accessed via TLS came from a particular website and optionally prove statements about such data in zero-knowledge, keeping the data itself secret. DECO is the first such system that works without trusted hardware or server-side modifications. DECO can liberate data from centralized web-service silos, making it accessible to a rich spectrum of applications. To demonstrate the power of DECO, we implement three applications that are hard to achieve without it: a private financial instrument using smart contracts, converting legacy credentials to anonymous credentials, and verifiable claims against price discrimination.

Philip Daian, Steven Goldfeder, Tyler Kell et al.

Blockchains, and specifically smart contracts, have promised to create fair and transparent trading ecosystems. Unfortunately, we show that this promise has not been met. We document and quantify the widespread and rising deployment of arbitrage bots in blockchain systems, specifically in decentralized exchanges (or "DEXes"). Like high-frequency traders on Wall Street, these bots exploit inefficiencies in DEXes, paying high transaction fees and optimizing network latency to frontrun, i.e., anticipate and exploit, ordinary users' DEX trades. We study the breadth of DEX arbitrage bots in a subset of transactions that yield quantifiable revenue to these bots. We also study bots' profit-making strategies, with a focus on blockchain-specific elements. We observe bots engage in what we call priority gas auctions (PGAs), competitively bidding up transaction fees in order to obtain priority ordering, i.e., early block position and execution, for their transactions. PGAs present an interesting and complex new continuous-time, partial-information, game-theoretic model that we formalize and study. We release an interactive web portal, http://frontrun.me/, to provide the community with real-time data on PGAs. We additionally show that high fees paid for priority transaction ordering poses a systemic risk to consensus-layer security. We explain that such fees are just one form of a general phenomenon in DEXes and beyond---what we call miner extractable value (MEV)---that poses concrete, measurable, consensus-layer security risks. We show empirically that MEV poses a realistic threat to Ethereum today. Our work highlights the large, complex risks created by transaction-ordering dependencies in smart contracts and the ways in which traditional forms of financial-market exploitation are adapting to and penetrating blockchain economies.

Raymond Cheng, Fan Zhang, Jernej Kos et al.

Smart contracts are applications that execute on blockchains. Today they manage billions of dollars in value and motivate visionary plans for pervasive blockchain deployment. While smart contracts inherit the availability and other security assurances of blockchains, however, they are impeded by blockchains' lack of confidentiality and poor performance. We present Ekiden, a system that addresses these critical gaps by combining blockchains with Trusted Execution Environments (TEEs). Ekiden leverages a novel architecture that separates consensus from execution, enabling efficient TEE-backed confidentiality-preserving smart-contracts and high scalability. Our prototype (with Tendermint as the consensus layer) achieves example performance of 600x more throughput and 400x less latency at 1000x less cost than the Ethereum mainnet. Another contribution of this paper is that we systematically identify and treat the pitfalls arising from harmonizing TEEs and blockchains. Treated separately, both TEEs and blockchains provide powerful guarantees, but hybridized, though, they engender new attacks. For example, in naive designs, privacy in TEE-backed contracts can be jeopardized by forgery of blocks, a seemingly unrelated attack vector. We believe the insights learned from Ekiden will prove to be of broad importance in hybridized TEE-blockchain systems.

Florian Tramèr, Fan Zhang, Ari Juels et al.

Machine learning (ML) models may be deemed confidential due to their sensitive training data, commercial value, or use in security applications. Increasingly often, confidential ML models are being deployed with publicly accessible query interfaces. ML-as-a-service ("predictive analytics") systems are an example: Some allow users to train models on potentially sensitive data and charge others for access on a pay-per-query basis. The tension between model confidentiality and public access motivates our investigation of model extraction attacks. In such attacks, an adversary with black-box access, but no prior knowledge of an ML model's parameters or training data, aims to duplicate the functionality of (i.e., "steal") the model. Unlike in classical learning theory settings, ML-as-a-service offerings may accept partial feature vectors as inputs and include confidence values with predictions. Given these practices, we show simple, efficient attacks that extract target ML models with near-perfect fidelity for popular model classes including logistic regression, neural networks, and decision trees. We demonstrate these attacks against the online services of BigML and Amazon Machine Learning. We further show that the natural countermeasure of omitting confidence values from model outputs still admits potentially harmful model extraction attacks. Our results highlight the need for careful ML model deployment and new model extraction countermeasures.