Detecting vulnerabilities in smart contracts is vital for the security and reliability of decentralized apps. To facilitate vulnerability detection, contract codes, including bug patterns, are represented as heterogeneous graphs with various nodes and edges, like control-flow and function-call graphs. However, existing graph-learning techniques struggle with large, complex graphs. This article presents MANDO-LLM, a novel framework that combines heterogeneous graph transformers (HGTs) with large language models (LLMs) for detecting vulnerabilities in smart contracts represented as heterogeneous contract graphs built upon control-flow and call graphs. MANDO-LLM uses LLMs to capture code features from control-flow and call data, customizes HGTs to learn embeddings with specific node-edge meta relations, and employs classifiers for vulnerability detection in Solidity code at both contract and line levels. Our evaluation shows that MANDO-LLM significantly outperforms existing methods on real-world large-scale imbalanced datasets, with F1-score improvements from 0.59% to 80.72% at the contract level. It is also one of the first effective methods for identifying line-level vulnerabilities, with performance boosts ranging from 3.09% to over 95% across different vulnerability types. MANDO-LLM’s versatility allows easy retraining for various vulnerabilities without needing manually defined patterns.

Recent advancements in deep learning for Acoustic Scene Classification (ASC) have significantly improved the ability to discern and categorize complex soundscapes. Nonetheless, the real-world application of these technologies presents notable challenges, particularly regarding computational demands and adaptability to diverse environmental conditions. To address key issues such as recording device mismatch, strict memory and complexity constraints, and the limited availability of labeled data, we introduce ASCDomain - an innovative framework designed for efficient and adaptable acoustic scene analysis. Our approach combines a teacher-student knowledge distillation mechanism, where an ensemble of high-performance models serves as the teacher, enhancing the training process. This is further supported by a domain adversarial neural network that ensures robust domain adaptation. By integrating a compact isotropic neural network as the student, our system minimizes computational requirements while reducing the dependency on extensive labeled datasets. Validated on the TAU Urban Acoustic Scenes 2022 Mobile dataset, ASCDomain demonstrates leading-edge performance. The code is available at https://github.com/hubtru/ASCDomain.

An automatic classification of the malignancy of lung nodules in computed tomography (CT) scans can support early detection of lung cancer, which is crucial for the treatment success. The novel photon-counting CT (PCCT) technology enables high image quality with a low radiation dose and provides additional spectral information. This research focuses on whether PCCT scans offer a benefit in the automatic classification of lung nodules. Establishing a dataset of PCCT images poses several challenges, such as the extraction of annotations or the data imbalance.

Multi-label data is prevalent across various applications, where instances can be annotated with a set of classes. Although multi-label data can take various forms, such as images and text, tabular multi-label data stands out as the predominant data type in many real-world scenarios. Over the past decades, numerous methods have been proposed for tabular multi-label classification. Effectively addressing challenges like class imbalance, correlation among labels and features, and scalability is crucial for a high-performance multi-label classifier. However, many existing methods fall short of fully considering the correlation between labels and features. In cases where attempts are made, they often encounter high computational costs, rendering them impractical for large datasets. This paper in- troduces an innovative classification method for tabular multi-label data, utilizing a fusion of transformers and graph convolutional networks (GCN). The central concept of the proposed approach involves transforming tabular data into images, leveraging state-of-the-art methods in image processing, including image-based transformers and pre-trained models to capture correlation among labels effectively. Our approach jointly learns the representation of feature space and the correlation among labels within a unified network. To substantiate the performance of our proposed method, we conducted a rigorous series of experiments across diverse multi-label datasets1. The results underscore the superior performance and scalability of our approach compared to other existing state-of-the-art methods. This work not only contributes a novel perspective to the field of tabular multi-label classification but also showcases advancements in both accuracy and scalability.

