Interpreting user actions to better understand their needs provides an important tool for improving information access services. In the context of organic Web search, considerable effort has been made to model user behavior and infer query intent, with the goal of improving the overall user experience. Much less work has been done in the area of sponsored search, i.e., with respect to the advertisement links (ads) displayed on search result pages by many commercial search engines. This thesis develops and evaluates new models and methods required to interpret user browsing and click behavior and understand query intent in this very different context. The concern of the initial part of the thesis is on extending the query categories for commercial search and on inferring query intent, with a focus on two major tasks: i) enriching queries with contextual information obtained from search result pages returned for these queries, and ii) developing relatively simple methods for the reliable labeling of training data via crowdsourcing. A central idea of this thesis work is to study the impact of contextual factors (including query intent, ad placement, and page structure) on user behavior. Later, this information is incorporated into probabilistic models to evaluate the quality of advertisement links within the context that they are displayed in their history of appearance. In order to account for these factors, a number of query and location biases are proposed and formulated into a group of browsing and click models. To explore user intent and behavior and to evaluate the performance of the proposed models and methods, logs of query and click information provided for research purposes are used. Overall, query intent is found to have substantial impact on predictions of user click behavior in sponsored search. Predictions are further improved by considering ads in the context of the other ads displayed on a result page. The parameters of the browsing and click models are learned using an expectation maximization technique applied to click signals recorded in the logs. The initial motivation of the user to browse the ad list and their browsing persistence are found to be related to query intent and browsing/click behavior. Accommodating these biases along with the location bias in user models appear as effective contextual signals, improving the performance of the existing models.

With the development of artificial intelligence , the application of recommender systems has been extended to fields such as e-commerce shopping cart analysis or video recommendation system. These systems provide user a recommended resource set based on their habits or behavior patterns to help users saving searching cost. However, these techniques have not been successfully adopted to help users search functions on smart-phones more efficiency. This research is designated to build the context-aware system, which can generate the list of operations predicting which function user might use under certain contexts through continuously learning users operation patterns and related device perceived scenario. The system utilize event-condition-action patterns to describe user frequent behaviors, and the research will focus on developing innovative Action-Condition-Fit algorithm to figure the similarity between action pattern sets and real-time scenario. Proposed system and algorithm will then be built on Google App Engine and Android device to empirically validate its performance through field test.

Denna explorativa studie analyserar gemensamma och differentiella mönster för kommunikation och beteende hos användare i Hjälpforum. Målet med studien är att kartlägga aktiva användartyper med syfte att ge SVT kunskap om Hjälpforumets användare. Om forumadministratörer har mer kunskap om användarna i Hjälpforum kan de anpassa sin kommunikation utifrån användarnas individuella behov och öka deras förmåga att bidra med mer kvalitativ återkoppling. Detta kan i sin tur bidra till en positivare inställning och ökat förtroende hos användare i Hjälpforum. I denna studie kunde fem användartyper kartläggas i Hjälpforum med stöd av egen empiri och tidigare forskning. Nybörjare utgör främsta användartypen i Hjälpforum och SVT bör därför tillgodose deras behov för att eventuellt minska antalet nya användarinlägg. Dessutom kunde denna studie med hjälp av lämpliga databearbetningsverktyg kartlägga, en för tillfället, okänd användartyp i Hjälpforum utifrån ovanliga mönster i beteende och kommunikation vilket kan vara intressant för vidare analys. Studien presenterar förslag på hur resultatet kan användas för att skapa riktlinjer för framtida kommunikation och utveckling av Hjälpforum och SVT:s webbplats. / This exploratory study analyzes the common and differential patterns of communication and behavior of users in Help Forum. The goal of the study is to map the active user types with the aim of providing SVT knowledge of the Help Forum users. If the forum administrators have more knowledge of the users in the forum, they can adapt their communication based on the individual needs and enhance their ability to contribute more qualitative feedback. This may in turn contribute to a more positive attitude and greater confidence among users in Help Forum. In this study, five types of users could be identified in the Forum with the support of its own empirical data and previous research. Beginners are primary user type in the forum and SVT should cater to their needs in order to possibly reduce the number of new user posts. Furthermore, this study using appropriate data processing tool map, one for the moment, unknown user type in the forum by unusual patterns of behavior and communication which may be of interest for further analysis. The study presents suggestions on how the results can be used to create guidelines for future communication and development of Help Forum and SVT's website.

