arXiv subjects

Roger Wattenhofer

Publications and source records attributed to Roger Wattenhofer.

At least 109 records · Page 6Linked to original sources

arXiv2024

Can an AI Agent Safely Run a Government? Existence of Probably Approximately Aligned Policies

While autonomous agents often surpass humans in their ability to handle vast and complex data, their potential misalignment (i.e., lack of transparency regarding their true objective) has thus far hindered their use in critical applications such as social decision processes. More importantly, existing alignment methods provide no formal guarantees on the safety of such models. Drawing from utility and social choice theory, we provide a novel quantitative definition of alignment in the context of social decision-making. Building on this definition, we introduce probably approximately aligned (i.e., near-optimal) policies, and we derive a sufficient condition for their existence. Lastly, recognizing the practical difficulty of satisfying this condition, we introduce the relaxed concept of safe (i.e., nondestructive) policies, and we propose a simple yet robust method to safeguard the black-box policy of any autonomous agent, ensuring all its actions are verifiably safe for the society.

cs.AI↗

arXiv2024

Tiny Transformers Excel at Sentence Compression

It is staggering that words of the English language, which are on average represented by 5--6 bytes of ASCII, require as much as 24 kilobytes when served to large language models. We show that there is room for more information in every token embedding. We demonstrate that 1--3-layer transformers are capable of encoding and subsequently decoding standard English sentences into as little as a single 3-kilobyte token. Our work implies that even small networks can learn to construct valid English sentences and suggests the possibility of optimising large language models by moving from sub-word token embeddings towards larger fragments of text.

cs.LG↗

arXiv2024

On the Lifecycle of a Lightning Network Payment Channel

The Bitcoin Lightning Network, launched in 2018, serves as a layer 2 scaling solution for Bitcoin. The Lightning Network allows users to establish channels between each other and subsequently exchange off-chain payments. Together, these channels form a network that facilitates payments between parties even if they do not have a channel in common. The Lightning Network has gained popularity over the past five years as it offers an attractive alternative to on-chain transactions by substantially reducing transaction costs and processing times. Nevertheless, due to the privacy-centric design of the Lightning Network, little is understood about its inner workings. In this work, we conduct a measurement study of the Lightning Network to shed light on the lifecycle of channels. By combining Lightning gossip messages with on-chain Bitcoin data, we investigate the lifecycle of a channel from its opening through its lifetime to its closing. In particular, our analysis offers unique insights into the utilization patterns of the Lightning Network. Even more so, through decoding the channel closing transactions, we obtain the first dataset of Lightning Network payments, observe the imbalance of channels during the closing, and investigate whether both parties are involved in the closing, or one closes the channel unilaterally. For instance, we find nearly 60% of cooperatively closed channels are resurrected, i.e., their outputs were used to fund another channel.

cs.DC↗

arXiv2024

Sybil Detection using Graph Neural Networks

This paper presents SYBILGAT, a novel approach to Sybil detection in social networks using Graph Attention Networks (GATs). Traditional methods for Sybil detection primarily leverage structural properties of networks; however, they tend to struggle with a large number of attack edges and are often unable to simultaneously utilize both known Sybil and honest nodes. Our proposed method addresses these limitations by dynamically assigning attention weights to different nodes during aggregations, enhancing detection performance. We conducted extensive experiments in various scenarios, including pretraining in sampled subgraphs, synthetic networks, and networks under targeted attacks. The results show that SYBILGAT significantly outperforms the state-of-the-art algorithms, particularly in scenarios with high attack complexity and when the number of attack edges increases. Our approach shows robust performance across different network models and sizes, even as the detection task becomes more challenging. We successfully applied the model to a real-world Twitter graph with more than 269k nodes and 6.8M edges. The flexibility and generalizability of SYBILGAT make it a promising tool to defend against Sybil attacks in online social networks with only structural information.

cs.SI↗

arXiv2024

Towards Leveraging Contrastively Pretrained Neural Audio Embeddings for Recommender Tasks