The progress of deep-learning-based forecasting architectures is evident through their expanding parameter configurations. However, the need for rapid online decision making in practical scenarios calls for an alternative strategy, highlighting the necessity for networks that are not only adaptive but also efficient in real-time operations. This shift is critical as we confront three principal challenges in deep-learning-based forecasting frameworks: (i) the inherent limitations of transformers, which, despite their attempts to preserve ordering information, the temporal information loss due to the permutation-invariant nature of self-attention mechanisms is inevitable, (ii) the inefficacy of linear models in capturing the dynamic interactions within swiftly evolving signals; and (iii) the incapacity of tree-based approaches to extrapolating beyond values present in the training set. In response to these challenges, we introduce LTBoost, an innovative boosted hybrid of linear and tree-based ensemble gradient algorithms tailored for long-term time series forecasting (LTSF) tasks, scalable to high data dimensions. LTBoost employs a dual strategy, beginning with a linear regression model to capture trends and extrapolate beyond known data, complemented by a robust nonlinear tree-based model that focuses on the residuals. This boosted hybrid approach not only addresses the challenges posed by existing models but also significantly improves forecast accuracy. The effectiveness of LTBoost is validated through empirical experiments conducted on nine well-established benchmark datasets, demonstrating superior performance and achieving state-of-the-art results in 32 out of 36 cases, measured by mean absolute error (MAE). Our findings also explore the impact of lag features and signal normalization techniques, demonstrating further improvements in predictive accuracy. This hybrid and highly effective approach highlights LTBoost's innovation and its resolution of specific forecasting challenges, setting the stage for its contribution to the field of time series forecasting, paving the way for its application in diverse real-world scenarios.

Entity matching, also known as user identity linkage, is a critical task in data integration. While established techniques primarily focus on large-scale networks, there are several applications where small networks pose challenges due to limited training data and sparsity. This study addresses entity matching in the field of criminology, where small networks are common and the number of known matching nodes is restricted. To support this research, we exploit a multimodal dataset, collected as part of a security-related project, consisting of an intercepted telephone calls network (i.e., ROXSD data) and a network of social forum interactions (i.e., ROXHOOD data) collected in a simulated environment, although following real investigation scenario. To improve accuracy and efficiency, we propose a novel approach for entity matching across these two small networks using node attributes. Existing techniques often merely focus on topology consistency between two networks and overlook valuable information, such as network node attributes, making them vulnerable to structural changes. Inspired by the remarkable success of deep learning, we present UGC-DeepLink, an end-to-end semi-supervised learning framework that leverages user-generated content. UGC-DeepLink encodes network nodes into vector representations, capturing both local and global network structures to align anchor nodes using deep neural networks. A dual learning paradigm and the policy gradient method transfer knowledge and update the linkage. Additionally, node attributes, such as call contents and forum exchanged texts, enhance the ranking of matching nodes. Experimental results on ROXSD and ROXHOOD demonstrate that UGC-DeepLink surpasses baselines and state-of-the-art methods in terms of identity-match ranking. The code and dataset are available at https://github.com/erichoang/UGC-DeepLink.

Time Series Classification (TSC) aims to develop predictive models for discrete target variables using ordered, real-valued attributes. However, existing deep learning approaches face challenges in addressing the inherent dependencies across data dimensions and the dynamic characteristics of time series data, often resulting in insufficient feature extraction and suboptimal classification accuracy. This paper introduces HyperTime, a novel framework designed to overcome these challenges by modeling time series data as hypergraphs, where hyperedges connect arbitrary sets of nodes. HyperTime distinguishes itself by processing time series data into segmented windows, which form nodes in a hypergraph connected by hyperedges. This approach captures complex temporal relationships and dynamics within the data. The framework employs ‘hypernode embedding’, applying an attention mechanism to each time series dimension within a window, and ‘hyperedge embedding’, incorporating an LSTM layer for deeper temporal analysis. A key innovation in HyperTime is the ‘hyper convolution’ operation, which includes both hypernode and hyperedge convolutions. Nodes are convoluted using the embeddings of their associated hyperedges, enriching the learning and understanding of their role within the hypergraph. The hyperedge convolution layer integrates local and global information, aggregating node features to capture complex interactions and relationships. The effectiveness of HyperTime is validated through extensive experiments on 26 datasets from the UEA archive, the Human Activity Recognition (HAR) and PAMAP2 Physical Activity Monitoring (PAM) datasets, and 10 datasets from the UCR archive.