Interpreting user actions to better understand their needs provides an important tool for improving information access services. In the context of organic Web search, considerable effort has been made to model user behavior and infer query intent, with the goal of improving the overall user experience. Much less work has been done in the area of sponsored search, i.e., with respect to the advertisement links (ads) displayed on search result pages by many commercial search engines. This thesis develops and evaluates new models and methods required to interpret user browsing and click behavior and understand query intent in this very different context. The concern of the initial part of the thesis is on extending the query categories for commercial search and on inferring query intent, with a focus on two major tasks: i) enriching queries with contextual information obtained from search result pages returned for these queries, and ii) developing relatively simple methods for the reliable labeling of training data via crowdsourcing. A central idea of this thesis work is to study the impact of contextual factors (including query intent, ad placement, and page structure) on user behavior. Later, this information is incorporated into probabilistic models to evaluate the quality of advertisement links within the context that they are displayed in their history of appearance. In order to account for these factors, a number of query and location biases are proposed and formulated into a group of browsing and click models. To explore user intent and behavior and to evaluate the performance of the proposed models and methods, logs of query and click information provided for research purposes are used. Overall, query intent is found to have substantial impact on predictions of user click behavior in sponsored search. Predictions are further improved by considering ads in the context of the other ads displayed on a result page. The parameters of the browsing and click models are learned using an expectation maximization technique applied to click signals recorded in the logs. The initial motivation of the user to browse the ad list and their browsing persistence are found to be related to query intent and browsing/click behavior. Accommodating these biases along with the location bias in user models appear as effective contextual signals, improving the performance of the existing models.

The purpose of this paper has been to study how users experiences and behaviors are influenced by a "like" function, when publication of images in social media with the photo app Instagram as the selected base. In this study respondents has been part of a study regarding their experiences and behavior related to "like" function in Instagram. The results of the question form have then been analyzed to find patterns in their behavior and to exemplify/reinforce different type of user-behavior found in related scientific articles. The study has shown that the behavior of Instagram users is affected by the "like" function, when they believes that the function is a status marker that raises the interest for certain users by the number of likes. Through my results, I have been able to amplify and illustrate several theories of Goffman, Buckingham and Gripsrud which are presented in the report.

Context. With the development of computer software, hardware, and communication technologies, a new type of human-centered computing model, called Cloud Computing (CC) has been established as a commercial computer network service. However, the openness of CC brings huge security challenge to the identity-based access control system, as it not able to effectively prevent malicious users accessing; information security problems, system stability problems, and also the trust issues between cloud service users (CSUs) and cloud service providers (CSPs) are arising therefrom. User behavior trust (UBT) evaluation is a valid method to solve security dilemmas of identity-based access control system, but current studies of UBT based access control model is still not mature enough, existing the problems like UBT evaluation complexity, trust dynamic update efficiency, evaluation accuracy, etc. Objective. The aim of the study is to design and develop an improved UBT based CC access control model compare to the current state-of-art. Including an improved UBT evaluation method, able to reflect the user’s credibility according to the user’s interaction behavior, provides access control model with valid evidence to making access control decision; and a dynamic authorization control and re-allocation strategy, able to timely response to user’s malicious behavior during entire interaction process through real-time behavior trust evaluation. Timely updating CSUs trust value and re-allocating authority degree. Methods. This study presented a systematical literature review (SLR) to identify the working structure of UBT based access control model; summarize the CSUs’ behaviors that can be collected as UBT evaluation evidence; identify the attributes of trust that will affect the accuracy of UBT evaluation; and evaluated the current state-of-art of UBT based access control models and their potential advantages, opportunities, and weaknesses. Using the acquired knowledge, design a UBT based access control model, and adopt prototype method to simulate the performance of the model, in order to verify its validation, verify improvements, and limitations. Results. Through the SLR, two types of UBT based access control model working structures are identified and illustrated, essential elements are summarized, and a dynamic trust and access update module is described; 23 CSU’s behavior evidence items are identified and classified into three classes; four important trust attributes, influences, and corresponding countermeasures are identified and summarized; and eight current state-of-art of UBT based access control models are identified and evaluated. A Triple Dynamic Window based Access Control model (TDW) was designed and established as a prototype, the simulation result indicates the TDW model is well performed on the trust fraud problem and trust expiration problem. Conclusions. From the research results that we obtained from this study, we have identified several basic elements of UBT evaluation method, evaluated the current state-of-art UBT based access control models. Towards the weaknesses of trust fraud prevention and trust expiration problem, this paper designed a TDW based access control model. In comparing to the current state-of-art of UBT models, the TDW model has the following advantages, such as it is effectively preventing trust fraud problem with “slow rise” principle, able to timely response to malicious behavior by constantly aggravate punishment strategy (“rapid decrease” principle), effectively prevent malicious behavior and malicious user, and able to reflect the recent credibility of accessing user by expired trust update strategy and most recent trust calculation; finally, it has simple and customizable data structure, simple trust evaluation method, which has good scalability.

