Online social networks can be modelled as graphs; in this paper, we analyze the use of graph metrics for identifying users with anomalous relationships to other users. A framework is proposed for analyzing the effectiveness of various graph theoretic properties such as the number of neighbouring nodes and edges, betweenness centrality, and community cohesiveness in detecting anomalous users. Experimental results on real-world data collected from online social networks show that the majority of users typically have friends who are friends themselves, whereas anomalous users’ graphs typically do not follow this common rule. Empirical analysis also shows that the relationship between average betweenness centrality and edges identifies anomalies more accurately than other approaches.

Analyzing the effectiveness of graph metrics for anomaly detection in online social networks

Motivated by the recent surge of criminal activities with cross-cryptocurrency trades, we introduce a new topological perspective to structural anomaly detection in dynamic multilayer networks. We postulate that anomalies in the underlying blockchain transaction graph that are composed of multiple layers are likely to also be manifested in anomalous patterns of the network shape properties. As such, we invoke the machinery of clique persistent homology on graphs to systematically and efficiently track evolution of the network shape and, as a result, to detect changes in the underlying network topology and geometry. We develop a new persistence summary for multilayer networks, called stacked persistence diagram, and prove its stability under input data perturbations. We validate our new topological anomaly detection framework in application to dynamic multilayer networks from the Ethereum Blockchain and the Ripple Credit Network, and demonstrate that our stacked PD approach substantially outperforms state-of-art techniques.

Topological Anomaly Detection in Dynamic Multilayer Blockchain Networks.

Abstract Graph learning is being increasingly applied to image clustering to reveal intra-class and inter-class relationships in data. However, existing graph learning-based image clustering focuses on grouping images under a single view, which under-utilises the information provided by the data. To address that, we propose a self-supervised multi-view image clustering technique under contrastive heterogeneous graph learning. Our method computes a heterogeneous affinity graph for multi-view image data. It conducts Local Feature Propagation (LFP) for reasoning over the local neighbourhood of each node and executes an Influence-aware Feature Propagation (IFP) from each node to its influential node for learning the clustering intention. The proposed framework pioneeringly employs two contrastive objectives. The first targets to contrast and fuse multiple views for the overall LFP embedding, and the second maximises the mutual information between LFP and IFP representations. We conduct extensive experiments on the benchmark datasets for the problem, i.e. COIL-20, Caltech7 and CASIA-WebFace. Our evaluation shows that our method outperforms the state-of-the-art methods, including the popular techniques MVGL, MCGC and HeCo. 

/pdf/gomic-multi-view-image-clustering-via-self-supervised-nbb50rkd.pdf

GoMIC: Multi-view image clustering via self-supervised contrastive heterogeneous graph co-learning

Both node classification and link prediction are popular topics of supervised learning on the graph data, but previous works seldom integrate them together to capture their complementary information. In this paper, we propose a Multi-Task and Multi-Graph Convolutional Network (MTGCN) to jointly conduct node classification and link prediction in a unified framework. Specifically, MTGCN consists of multiple multi-task learning so that each multi-task learning learns the complementary information between node classification and link prediction. In particular, each multi-task learning uses different inputs to output representations of the graph data. Moreover, the parameters of one multi-task learning initialize the parameters of the other multi-task learning, so that the useful information in the former multi-task learning can be propagated to the other multi-task learning. As a result, the information is augmented to guarantee the quality of representations by exploring the complex constructure inherent in the graph data. Experimental results on six datasets show that our MTGCN outperforms the comparison methods in terms of both node classification and link prediction. 

MTGCN: A multi-task approach for node classification and link prediction in graph data

Drivers can make significant impacts on transportation systems. They can leave important information due to their social behaviors. But, the role of drivers has been overlooked yet. In this paper, for the first time, drivers' collaboration network is introduced. The network is considered in a heterogenous form, because of existence multiple relationships between drivers in the real-world situation. Since drivers do not belong to only one community, the overlapping of communities is considered and a new overlapping community detection algorithm is developed to discover the hidden structure of the network. Also, we present a new overlapping score to improve the community detection algorithm, using the adjacencies among non-memeber neighbor nodes and communities. Solving the algorithm will lead to discovering dense communities of drivers that have meaningful relationships with each other. This will result in a better understanding of the transportation network and also improving the overall performance of the system. A comparison of the developed algorithm with the others demonstrates the effectiveness of the algorithm. To evaluate the applicability of the algorithm, a real drivers' collaboration network is presented and the developed algorithm is applied to derive insights.