Conventional time series classification approaches based on bags of patterns or shapelets face significant challenges in dealing with a vast amount of feature candidates from high-dimensional multivariate data. In contrast, deep neural networks can learn low-dimensional features efficiently, and in particular, convolutional neural networks have shown promising results in classifying multivariate time series data. A key factor in the success of deep neural networks is this astonishing expressive power. However, this power comes at the cost of complex, black-boxed models, conflicting with the goals of building reliable and human-understandable models. In this work1, we introduce a new interpretable framework for multivariate time series data that by extracting and clustering the input quantifies the contribution of time-varying input variables and each signal’s role to the classification. We construct a graph that captures the temporal relationship between the extracted patterns for each layer and propose an effective merging strategy to aggregate those graphs into one. Finally, a graph embedding algorithm generates new representations of the created interpretable time-series features. Our extensive experiments indicate the benefit of our time-aware graph-based representation in multivariate time series classification while enriching them with more interpretability.

Although Internet of Things (IoT) systems are widely used in various industries, they are prone to data collection errors due to device limitations and environmental factors. These errors can significantly degrade the quality of collected data and the event log extracted from raw sensor readings, impact data analysis and lead to inaccurate or distorted results. This article emphasizes the importance of evaluating data quality and errors before proceeding with analysis. The effectiveness of three error correction methods, a rule-based method and a Process Mining (PM)-based method which are adjusted for a smart home use case, and their combination was also investigated in resolving log errors. The study found that understanding different types and sources of errors, and adapting the error correction algorithm based on this knowledge of error sources, can greatly improve the algorithm’s efficiency in addressing various error types.

Increasing privacy concerns have led to decentralized and federated machine learning techniques that allow individual clients to consult and train models collaboratively without sharing private information. Some of these applications, such as medical and healthcare, require the final decisions to be interpretable. One common form of data in these applications is multivariate time series, where deep neural networks, especially convolutional neural networks based approaches, have established excellent performance in their classification tasks. However, promising results and performance of deep learning models are a black box, and their decisions cannot always be guaranteed and trusted. While several approaches address the interpretability of deep learning models for multivariate time series data in a centralized environment, less effort has been made in a federated setting. In this work, we introduce FLAMES2Graph, a new horizontal federated learning framework designed to interpret the deep learning decisions of each client. FLAMES2Graph extracts and visualizes those input subsequences that are highly activated by a convolutional neural network. Besides, an evolution graph is created to capture the temporal dependencies between the extracted distinct subsequences. The federated learning clients only share this temporal evolution graph with the centralized server instead of trained model weights to create a global evolution graph. Our extensive experiments on various datasets from well-known multivariate benchmarks indicate that the FLAMES2Graph framework significantly outperforms other state-of-the-art federated methods while keeping privacy and augmenting network decision interpretation.

Smart contracts in blockchains have been increasingly used for high-value business applications. It is essential to check smart contracts' reliability before and after deployment. Although various program analysis and deep learning techniques have been proposed to detect vulnerabilities in either Ethereum smart contract source code or bytecode, their detection accuracy and scalability are still limited. This paper presents a novel framework named MANDO-HGT for detecting smart contract vulnerabilities. Given Ethereum smart contracts, either in source code or bytecode form, and vulnerable or clean, MANDO-HGT custom-builds heterogeneous contract graphs (HCGs) to represent control-flow and/or function-call information of the code. It then adapts heterogeneous graph transformers (HGTs) with customized meta relations for graph nodes and edges to learn their embeddings and train classifiers for detecting various vulnerability types in the nodes and graphs of the contracts more accurately. We have collected more than 55K Ethereum smart contracts from various data sources and verified the labels for 423 buggy and 2,742 clean contracts to evaluate MANDO-HGT. Our empirical results show that MANDO-HGT can significantly improve the detection accuracy of other state-of-the-art vulnerability detection techniques that are based on either machine learning or conventional analysis techniques. The accuracy improvements in terms of F1-score range from 0.7% to more than 76% at either the coarse-grained contract level or the fine-grained line level for various vulnerability types in either source code or bytecode. Our method is general and can be retrained easily for different vulnerability types without the need for manually defined vulnerability patterns.