Online Social Networks (OSNs) have seen an exponential growth over the last decade, with Facebook having more than 1.49 billion monthly active users and Twitter having 135,000 new users signing up every day as of 2015. Users are sharing 70 million photos per day on the Instagram photo-sharing network. Yahoo Answers question-answering community has more than 1 billion posted answers. The meteoric rise in popularity has made OSNs important social platforms for computer-mediated communications and embedded themselves into society’s daily life, with direct consequences to the offline world and activities. OSNs are built on a foundation of trust, where users connect to other users with common interests or overlapping personal trajectories. They leverage real-world social relationships and/or common preferences, and enable users to communicate online by providing them with a variety of interaction mechanisms. This dissertation studies abuse and privacy in online social networks. More specifically, we look at two issues: (1) the content abusers in the community question answering (CQA) social network and, (2) the privacy risks that comes from the default permissive privacy settings of the OSNs. Abusive users have negative consequences for the community and its users, as they decrease the community’s cohesion, performance, and participation. We investigate the reporting of 10 million editorially curated abuse reports from 1.5 million users in Yahoo Answers, one of the oldest, largest, and most popular CQA platforms. We characterize the contribution and position of the content abusers in Yahoo Answers social networks. Based on our empirical observations, we build machine learning models to predict such users. Users not only face the risk of exposing themselves to abusive users or content, but also face leakage risks of their personal information due to weak and permissive default privacy policies. We study the relationship between users’ privacy concerns and their engagement in Yahoo Answers social networks. We find privacy-concerned users have higher qualitative and quantitative contributions, show higher retention, report more abuses, have higher perception on answer quality and have larger social circles. Next, we look at users’ privacy concerns, abusive behavior, and engagement through the lenses of national cultures and discover cross-cultural variations in CQA social networks. However, our study in Yahoo Answers reveals that the majority of users (about 87%) do not change the default privacy policies. Moreover, we find a similar story in a different type of social network (blogging): 92% bloggers’ do not change their default privacy settings. These results on default privacy are consistent with general-purpose social networks (such as Facebook) and warn about the importance of user-protecting default privacy settings. We model and implement default privacy as contextual integrity in OSNs. We present a privacy framework, Aegis, and provide a reference implementation. Aegis models expected privacy as contextual integrity using semantic web tools and focuses on defining default privacy policies. Finally, this dissertation presents a comprehensive overview of the privacy and security attacks in the online social networks projecting them in two directions: attacks that exploit users’ personal information and declared social relationships for unintended purposes; and attacks that are aimed at the OSN service provider itself, by threatening its core business.

