This paper introduces 3MEthTaskforce (https://3meth.github.io), a multi-source, multi-level, and multi-token Ethereum dataset addressing the limitations of single-source datasets. Integrating over 300 million transaction records, 3,880 token profiles, global market indicators, and Reddit sentiment data from 2014-2024, it enables comprehensive studies on user behavior, market sentiment, and token performance. 3MEthTaskforce defines benchmarks for user behavior prediction and token price prediction tasks, using 6 dynamic graph networks and 19 time-series models to evaluate performance. Its multimodal design supports risk analysis and market fluctuation modeling, providing a valuable resource for advancing blockchain analytics and decentralized finance research.
Directed Acyclic Graph (DAG)-based Distributed Ledger Technologies (DLTs) have emerged as a promising solution to the scalability issues inherent in traditional blockchains. However, amidst the focus on scalability, the crucial aspect of privacy within DAG-based DLTs has been largely overlooked. This paper seeks to address this gap by providing a comprehensive examination of privacy notions and challenges within DAG-based DLTs. We delve into potential methodologies to enhance privacy within these systems, while also analyzing the associated hurdles and real-world implementations within state-of-the-art DAG-based DLTs. By exploring these methodologies, we not only illuminate the current landscape of privacy in DAG-based DLTs but also outline future research directions in this evolving field.
This paper presents an in-depth exploration of Data Availability Sampling (DAS) and sharding mechanisms within decentralized systems through simulation-based analysis. DAS, a pivotal concept in blockchain technology and decentralized networks, is thoroughly examined to unravel its intricacies and assess its impact on system performance. Through the development of a simulator tailored explicitly for DAS, we embark on a comprehensive investigation into the parameters that influence system behavior and efficiency. A series of experiments are conducted within the simulated environment to validate theoretical formulations and dissect the interplay of DAS parameters. This includes an exploration of approaches such as custody by row, variations in validators per node, and malicious nodes. The outcomes of these experiments furnish insights into the efficacy of DAS protocols and pave the way for the formulation of optimization strategies geared towards enhancing decentralized network performance. Moreover, the findings serve as guidelines for future research endeavors, offering a nuanced understanding of the complexities inherent in decentralized systems. This study not only contributes to the theoretical understanding of DAS but also offers practical implications for the design, implementation, and optimization of decentralized systems.
Daniel Kang, Tatsunori Hashimoto, Ion Stoica, Yi Sun
As ML models have increased in capabilities and accuracy, so has the complexity of their deployments. Increasingly, ML model consumers are turning to service providers to serve the ML models in the ML-as-a-service (MLaaS) paradigm. As MLaaS proliferates, a critical requirement emerges: how can model consumers verify that the correct predictions were served, in the face of malicious, lazy, or buggy service providers? In this work, we present the first practical ImageNet-scale method to verify ML model inference non-interactively, i.e., after the inference has been done. To do so, we leverage recent developments in ZK-SNARKs (zero-knowledge succinct non-interactive argument of knowledge), a form of zero-knowledge proofs. ZK-SNARKs allows us to verify ML model execution non-interactively and with only standard cryptographic hardness assumptions. In particular, we provide the first ZK-SNARK proof of valid inference for a full resolution ImageNet model, achieving 79\% top-5 accuracy. We further use these ZK-SNARKs to design protocols to verify ML model execution in a variety of scenarios, including for verifying MLaaS predictions, verifying MLaaS model accuracy, and using ML models for trustless retrieval. Together, our results show that ZK-SNARKs have the promise to make verified ML model inference practical.
