User Profiles In Social Networks Research Articles

Hyperspherical Variational Co-embedding for Attributed Networks

Network-based information has been widely explored and exploited in the information retrieval literature. Attributed networks, consisting of nodes, edges as well as attributes describing properties of nodes, are a basic type of network-based data, and are especially useful for many applications. Examples include user profiling in social networks and item recommendation in user-item purchase networks. Learning useful and expressive representations of entities in attributed networks can provide more effective building blocks to down-stream network-based tasks such as link prediction and attribute inference. Practically, input features of attributed networks are normalized as unit directional vectors. However, most network embedding techniques ignore the spherical nature of inputs and focus on learning representations in a Gaussian or Euclidean space, which, we hypothesize, might lead to less effective representations. To obtain more effective representations of attributed networks, we investigate the problem of mapping an attributed network with unit normalized directional features into a non-Gaussian and non-Euclidean space. Specifically, we propose a hyperspherical variational co-embedding for attributed networks (HCAN), which is based on generalized variational auto-encoders for heterogeneous data with multiple types of entities. HCAN jointly learns latent embeddings for both nodes and attributes in a unified hyperspherical space such that the affinities between nodes and attributes can be captured effectively. We argue that this is a crucial feature in many real-world applications of attributed networks. Previous Gaussian network embedding algorithms break the assumption of uninformative prior, which leads to unstable results and poor performance. In contrast, HCAN embeds nodes and attributes as von Mises-Fisher distributions, and allows one to capture the uncertainty of the inferred representations. Experimental results on eight datasets show that HCAN yields better performance in a number of applications compared with nine state-of-the-art baselines.

Research of the aviation personnel vulnerability profile to social engineering attacks

In conditions of strengthening the informational component of aviation activity, the task of ensuring aviation cybersecurity becomes extremely urgent. Currently, a regulatory framework is being developed that regulates activities in this area, both on the part of the International Civil Aviation Organization and at the Russian Federation level. In the complex of aviation cybersecurity threats, which include deliberate attacks, errors of third-party companies, system errors, natural phenomena, the human factor occupies an important place. In this work, this negative phenomenon is considered from the point of view of the aviation personnel vulnerability to social engineering attacks. Such type of attack by an attacker involves a set of applied psychological and analytical techniques that facilitate the receipt of confidential information or the violation of information security rules by legitimate company employees. The existing approach to building a profile of user vulnerabilities to social engineering attacks involves a series of psychological tests, the results of which are used to predict the user vulnerability through its psychological characteristics. In this work a slightly different task is posed, the main idea is to restore the vulnerability profile of aviation personnel from activity data in a social network. This is due to the fact that studying the user profile of a social network will more quickly solve the problem of choosing the most vulnerable employee for a particular type of social engineering attack and introduce preventive measures. The research was conducted on the basis of JSC «Surgut International Airport». 36 aviation security inspectors were selected as the respondents. Empirical data have been obtained including profiles of social network user profiles and a number of psychological tests. Using factor analysis the problem of reducing dimensionality and choosing the most informative indicators characterizing the activity of a social network user has been solved. A discriminant model that allows predicting the vulnerability profile of personnel according to the social network has been developed. Possible types of social engineering attacks on aviation personnel are presented.

Making Decisions by Unlabeled Bits

In a statistical decision problem, one observes independent samples and is asked to decide between two mutually exclusive states of nature. With knowledge of the data distributions under the two hypotheses, it is known that the asymptotically optimal error probabilities converge to zero at a rate given by the error exponent function, which quantifies the information for detection contained in the data. As a new element, suppose that the decision problem is to be solved without data labeling, i.e., with no knowledge of which observation comes from which distribution. What is the error exponent in this case? How much information for detection is contained in the samples and how much is lost with the labels? This problem — unlabeled detection — is of theoretical relevance per se , but is also attracting growing practical interest, because loss of labels often occurs naturally in big-data settings, user profiling in social networks; and may be deliberate, for example to save bandwidth in inference involving large sensor networks. For binary unlabeled observations and focusing on the low-detectability regime, we derive simple closed-form expressions for the error exponent and related quantities (Chernoff-Stein exponent, Chernoff information), providing new insights. Practical algorithms are then discussed, showing that many decision algorithms proposed in the literature reduce for binary data to simple forms, which highlight their properties and relative merits. A detector with close-to-optimum performance for a wide class of detection problems is proposed.

Behavioral Habits-Based User Identification Across Social Networks

Social networking is an interactive Internet of Things. The symmetry of the network can reflect the similar friendships of users on different social networks. A user’s behavior habits are not easy to change, and users usually have the same or similar display names and published contents among multiple social networks. Therefore, the symmetry concept can be used to analyze the information generated by the user for user identification. User identification plays a key role in building better information about social network user profiles. As a consequence, it has very important practical significance in many network applications and has attracted a great deal of attention from researchers. However, existing works are primarily focused on rich network data and ignore the difficulty involved in data acquisition. Display names and user-published content are very easy to obtain compared to other types of user data across different social networks. Therefore, this paper proposes an across social networks user identification method based on user behavior habits (ANIUBH). We analyzed the user’s personalized naming habits in terms of display names, then utilized different similarity calculation methods to measure the similarity of the features contained in the display names. The variant entropy value was adopted to assign weights to the features mentioned above. In addition, we also measured and analyzed the user’s interest graph to further improve user identification performance. Finally, we combined one-to-one constraint with the Gale–Shapley algorithm to eliminate the one-to-many and many-to-many account-matching problems that often occur during the results-matching process. Experimental results demonstrated that our proposed method enables the possibility of user identification using only a small amount of online data.