Online communities provide a fertile ground for analyzing people\'s behavior and improving our understanding of social processes. For instance, when modeling social interaction online, it is important to understand when people are reacting to each other. Also, since both people and communities change over time, we argue that analyses of online communities that take time into account will lead to deeper and more accurate results. In many cases, however, users behavior can be easily missed: users react to content in many more ways than observed by explicit indicators (such as likes on Facebook or replies on Twitter) and poorly aggregated temporal data might hide, misrepresent and even lead to wrong conclusions about how users are evolving. In order to address the problem of detecting non-explicit responses, we present a new approach that uses tf-idf similarity between a user\'s own tweets and recent tweets by people they follow. Based on a month\'s worth of posting data from 449 ego networks in Twitter, this method demonstrates that it is likely that at least 11% of reactions are not captured by the explicit reply and retweet mechanisms. Further, these uncaptured reactions are not evenly distributed between users: some users, who create replies and retweets without using the official interface mechanisms, are much more responsive to followees than they appear. This suggests that detecting non-explicit responses is an important consideration in mitigating biases and building more accurate models when using these markers to study social interaction and information diffusion. We also address the problem of users evolution in Reddit based on comment and submission data from 2007 to 2014. Even using one of the simplest temporal differences between usersyearly cohortswe find wide differences in people\'s behavior, including comment activity, effort, and survival. Furthermore, not accounting for time can lead us to misinterpret important phenomena. For instance, we observe that average comment length decreases over any fixed period of time, but comment length in each cohort of users steadily increases during the same period after an abrupt initial drop, an example of Simpson\'s Paradox. Dividing cohorts into sub-cohorts based on the survival time in the community provides further insights; in particular, longer-lived users start at a higher activity level and make more and shorter comments than those who leave earlier. These findings both give more insight into user evolution in Reddit in particular, and raise a number of interesting questions around studying online behavior going forward. / Comunidades online proporcionam um ambiente fértil para análise do comportamento de indivíduos e processos sociais. Por exemplo, ao modelarmos interações sociais online, é importante compreendemos quando indivíduos estão reagindo a outros indivíduos. Além disso, pessoas e comunidades mudam com o passar do tempo, e levar em consideração sua evolução temporal nos leva a resultados mais precisos. Entretanto, em muitos casos, o comportamento dos usuários pode ser perdido: suas reações ao conteúdo ao qual são expostos não são capturadas por indicadores explícitos (likes no Facebook, replies no Twitter). Agregações temporais de dados pouco criteriosas podem ocultar, enviesar ou até levar a conclusões equivocadas sobre como usuários evoluem. Apresentamos uma nova abordagem para o problema de detectar respostas não-explicitas que utiliza similaridade tf-idf entre tweets de um usuário e tweets recentes que este usuário recebeu de quem segue. Com base em dados de postagens de um mês para 449 redes egocêntricas do Twitter, este método evidencia que temos um volume de ao menos 11% de reações não capturadas pelos mecanismos explicitos de reply e retweet. Além disso, essas reações não capturadas não estão uniformemente distribuídas entre os usuários: alguns usuários que criam replies e retweets sem utilizar os mecanismos formais da interface são muito mais responsivos a quem eles seguem do que aparentam. Isso sugere que detectar respostas não-explicitas é importante para mitigar viéses e construir modelos mais precisos a fim de estudar interações sociais e difusão de informação. Abordamos o problema de evolução de usuários no Reddit com base em dados entre o período de 2007 a 2014. Utilizando métodos simples de diferenciação temporal dos usuários -- cohorts anuais -- encontramos amplas diferenças entre o comportamento, que incluem criação de comentários, métricas de esforço e sobrevivência. Desconsiderar a evolução temporal pode levar a equívocos a respeito de fenômenos importantes. Por exemplo, o tamanho médio dos comentários na rede decresce ao longo de qualquer intervalo de tempo, mas este tamanho é crescente em cada uma das cohorts de usuários no mesmo período, salvo de uma queda inicial. Esta é uma observação do Paradoxo de Simpson. Dividir as cohorts de usuários em sub-cohorts baseadas em anos de sobrevivência na rede nos fornece uma perspectiva melhor; usuários que sobrevivem por mais tempo apresentam um maior nível de atividade inicial, com comentários mais curtos do que aqueles que sobrevivem menos. Com isto, compreendemos melhor como usuários evoluem no Reddit e levantamos uma série de questões a respeito de futuros desdobramentos do estudo de comportamento online.

There are a variety of communication mediums or devices for interaction. Users hop from one medium to another frequently. Though the increase in the number of devices brings convenience, it also raises security concerns. Provision of platform to users is as much important as its security. In this dissertation we propose a security approach that captures user behavior for identifying malicious activities. System users exhibit certain behavioral patterns while utilizing the resources. User behaviors such as device location, accessing certain files in a server, using a designated or specific user account etc. If this behavior is captured and compared with normal users’ behavior, anomalies can be detected. In our model, we have identified malicious users and have assigned trust value to each user accessing the system. When a user accesses new files on the servers that have not been previously accessed, accessing multiple accounts from the same device etc., these users are considered suspicious. If this behavior continues, they are categorized as ingenuine. A trust value is assigned to users. This value determines the trustworthiness of a user. Genuine users get higher trust value and ingenuine users get a lower trust value. The range of trust value varies from zero to one, with one being the highest trustworthiness and zero being the lowest. In our model, we have sixteen different features to track user behavior. These features evaluate users’ activities. From the time users’ log in to the system till they log out, users are monitored based on these sixteen features. These features determine whether the user is malicious. For instance, features such as accessing too many accounts, using proxy servers, too many incorrect logins attribute to suspicious activity. Higher the number of these features, more suspicious is the user. More such additional features contribute to lower trust value. Identifying malicious users could prevent and/or mitigate the attacks. This will enable in taking timely action against these users from performing any unauthorized or illegal actions. This could prevent insider and masquerade attacks. This application could be utilized in mobile, cloud and pervasive computing platforms.

Digital footprint designates data left behind a user movement in the digital environment or with the interaction with it. We distinguish active and passive digital footprint, together creating the digital identity of a user. The biggest benefit of the digital footprint for a user is the personalization of internet content. It also creates the reputation of a user on the internet. Other parties can utilize a digital footprint for purposes of marketing, science, HR research, and criminology. Digitial footprint can be actively affected and limited by users and different strategies for managing it exists. The aim of this thesis is to analyze user behavior of students in digital environment and their knowledge of digital footprint. It's focused on the relation between users knowledge about digital footprint existence and his real behavior. To fulfill the main purpose of the thesis researched questions are stated and the strategies of user behavior and managing their data are investigated afterwards. The knowledge of students and their real behavior with an accent on their motivation and reasons are determined in semi structured interviews. The relation between knowledge and real behavior has been found, as well as a strong influence of this relation on a user behavior strategy online.