Learning heterogeneous graphs consisting of different types of nodes and edges enhances the results of homogeneous graph techniques. An interesting example of such graphs is control-flow graphs representing possible software code execution flows. As such graphs represent more semantic information of code, developing techniques and tools for such graphs can be highly beneficial for detecting vulnerabilities in software for its reliability. However, existing heterogeneous graph techniques are still insufficient in handling complex graphs where the number of different types of nodes and edges is large and variable. This paper concentrates on the Ethereum smart contracts as a sample of software codes represented by heterogeneous contract graphs built upon both control-flow graphs and call graphs containing different types of nodes and links. We propose MANDO, a new heterogeneous graph representation to learn such heterogeneous contract graphs’ structures. MANDO extracts customized meta-paths, which compose relational connections between different types of nodes and their neighbors. Moreover, it develops a multi-metapath heterogeneous graph attention network to learn multi-level embeddings of different types of nodes and their metapaths in the heterogeneous contract graphs, which can capture the code semantics of smart contracts more accurately and facilitate both fine-grained line-level and coarse-grained contract-level vulnerability detection. Our extensive evaluation of large smart contract datasets shows that MANDO improves the vulnerability detection results of other techniques at the coarse-grained contract level. More importantly, it is the first learning-based approach capable of identifying vulnerabilities at the fine-grained line-level, and significantly improves the traditional code analysis-based vulnerability detection approaches by 11.35% to 70.81% in terms of F1-score.

Deep neural networks, especially the Convolutional Neural Network (CNN) models, have shown promising results in multivariate time series data analysis. However, the predictions of these data-driven black-box models are tough to interpret from a human perspective, making it questionable to trust and rely on the predictions made by these models, specifically for time series data with the append-only feature. This paper proposes a new approach to interpret the CNN outputs by extracting and clustering the activated time series sequences learned from a trained network. These sequences show the representative features for each output label and form interpretable representations from the original time series data. Our approach is the first framework to identify each signal’s role and dependencies, consider all possible combinations of signals in the multivariate time-series input, and visualize the data representative features. Our experiments on the Baydogan’s archive indicate remarkable improvements in the interpretability of the network predictions and relation identification of each input signal to the output label and the channels of the network layers. Furthermore, the conducted experiments confirm that the extracted patterns are representative of the multivariate input and changing them results in a drastic reduction in the prediction accuracy.

Social networks have become an inseparable part of human activities. Most existing social networks follow a centralized system model, which despite storing valuable information of users, arise many critical concerns such as content ownership and over-commercialization. Recently, decentralized social networks, built primarily on blockchain technology, have been proposed as a substitution to eliminate these concerns. Since decentralized architectures are mature enough to be on par with the centralized ones, decentralized social networks are becoming more and more popular. Decentralized social networks can offer both common options like writing posts and comments and more advanced options such as reward systems and voting mechanisms. They provide rich eco-systems for the influencers to interact with their followers and other users via staking systems based on cryptocurrency tokens. The vast and valuable data of the decentralized social networks open several new directions for the research community to extend human behavior knowledge. However, accessing and collecting data from these social networks is not easy because it requires strong blockchain knowledge, which is not the main focus of computer science and social science researchers. Hence, our work proposes the SoChainDB framework that facilitates obtaining data from these new social networks. To show the capacity and strength of SoChainDB, we crawl and publish Hive data - one of the largest blockchain-based social networks. We conduct extensive analyses to understand the insight of Hive data and discuss some interesting applications, e.g., game, non-fungible tokens market built upon Hive. It is worth mentioning that our framework is well-adaptable to other blockchain social networks with minimal modification. SoChainDB is publicly accessible at http://sochaindb.com and the dataset is available under the CC BY-SA 4.0 license.

Smart contracts are increasingly used with blockchain systems for high-value applications. It is highly desired to ensure the quality of smart contract source code before they are deployed. This paper proposes a new deep learning-based tool, MANDO-GURU, that aims to accurately detect vulnerabilities in smart contracts at both coarse-grained contract-level and fine-grained line-level. Using a combination of control-flow graphs and call graphs of Solidity code, we design new heterogeneous graph attention neural networks to encode more structural and potentially semantic relations among different types of nodes and edges of such graphs and use the encoded embeddings of the graphs and nodes to detect vulnerabilities. Our validation of real-world smart contract datasets shows that MANDO-GURU can significantly improve many other vulnerability detection techniques by up to 24% in terms of the F1-score at the contract level, depending on vulnerability types. It is the first learning-based tool for Ethereum smart contracts that identify vulnerabilities at the line level and significantly improves the traditional code analysis-based techniques by up to 63.4%. Our tool is publicly available at https://github.com/MANDO-Project/ge-sc-machine. A test version is currently deployed at http://mandoguru.com, and a demo video of our tool is available at http://mandoguru.com/demo-video.