Introducing drivers' collaboration network: A two-layers social network perspective in road transportation system analysis

Multi-view data contains a set of features representing different perspectives associated with the same data and this phenomenon can be commonly observed in real-world applications. Multi-view clustering in terms of text and image data faces substantial challenges such as Structure-preserving and Sparsity. Existing methods do not conserve the structure of data space and the recent improvements have earmarked only the local layout. Preserving the local structure of data space is not sufficient to handle sparsity in these data. In this paper, we propose a novel clustering approach, called Proximity-based Multi-View Non-negative Matrix Factorization (PMVNMF), which utilizes both the local and global structure of data space conjointly to handle sparsity in real-world multimedia (text and image) data. For each view, the 1-step and 2-step transition probability matrices as the first-order and second-order proximity matrices are constructed to uncover their respective latent local and global geometric structures. Then, view-specific proximity matrices as an integration of the above two types of proximity matrices are constructed. Eventually, Non-negative Matrix Factorization (NMF) is explored via graph regularization and consensus regularization, to consider the obtained integrated graph structures as well as to disclose the indistinct common structure shared by all representations. The algorithm can capture elementary structure of data space and is robust to sparse data. We conduct experiments on six real-world datasets including two text and four image datasets; and compare the performance of the proposed algorithm with eight baseline approaches. Six evaluation metrics including accuracy, f-score, precision, recall, NMI, and entropy are employed to evaluate the performance of algorithm. The results show the outperformance of proposed algorithm over baselines.

A novel multi-view clustering approach via proximity-based factorization targeting structural maintenance and sparsity challenges for text and image categorization

The attributed multiplex network is a set of attributed networks in which each network represents a different type of interaction between the same set of nodes. Individual networks are termed as layers or dimensions and network nodes are characterized by attribute vectors. Neighborhood, in general, refers to any dense connected subgraph. We refer neighborhood1 as subgraph induced on graph node and its neighbors. It is usually observed that majority of the nodes in multilayer networks are active only on small number of layers except some outliers [18]. However, node activity is not strictly correlated to the edges incident in a node. A node might be active at few layers with relatively large number of incident edges and at the same time, multi-active node might not have many links even on single layer. Moreover, each layer has distinct importance in the multiplex networks2 and the structure and size of neighborhood formed by these multiplex nodes are different on each layer. Nodes with different attributes come together on different layers in the attributed multiplex networks. This node and layer heterogeneity should be considered while identifying anomalous neighborhoods in the attributed multiplex networks. Thus, a measure is required to quantify the quality of neighborhoods formed by active nodes on different layers. Existing approaches do not consider heterogeneity among network layers and do quantify the structure of networks either separately for each layer or its aggregated network and ignore the attributes of nodes. In this work, we define a novel quality measure Multi-Normality which utilizes the structure and attributes together of each layer and detect attribute coherence in neighborhoods between layers. We also propose an algorithm exhausting multi-normality to identify anomalous neighborhoods in multiplex networks and is named as Anomaly Detection of Entity Neighborhoods in Multiplex Networks (ADENMN). We evaluate the effectiveness of the proposed algorithm in anomaly detection by comparing its performance with three existing baseline approaches including ADOMS, AMM and AGG+AD on five real-world attributed multiplex networks including Amazon, YouTube, Noordin top terrorist network, DBLP_C, and Aarhus. The results of experiments demonstrate that multi-normality outperforms baseline algorithms.

Ranking and Discovering Anomalous Neighborhoods in Attributed Multiplex Networks

In this paper, an extension of Precision Time Protocol (PTP) to enable energy-efficient clock synchronization between the nodes within Wireless Sensor Network (WSN) is proposed. PTP is nanosecond accuracy clock synchronization protocol in which nodes are organized in master-slave hierarchy on the basis of clock accuracy by means of Best Master Clock (BMC) algorithm. The algorithm considers clock accuracy to select best clock in the system. A novel modification of IEEE 1588 BMC algorithm for energy-constraint multi-hop WSN has been proposed to reduce clock convergence time and energy needed by considering out-degree of clocks without sacrificing synchronization accuracy. The new algorithm results in energy efficient clock synchronization that makes it most appropriate for low-power multi-hop wireless sensor networks. We present NS-3 simulation data that confirms the effectiveness of work.

Monika Bansal

Papers

A novel multi-view clustering approach via proximity-based factorization targeting structural maintenance and sparsity challenges for text and image categorization

Ranking and Discovering Anomalous Neighborhoods in Attributed Multiplex Networks

Out-Degree Based Clock Synchronization In Wireless Networks Using Precision Time Protocol