Music recommender systems frequently utilize network-based models to capture relationships between music pieces, artists, and users. Although these relationships provide valuable insights for predictions, new music pieces or artists often face the cold-start problem due to insufficient initial information. To address this, one can extract content-based information directly from the music to enhance collaborative-filtering-based methods. While previous approaches have relied on hand-crafted audio features for this purpose, we explore the use of contrastively pretrained neural audio embedding models, which offer a richer and more nuanced representation of music. Our experiments demonstrate that neural embeddings, particularly those generated with the Contrastive Language-Audio Pretraining (CLAP) model, present a promising approach to enhancing music recommendation tasks within graph-based frameworks.

cs.SD↗

arXiv2024

Breaking reCAPTCHAv2

Our work examines the efficacy of employing advanced machine learning methods to solve captchas from Google's reCAPTCHAv2 system. We evaluate the effectiveness of automated systems in solving captchas by utilizing advanced YOLO models for image segmentation and classification. Our main result is that we can solve 100% of the captchas, while previous work only solved 68-71%. Furthermore, our findings suggest that there is no significant difference in the number of challenges humans and bots must solve to pass the captchas in reCAPTCHAv2. This implies that current AI technologies can exploit advanced image-based captchas. We also look under the hood of reCAPTCHAv2, and find evidence that reCAPTCHAv2 is heavily based on cookie and browser history data when evaluating whether a user is human or not. The code is provided alongside this paper.

cs.CV↗

arXiv2024

Assessing Adversarial Robustness of Large Language Models: An Empirical Study

Large Language Models (LLMs) have revolutionized natural language processing, but their robustness against adversarial attacks remains a critical concern. We presents a novel white-box style attack approach that exposes vulnerabilities in leading open-source LLMs, including Llama, OPT, and T5. We assess the impact of model size, structure, and fine-tuning strategies on their resistance to adversarial perturbations. Our comprehensive evaluation across five diverse text classification tasks establishes a new benchmark for LLM robustness. The findings of this study have far-reaching implications for the reliable deployment of LLMs in real-world applications and contribute to the advancement of trustworthy AI systems.

cs.CL↗

arXiv2024

Seeing Through the Mask: Rethinking Adversarial Examples for CAPTCHAs

Modern CAPTCHAs rely heavily on vision tasks that are supposedly hard for computers but easy for humans. However, advances in image recognition models pose a significant threat to such CAPTCHAs. These models can easily be fooled by generating some well-hidden "random" noise and adding it to the image, or hiding objects in the image. However, these methods are model-specific and thus can not aid CAPTCHAs in fooling all models. We show in this work that by allowing for more significant changes to the images while preserving the semantic information and keeping it solvable by humans, we can fool many state-of-the-art models. Specifically, we demonstrate that by adding masks of various intensities the Accuracy @ 1 (Acc@1) drops by more than 50%-points for all models, and supposedly robust models such as vision transformers see an Acc@1 drop of 80%-points. These masks can therefore effectively fool modern image classifiers, thus showing that machines have not caught up with humans -- yet.

cs.CV↗

arXiv2024

The Potential of Self-Regulation for Front-Running Prevention on DEXes

The transaction ordering dependency of the smart contracts building decentralized exchanges (DEXes) allow for predatory trading strategies. In particular, front-running attacks present a constant risk for traders on DEXes. Whereas legal regulation outlaws most front-running practices in traditional finance, such measures are ineffective in preventing front-running on DEXes. While novel market designs hindering front-running may emerge, it remains unclear whether the market's participants, in particular, liquidity providers, would be willing to adopt these new designs. A misalignment of the participant's private incentives and the market's social incentives can hinder the market from adopting an effective prevention mechanism. We present a game-theoretic model to study the behavior of sophisticated traders, retail traders, and liquidity providers in DEXes. Sophisticated traders adjust for front-running attacks, while retail traders do not, likely due to lack of knowledge or irrationality. Our findings show that with less than 1% of order flow from retail traders, traders' and liquidity providers' interests align with the market's social incentives - eliminating front-running attacks. However, the benefit from embracing this novel market is often small and may not suffice to entice them. With retail traders making up a larger proportion (around 10%) of the order flow, liquidity providers tend to stay in pools that do not protect against front-running. This suggests both educating traders and providing additional incentives for liquidity providers are necessary for market self-regulation.

cs.GT↗

arXiv2024

Tyche: Collateral-Free Coalition-Resistant Multiparty Lotteries with Arbitrary Payouts

