Martin Harrigan

An Analysis of Anonymity in the Bitcoin System

Anonymity in Bit coin, a peer-to-peer electronic currency system, is a complicated issue. Within the system, users are identified by public-keys only. An attacker wishing to de-anonymize its users will attempt to construct the one to-many mapping between users and public-keys and associate information external to the system with the users. Bitcoinfrustrates this attack by storing the mapping of a user to his or her public-keys on that users node only and by allowing each user to generate as many public-keys as required. In this paper we consider the topological structure of two networks derived from Bitcoins public transaction history. We show that the two networks have a non-trivial topological structure, provide complementary views of the Bit coin system and have implications for anonymity. We combine these structures with external information and techniques such as context discovery and flow analysis to investigate an alleged theft of Bit coins, which, at the time of the theft, had a market value of approximately half a million U.S. dollars.

Practical level planarity testing and layout with embedding constraints

We describe a practical method to test a leveled graph for level planarity and provide a level planar layout of the graph if the test succeeds, all in quadratic running-time. Embedding constraints restricting the order of incident edges around the vertices are allowed.

Classifying Wikipedia articles using network motif counts and ratios

Because the production of Wikipedia articles is a collaborative process, the edit network around a article can tell us something about the quality of that article. Articles that have received little attention will have sparse networks; at the other end of the spectrum, articles that are Wikipedia battle grounds will have very crowded networks. In this paper we evaluate the idea of characterizing edit networks as a vector of motif counts that can be used in clustering and classification. Our objective is not immediately to develop a powerful classifier but to assess what is the signal in network motifs. We show that this motif count vector representation is effective for classifying articles on the Wikipedia quality scale. We further show that ratios of motif counts can effectively overcome normalization problems when comparing networks of radically different sizes.

EgoNav: exploring networks through egocentric spatializations

EgoNav is a visual analytics system that characterizes egos based on the relationship structure of their egocentric networks and presents the results as a spatialization. An ego, or individual node in a network, is most closely related to its neighbors, and to a lesser degree, to its neighbors neighbors. For example, in social networks, people are closely related to their friends and family. In financial networks, the affairs of borrowers and lenders are more closely tied to each other. In fact, the relationship structure surrounding an ego, or an egocentric network, can provide characteristic information about the ego itself. Using network motif analysis and dimensionality reduction techniques, the system places egos in similar areas of a spatialization if their egocentric networks are structurally similar. This view of a network discriminates between the various classes of typical and exceptional egos. We demonstrate its effectiveness using appropriate synthetic datasets, real-world mobile phone call and peer-to-peer lending datasets. We subsequently elicit user feedback from experts involved in the investigation of financial fraud to assess the tools applicability in this domain.

The Unreasonable Effectiveness of Address Clustering

Address clustering tries to construct the one-to-many mapping from entities to addresses in the Bitcoin system. Simple heuristics based on the micro-structure of transactions have proved very effective in practice. In this paper we describe the primary reasons behind this effectiveness: address reuse, avoidable merging, super-clusters with high centrality,, the incremental growth of address clusters. We quantify their impact during Bitcoins first seven years of existence.

Characterizing ego-networks using motifs

We assess the potential of network motif profiles to characterize ego-networks in much the same way that a bag-of-words strategy allows text documents to be compared in a vector space framework. This is potentially valuable as a generic strategy for comparing nodes in a network in terms of the network structure in which they are embedded. In this paper, we consider the computational challenges and model selection decisions involved in network motif profiling. We also present three case studies concerning the analysis of Wikipedia edit networks, YouTube spam campaigns, and peer-to-peer lending in the Prosper marketplace.

Mining dense structures to uncover anomalous behaviour in financial network data

The identification of anomalous user behaviour is important in a number of application areas, since it may be indicative of fraudulent activity. In the work presented here, the focus is on the identification and subsequent investigation of suspicious interactions in a network of financial transactions. A network is constructed from data from a peer-to-peer lending system, with links between members representing the initiation of loans. The network is time-sliced to facilitate temporal analysis. Anomalous network structure is sought in the time-sliced network, illustrating the occurrences of unusual behaviour among members. In order to assess the significance of the dense structures returned the enrichment of member attributes within these structures is examined. It seems that dense structures are associated with geographic regions.

Efficiently drawing a significant spanning tree of a directed graph

A directed graph can model any ordered relationship between objects. However, visualizing such graphs can be a challenging task. If the graph is undirected, a popular strategy is to choose a significant spanning tree, nominate a vertex as the root, for example the vertex whose distance from all other vertices is minimal, hang the significant spanning subtrees from this root and add in the remaining edges in some unobtrusive manner. In the directed case the spanning tree is a tree DAG (a directed graph without any undirected cycles) and not simply a directed tree with one appropriate root. It may have multiple sources (vertices with indegree equal to zero) that all warrant root status and so the undirected approach must be modified somewhat. In this paper, we present a method of drawing directed graphs that emphasizes a significant spanning tree. It can be considered a variation of the Sugiyama framework in that it combines two steps of the framework (leveling and crossing minimisation) by finding, in linear time, a leveling of the graph that is level planar with respect to some spanning tree and restricting the permutations of the vertices on each level to those that constitute a level planar embedding of this subgraph. The edges of the spanning tree will therefore not cross each other. Using the globally oriented Fiedler vector we choose permutations of the vertices on each level that reduce the number of crossings between the remaining edges.

Using Vector Clocks to Visualize Communication Flow

Given a dataset comprising a temporal sequence of communications between actors, how can we visualize the ‘flow’ of communication over time? Current practice transforms the dataset into a dynamic graph – vertices represent the actors and directed edges represent the communications. The directed edges are added and removed over time. There are then several approaches to visualizing dynamic graphs that optimize aesthetic criteria, most producing animated node-link diagrams. However, dynamic graphs are not the only way to model this problem. One alternative from the field of distributed computing is vector clocks. Recent work employed vector clocks to analyze communication flow in social networks with much effect, arguing that they provide new insights into the problem. In this paper, we use vector clocks as a basis for visualizing communication flow. We show that communication patterns, e.g., random, partitioned and core-periphery, are easily discernible in the resulting visualizations. We also argue that, in the cases where vector clocks are used to analyze communication flow, it is most natural to base the accompanying visualizations on vector clocks also.

Reports on the Workshops Held at the Sixth International AAAI Conference on Weblogs and Social Media

In 2012 the International Conference on Weblogs and Social Media (ICWSM 2012), held in Dublin, Ireland, focused primarily on cutting-edge research in social media. Most notably, the increasing influence of user-generated content in the newsroom was discussed. As social media are becoming more and more relevant as a source of information, traditional media organizations are faced with new challenges. Aanyone equipped with a smartphone can now capture and publish events as they unfold. In many cases established news agencies are not the first point of call for information anymore. In order to stay competitive media organizations are increasingly depending on content from social networks to cover and present all perspectives of an event. However, they face one crucial question when it comes to using content from these networks: “Is the source reliable?” At this workshop, we discussed existing approaches and ways in which some of the prevailing challenges are encountered when developing new methodologies. The keynote speaker, Katrin Weller, questioned whether Twitter is actually a social network. No, she states. Status updates on Twitter are covering all sort of topics — from popular culture to neuroscience, from intimate, personal details to major press releases of world-leading companies. In their totality, they can be considered as a giant but completely unstructured and unorganized knowledge base of what is going on. On the one hand this enables browsing and discovering new interesting pieces of information (serendipity effects). On the other hand it may pose enormous challenges to people looking for particular information. How can we access this rich resource of social content, which Reports