Abstract The massive production of data in the web environment increases the need for efficient and effective solutions to store and manage data. NoSQL databases with the capability of storing various types of data were introduced to be used in the management of big data. Besides, supporting flexible querying has significant impacts on data management usability and users prefer to ask their requests using fuzzy terms and get crisp results to make efficient decisions. In this paper, a fuzzy‐based approach is presented which adopts fuzzy queries and delivers crisp recommendations to the users. For this purpose, a counselor, which acts as a fuzzy agent and has knowledge about the data, defines fuzzy terms, membership functions and fuzzy rule tables in a way that each user can ask queries through fuzzy terms. The evaluation results show the efficiency of the proposed approach which removes the need for high‐level programming skills for users. As a result, our approach's efficiency and user acceptance are increased as well and made it an appropriate solution for using ambiguous queries in big graph databases.

Due to their disclosure required by law, business management reports have become publicly available for a large number of companies, and these reports offer the opportunity to assess the financial health or distress of a company, both quantitatively from the balance sheets and qualitatively from the text. In this paper, we analyze the potential of deep sentiment mining from the textual parts of business management reports and aim to detect signals for financial distress. We (1) created the largest corpus of business reports analyzed qualitatively to date, (2) defined a non-trivial target variable based on the so-called Altman Z-score, (3) developed a filtering of sentences based on class-correlated pattern mining to reduce the complexity of these long and complex texts, and (4) employed one of the best-performing machine learning methods for this type of task, Dependency Sensitive Convolutional Neural Networks (DSCNNs). Experimental results show that strong prediction performance can be achieved by a suitable bundle of methods, with an F1-score of more than 0.86 and a Kappa score of more than 65%. To better understand the parts of management reports that indicate financial distress, the prediction engine is complemented by a visualization tool that highlights critical text passages.

Density estimation of streaming data is a relevant task in numerous domains. In this paper, a novel non-parametric density estimator called FRONT (forest of normalized trees) is introduced. It uses a structure of multiple normalized trees, segments the feature space of the data stream through a periodically updated linear transformation and is able to adapt to ever evolving data streams. FRONT provides accurate density estimation and performs favorably compared to existing online density estimators in terms of the average log score on multiple standard data sets. Its low complexity, linear runtime as well as constant memory usage, makes FRONT by design suitable for large data streams. Finally, the paper provides a variation of FRONT called N-FRONT suitable for statistically independent data streams and correction methods for badly initialized trees to further improve performance.

Subgroup discovery is the task of identifying subgroups that show the most unusual statistical (distributional) characteristics with respect to a given target variable, at the intersection of predictive and descriptive induction. Redundancy and lack of rule interpretability constitute the major challenges in subgroup discovery today. We address these two issues by constrained latent Dirichlet allocation (LDA) to identify co-occurring feature values (descriptions) for subgroup rule search, obtaining a less redundant and more diverse rule set. Latent Dirichlet Allocation, as a topic modeling approach, is able to identify diverse topics, from which the rules can be derived. The resulting rules are less redundant and can also be interpreted by the corresponding topic. Experimental results on six benchmark datasets show that the presented approach provides rule sets with better rule redundancy and diversity compared to those of four existing algorithms. One unique and interesting advantage of the proposed method is that it can categorize rules by topics as well as the assignment of a probability to each feature value of a discovered rule, which can be used in the interpretation of the results.

Data stream mining has gained growing attentions due to its wide emerging applications such as target marketing, email filtering and network intrusion detection. In this paper, we propose a prototype-based classification model for evolving data streams, called SyncStream, which dynamically models time-changing concepts and makes predictions in a local fashion. Instead of learning a single model on a sliding window or ensemble learning, SyncStream captures evolving concepts by dynamically maintaining a set of prototypes in a new data structure called the P-tree. The prototypes are obtained by error-driven representativeness learning and synchronization-inspired constrained clustering. To identify abrupt concept drift in data streams, PCA and statistics based heuristic approaches are employed. SyncStream has several attractive benefits: (a) It is capable of dynamically modeling evolving concepts from even a small set of prototypes and is robust against noisy examples. (b) Owing to synchronization-based constrained clustering and the P-Tree, it supports an efficient and effective data representation and maintenance. (c) Gradual and abrupt concept drift can be effectively detected. Empirical results shows that our method achieves good predictive performance compared to state-of-the-art algorithms and that it requires much less time than another instance-based stream mining algorithm.