We propose Tyche, a family of protocols for performing practically (as well as asymptotically) efficient multiparty lotteries, resistant against aborts and majority coalitions. Our protocols are based on a commit-and-reveal approach, requiring only a collision-resistant hash function. All our protocols use a blockchain as a public bulletin board and for buy-in collection and payout settlement. Importantly though, they do not rely on it or any other third party for providing randomness. Also, participants are not required to post any collateral beyond their buy-in. Any honest participant can eventually settle the lottery, and dishonest behavior never reduces the winning probability of any honest participant. Further, we adapt all three protocols into anonymous lotteries, where (under certain conditions) the winner is unlinkable to any particular participant. We show that our protocols are secure, fair, and some preserve the participants' privacy. Finally, we evaluate the performance of our protocols, particularly in terms of transaction fees, by implementing them on the Sui blockchain. There we see that per user transaction fees are reasonably low and our protocols could potentially support millions of participants.

cs.CR↗

arXiv2024

Quantifying Liveness and Safety of Avalanche's Snowball

This work examines the resilience properties of the Snowball and Avalanche protocols that underlie the popular Avalanche blockchain. We experimentally quantify the resilience of Snowball using a simulation implemented in Rust, where the adversary strategically rebalances the network to delay termination. We show that in a network of $n$ nodes of equal stake, the adversary is able to break liveness when controlling $Ω(\sqrt{n})$ nodes. Specifically, for $n = 2000$, a simple adversary controlling $5.2\%$ of stake can successfully attack liveness. When the adversary is given additional information about the state of the network (without any communication or other advantages), the stake needed for a successful attack is as little as $2.8\%$. We show that the adversary can break safety in time exponentially dependent on their stake, and inversely linearly related to the size of the network, e.g. in 265 rounds in expectation when the adversary controls $25\%$ of a network of 3000. We conclude that Snowball and Avalanche are akin to Byzantine reliable broadcast protocols as opposed to consensus.

cs.DC↗

arXiv2024

Towards Learning Abductive Reasoning using VSA Distributed Representations

We introduce the Abductive Rule Learner with Context-awareness (ARLC), a model that solves abstract reasoning tasks based on Learn-VRF. ARLC features a novel and more broadly applicable training objective for abductive reasoning, resulting in better interpretability and higher accuracy when solving Raven's progressive matrices (RPM). ARLC allows both programming domain knowledge and learning the rules underlying a data distribution. We evaluate ARLC on the I-RAVEN dataset, showcasing state-of-the-art accuracy across both in-distribution and out-of-distribution (unseen attribute-rule pairs) tests. ARLC surpasses neuro-symbolic and connectionist baselines, including large language models, despite having orders of magnitude fewer parameters. We show ARLC's robustness to post-programming training by incrementally learning from examples on top of programmed knowledge, which only improves its performance and does not result in catastrophic forgetting of the programmed solution. We validate ARLC's seamless transfer learning from a 2x2 RPM constellation to unseen constellations. Our code is available at https://github.com/IBM/abductive-rule-learner-with-context-awareness.

cs.LG↗

arXiv2024

Next Level Message-Passing with Hierarchical Support Graphs

Message-Passing Neural Networks (MPNNs) are extensively employed in graph learning tasks but suffer from limitations such as the restricted scope of information exchange, by being confined to neighboring nodes during each round of message passing. Various strategies have been proposed to address these limitations, including incorporating virtual nodes to facilitate global information exchange. In this study, we introduce the Hierarchical Support Graph (HSG), an extension of the virtual node concept created through recursive coarsening of the original graph. This approach provides a flexible framework for enhancing information flow in graphs, independent of the specific MPNN layers utilized. We present a theoretical analysis of HSGs, investigate their empirical performance, and demonstrate that HSGs can surpass other methods augmented with virtual nodes, achieving state-of-the-art results across multiple datasets.

cs.LG↗

arXiv2024

SUBER: An RL Environment with Simulated Human Behavior for Recommender Systems

Reinforcement learning (RL) has gained popularity in the realm of recommender systems due to its ability to optimize long-term rewards and guide users in discovering relevant content. However, the successful implementation of RL in recommender systems is challenging because of several factors, including the limited availability of online data for training on-policy methods. This scarcity requires expensive human interaction for online model training. Furthermore, the development of effective evaluation frameworks that accurately reflect the quality of models remains a fundamental challenge in recommender systems. To address these challenges, we propose a comprehensive framework for synthetic environments that simulate human behavior by harnessing the capabilities of large language models (LLMs). We complement our framework with in-depth ablation studies and demonstrate its effectiveness with experiments on movie and book recommendations. Using LLMs as synthetic users, this work introduces a modular and novel framework to train RL-based recommender systems. The software, including the RL environment, is publicly available on GitHub.