Typically, pseudo-Computerized Tomography (CT) synthesis schemes proposed in the literature rely on complete atlases acquired with the same field of view (FOV) as the input volume. However, clinical CTs are usually acquired in a reduced FOV to decrease patient ionization. In this work, we present the Franken-CT approach, showing how the use of a non-parametric atlas composed of diverse anatomical overlapping Magnetic Resonance (MR)-CT scans and deep learning methods based on the U-net architecture enable synthesizing extended head and neck pseudo-CTs. Visual inspection of the results shows the high quality of the pseudo-CT and the robustness of the method, which is able to capture the details of the bone contours despite synthesizing the resulting image from knowledge obtained from images acquired with a completely different FOV. The experimental Zero-Normalized Cross-Correlation (ZNCC) reports 0.9367 ± 0.0138 (mean ± SD) and 95% confidence interval (0.9221, 0.9512); the experimental Mean Absolute Error (MAE) reports 73.9149 ± 9.2101 HU and 95% confidence interval (66.3383, 81.4915); the Structural Similarity Index Measure (SSIM) reports 0.9943 ± 0.0009 and 95% confidence interval (0.9935, 0.9951); and the experimental Dice coefficient for bone tissue reports 0.7051 ± 0.1126 and 95% confidence interval (0.6125, 0.7977). The voxel-by-voxel correlation plot shows an excellent correlation between pseudo-CT and ground-truth CT Hounsfield Units (m = 0.87; adjusted R2 = 0.91; p < 0.001). The Bland–Altman plot shows that the average of the differences is low (−38.6471 ± 199.6100; 95% CI (−429.8827, 352.5884)). This work serves as a proof of concept to demonstrate the great potential of deep learning methods for pseudo-CT synthesis and their great potential using real clinical datasets.
Gayan Benedict, Chief Information Officer, Reserve Bank of Australia, Australia
Distributed Ledger Technology (DLT) and its blockchain subclass are emerging as a disruptive innovation with an expanding range of potential applications. While DLT systems promise significant benefits to participants and diverse industries, their use of cryptography and incentive models to displace intermediaries and central authorities present significant challenges to conventional models of institutionally-oriented governance. This is due to the disintermediation of such governance bodies displacing the accountabilities and decision rights such parties typically provide. This paper addresses the emerging research question of how standards can address the assurances lost by the displacement of conventional institutional governance mechanisms in DLT systems.
Zero-knowledge proofs are mathematical cryptographic methods to demonstrate the validity of a claim while providing no further information beyond the claim itself. The possibility of using such proofs to process classified and other sensitive physical data has attracted attention, especially in the field of nuclear arms control. Here we demonstrate a non-electronic fast neutron differential radiography technique using superheated emulsion detectors that can confirm that two objects are identical without revealing their geometry or composition. Such a technique could form the basis of a verification system that could confirm the authenticity of nuclear weapons without sharing any secret design information. More broadly, by demonstrating a physical zero-knowledge proof that can compare physical properties of objects, this experiment opens the door to developing other such secure proof-systems for other applications.
We designed a concept of high resolution and quantitative SPECT for imaging a selected small region-of-interest (ROI) of human brain. This system is aimed at achieving high resolution less than 1 mm and being applied for imaging neurons and evaluating drug delivery system. Pinhole or cone-beam collimators are useful for high-resolution imaging of small ROI. However, when the ROI is smaller than the object, the projection data are truncated by radioisotope outside ROI. In the reconstructed image, the truncation causes the artifact and the overestimation of voxel value, which deceases quantitative accuracy of physiological functions. We are introducing the new truncation compensated 3D-OSEM (TC-3DOSEM) reconstruction method. The truncated data can be successfully reconstructed within ROI by fulfilling the condition that ROI contains a priori knowledge. In addition to small field-of-view (FOV) detector, we are introducing the parallel-hole collimator attached large FOV detector covering the entire brain, to acquire the non-truncated data and provide the priori knowledge in small ROI, even if the resolution of the detector is low. For imaging with high resolution, we are using LaBr3(Ce) scintillator with optically coupled to position-sensitive photomultiplier tube (H8500, Hamamatsu, Japan) as the detector. And also, for proof of our concept, we performed preliminary experiment using pinhole SPECT and brain phantom. The reconstruction ROI contained the region outside the brain, that is, zero count as the priori knowledge. The truncated data were reconstructed by TC-3DOSEM. The reconstructed image without artifact and overestimation was obtained with high resolution. This preliminary experiment suggested feasibility of high resolution and quantitative SPECT for imaging a selected small ROI of human brain.
It is shown that the MART (multiplicative algebraic reconstruction technique) algorithm can be derived by POCS. This gives MART a new theoretical interpretation and a proof of convergence to a stable solution even when other convex constraints are introduced. However, MART, as a multiplicative algorithm, depends on the initial solution. It is noted that, far from being a flaw, this property can be used to introduce further a priori knowledge about the image to be reconstructed, to maximize the entropy, to keep the ratio between the regions of the original image constant, or to set to zero the area outside the reconstruction volume. MART should be preferred to MENT (a maximum entropy algorithm) for entropy maximization, for it performs as well but is much faster. ART is much less influenced by the initial solution than MART.>