Using partial evaluation in holistic subgraph search

Because of its wide application, the subgraph matching problem has been studied extensively during the past decade. However, most existing solutions assume that a data graph is a vertex/edge-labeled graph (i.e., each vertex/edge has a simple label). These solutions build structural indices by considering the vertex labels. However, some real graphs contain rich-content vertices such as user profiles in social networks and HTML pages on the World Wide Web. In this study, we consider the problem of subgraph matching using a more general scenario. We build a structural index that does not depend on any vertex content. Based on the index, we design a holistic subgraph matching algorithm that considers the query graph as a whole and finds one match at a time. In order to further improve efficiency, we propose a “partial evaluation and assembly” framework to find subgraph matches over large graphs. Last but not least, our index has light maintenance overhead. Therefore, our method can work well on dynamic graphs. Extensive experiments on real graphs show that our method outperforms the state-of-the-art algorithms.

Soft quantification in statistical relational learning

We present a new statistical relational learning (SRL) framework that supports reasoning with soft quantifiers, such as “most” and “a few.” We define the syntax and the semantics of this language, which we call $$\hbox {PSL}^Q$$ , and present a most probable explanation inference algorithm for it. To the best of our knowledge, $$\hbox {PSL}^Q$$ is the first SRL framework that combines soft quantifiers with first-order logic rules for modelling uncertain relational data. Our experimental results for two real-world applications, link prediction in social trust networks and user profiling in social networks, demonstrate that the use of soft quantifiers not only allows for a natural and intuitive formulation of domain knowledge, but also improves inference accuracy.

Word Embeddings for User Profiling in Online Social Networks

User profiling in social networks can besignificantly augmented by using available full-text itemssuch as posts or statuses and ratings (in the form oflikes) that users give them. In this work, we applymodern natural language processing techniques basedon word embeddings to several problems related touser profiling in social networks. First, we present anapproach to create user profiles that measure a user’sinterest in various topics mined from the full texts of theitems. As a result, we get a user profile that can be used,e.g., for cold start recommendations for items, targetedadvertisement, and other purposes; our experimentsshow that the interests mining method performs on alevel comparable with collaborative algorithms while atthe same time being a cold start approach, i.e., itdoes not use the likes of an item being recommended.Second, we study the problem of predicting a user’sdemographic attributes such as age and gender basedon his or her full-text items. We evaluate theefficiency of various age prediction algorithms based onword2vec word embeddings and conduct an extensiveexperimental evaluation, comparing these algorithmswith each other and with classical baseline approaches.

Linking user profiles in social networks: a comparative review

The existence of user profiles belonging to a single user across different social networking sites poses several challenges to the research community. The main technical issue in that context is to detect the same user profiles across several different social networks by leveraging a set of mechanisms that identify the similarity among the user profiles. This problem is commonly referred to as entity matching or identity linkage on social networks. In this review, we describe and compare the 27 most important (to the best of our knowledge) research papers in this area. The main contributions of this article are to provide a systematic and integrated review of papers in this area, to provide comparative points that simplify the understanding of such systems, and finally to discuss future research avenues.

Linking user profiles in social networks: a comparative review

Linking user profiles in social networks: a comparative review

Inferring social network user profiles using a partial social graph

User profile inference on online social networks is a key task for targeted advertising and building recommender systems that rely on social network data. However, current algorithms for user profiling suffer from one or more of the following limitations: (1) assuming that the full social graph or a large training set of crawled data is available for training, (2) not exploiting the rich information that is available in social networks such as group memberships and likes, (3) treating numeric attributes as nominal attributes, and (4) not assessing the certainty of their predictions. In this paper, to address these limitations, we propose an algorithm named Partial Graph Profile Inference+ (PGPI+). The PGPI+ algorithm can accurately infer user profiles under the constraint of a partial social graph. PGPI+ does not require training, and it lets the user select the trade-off between the amount of information to be crawled for inferring a user profile and the accuracy the inference. Besides, PGPI+ is designed to use rich information about users when available: user profiles, friendship links, group memberships, and the "views" and "likes" from social networks such as Facebook. Moreover, to also address limitations 3 and 4, PGPI+ considers numeric attributes in addition to nominal attributes, and can evaluate the certainty of its predictions. An experimental evaluation with 31,247 user profiles from the Facebook and Pokec social networks shows that PGPI+ predicts user profiles with a higher accuracy than several start-of-the-art algorithms, and by accessing (crawling) less information from the social graph. Furthermore, an interesting result is that some profile attributes such as the status (student/professor) and genre can be predicted with more than 95 % accuracy using PGPI+.

Finding email correspondents in online social networks

Email correspondents play an important role in many people’s social networks. Finding email correspondents in social networks accurately, though may seem to be straightforward at a first glance, is challenging. Most of the existing online social networking sites recommend possible matches by comparing the information of email accounts and social network profiles, such as display names and email addresses. However, as shown empirically in this paper, such methods may not be effective in practice. To the best of our knowledge, this problem has not been carefully and thoroughly addressed in research. In this paper, we systematically investigate the problem and develop a practical data mining approach. We find that using only the profiles or the graph structures is far from effective. Our method utilizes the similarity between email accounts and social network user profiles, and at the same time explores the similarity between the email communication network and the social network under investigation. We demonstrate the effectiveness of our method using two real data sets on emails and Facebook.