cs.IR↗

arXiv2024

GraphFSA: A Finite State Automaton Framework for Algorithmic Learning on Graphs

Many graph algorithms can be viewed as sets of rules that are iteratively applied, with the number of iterations dependent on the size and complexity of the input graph. Existing machine learning architectures often struggle to represent these algorithmic decisions as discrete state transitions. Therefore, we propose a novel framework: GraphFSA (Graph Finite State Automaton). GraphFSA is designed to learn a finite state automaton that runs on each node of a given graph. We test GraphFSA on cellular automata problems, showcasing its abilities in a straightforward algorithmic setting. For a comprehensive empirical evaluation of our framework, we create a diverse range of synthetic problems. As our main application, we then focus on learning more elaborate graph algorithms. Our findings suggest that GraphFSA exhibits strong generalization and extrapolation abilities, presenting an alternative approach to represent these algorithms.

cs.AI↗

arXiv2024

SoK: Attacks on DAOs

Decentralized Autonomous Organizations (DAOs) are blockchain-based organizations that facilitate decentralized governance. Today, DAOs not only hold billions of dollars in their treasury but also govern many of the most popular Decentralized Finance (DeFi) protocols. This paper systematically analyses security threats to DAOs, focusing on the types of attacks they face. We study attacks on DAOs that took place in the past, attacks that have been theorized to be possible, and potential attacks that were uncovered and prevented in audits. For each of these (potential) attacks, we describe and categorize the attack vectors utilized into four categories. This reveals that while many attacks on DAOs take advantage of the less tangible and more complex human nature involved in governance, audits tend to focus on code and protocol vulnerabilities. Thus, additionally, the paper examines empirical data on DAO vulnerabilities, outlines risk factors contributing to these attacks, and suggests mitigation strategies to safeguard against such vulnerabilities.

cs.CR↗

arXiv2024

AEye: A Visualization Tool for Image Datasets

Image datasets serve as the foundation for machine learning models in computer vision, significantly influencing model capabilities, performance, and biases alongside architectural considerations. Therefore, understanding the composition and distribution of these datasets has become increasingly crucial. To address the need for intuitive exploration of these datasets, we propose AEye, an extensible and scalable visualization tool tailored to image datasets. AEye utilizes a contrastively trained model to embed images into semantically meaningful high-dimensional representations, facilitating data clustering and organization. To visualize the high-dimensional representations, we project them onto a two-dimensional plane and arrange images in layers so users can seamlessly navigate and explore them interactively. AEye facilitates semantic search functionalities for both text and image queries, enabling users to search for content. We open-source the codebase for AEye, and provide a simple configuration to add datasets.

cs.CV↗

arXiv2024

Graph Dimension Attention Networks for Enterprise Credit Assessment

Enterprise credit assessment is critical for evaluating financial risk, and Graph Neural Networks (GNNs), with their advanced capability to model inter-entity relationships, are a natural tool to get a deeper understanding of these financial networks. However, existing GNN-based methodologies predominantly emphasize entity-level attention mechanisms for contagion risk aggregation, often overlooking the heterogeneous importance of different feature dimensions, thus falling short in adequately modeling credit risk levels. To address this issue, we propose a novel architecture named Graph Dimension Attention Network (GDAN), which incorporates a dimension-level attention mechanism to capture fine-grained risk-related characteristics. Furthermore, we explore the interpretability of the GNN-based method in financial scenarios and propose a simple but effective data-centric explainer for GDAN, called GDAN-DistShift. DistShift provides edge-level interpretability by quantifying distribution shifts during the message-passing process. Moreover, we collected a real-world, multi-source Enterprise Credit Assessment Dataset (ECAD) and have made it accessible to the research community since high-quality datasets are lacking in this field. Extensive experiments conducted on ECAD demonstrate the effectiveness of our methods. In addition, we ran GDAN on the well-known datasets SMEsD and DBLP, also with excellent results.

cs.LG↗