Thesis (M.S.C.I.T.)--Regis University, Denver, Colo., 2005. / Title from PDF title page (viewed Nov. 23, 2005). Includes bibliographical references.

Orientadores: Akebo Yamakami, Tiago Agostinho de Almeida / Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-22T13:22:48Z (GMT). No. of bitstreams: 1 Silva_RenatoMoraes_M.pdf: 4136928 bytes, checksum: 218846058592353cb167c8c2d61e1bfd (MD5) Previous issue date: 2013 / Resumo: Com o crescente aumento do volume de informações disponíveis na Web, as ferramentas de busca tornam-se cada vez mais importantes para os usuários da Internet. Consequentemente, com o objetivo de se tornar mais visíveis, os sites concorrem entre si para ganhar melhores posições nos resultados das buscas feitas por esses usuários. Porém, muitos ganham maior visibilidade através de estratégias que enganam as ferramentas de busca. Esses sites, conhecidos como Web spam, causam prejuízos pessoais e econômicos aos usuários. Diante desse cenário, este trabalho apresenta uma análise do desempenho de diversos métodos de aprendizado de máquina aplicados na detecção automática de Web hosts que propagam Web spam. Os experimentos foram realizados usando duas bases de dados reais, públicas e de grande porte, das quais foram extraídos três diferentes conjuntos de vetores de atributos: baseados no conteúdo das páginas Web, baseados nos links das páginas Web e formados pela transformação dos atributos baseados nos links. Também foi analisada a viabilidade da redução de dimensionalidade do espaço dos atributos. Outra contribuição desse trabalho é a proposta de uma abordagem de classificação de Web spam, em que as predições obtidas com cada tipo de vetor de atributos são combinadas e uma decisão final é obtida usando-se voto majoritário simples. Os resultados obtidos indicam que os métodos de bagging de árvores de decisão, redes neurais perceptron de múltiplas camadas, floresta aleatória e boosting adaptativo de árvores de decisão são promissores na tarefa de detecção de Web spam. Além disso, verificou-se que os métodos de aprendizado tem melhor desempenho quando os vetores de atributos baseados no conteúdo e os vetores formados pela transformação dos atributos baseados nos links são combinados. Por fim, a combinação das predições obtidas com cada tipo de vetor de atributos gera bons resultados e por isso, essa é uma abordagem recomendada para o combate de Web spamming / Abstract: Due to the increasing volume of information available on the Web, search engines become increasingly important to Internet users. Consequently, with the purpose of becoming more visible, the Web sites compete to achieve better positions in the results of the searches made by such users. However, many of them achieve a good visibility through strategies that try to circumvent the search engines. This kind of Web sites are known as Web spam and they are responsible for personal injury and economic losses to users. Given this scenario, this work presents a performance analysis of established machine learning techniques employed to automatically detect Web hosts that disseminate Web spam. The experiments were performed with two real, public and large datasets, from which were extracted three different sets of features vectors: contentbased ones, link-based ones and features vectors generated by the transformation of the link-based features. We also analyzed the viability of the dimensionality reduction of the feature space. Another contribution of this work is the proposal of a Web spam classification approach which combines the predictions achieved by each type of features vector and using a simple majority voting. The results indicate that bagging of decision trees, multilayer perceptron neural networks, random forest and adaptive boosting of decision trees are promising in the task of spam hosts classification. Furthermore, we have conclude that the learning techniques perform better when we have combined the content-based features vectors and the features vectors generated by the transformation of the link-based features. Finally, the combination of the predictions achieved with each type of features vector has achieved superior results and therefore it is a recommended approach to automatically detect Web spam / Mestrado / Automação / Mestre em Engenharia Elétrica

The classifier discussed in this thesis considers the path traversed by an email (instead of its content) and reputation of the relays, features inaccessible to spammers. Groups of spammers and individual behaviors of a spammer in a given domain were analyzed to yield association patterns, which were then used to identify similar spammers. Unsolicited and phishing emails were successfully isolated from legitimate emails, using analysis results. Spammers and phishers are also categorized into serial spammers/phishers, recent spammers/phishers, prospective spammers/phishers, and suspects. Legitimate emails and trusted domains are classified into socially close (family members, friends), socially distinct (strangers etc), and opt-outs (resolved false positives and false negatives). Overall this classifier resulted in far less false positives when compared to current filters like SpamAssassin, achieving a 98.65% precision, which is well comparable to the precisions achieved by SPF, DNSRBL blacklists.

With the amount of email spam received these days it is hard to imagine that spammers act individually. Nowadays, most of the spam emails have been sent from a collection of compromised machines controlled by some spammers. These compromised computers are often called bots, using which the spammers can send massive volume of spam within a short period of time. The motivation of this work is to understand and analyze the behavior of spammers through a large collection of spam mails. My research examined a the data set collected over a 2.5-year period and developed an algorithm which would give the botnet features and then classify them into various groups. Principal component analysis was used to study the association patterns of group of spammers and the individual behavior of a spammer in a given domain. This is based on the features which capture maximum variance of information we have clustered. Presence information is a growing tool towards more efficient communication and providing new services and features within a business setting and much more. The main contribution in my thesis is to propose the willingness estimator that can estimate the callee's willingness without his/her involvement, the model estimates willingness level based on call history. Finally, the accuracy of the proposed willingness estimator is validated with the actual call logs.

The purpose of this project is to design, build and implement a secure, payment-based email delivery system that provides privacy and reduces spam.

Orientador: Akedo Yamakami / Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Elétrica e de Computação / Made available in DSpace on 2018-08-16T13:44:58Z (GMT). No. of bitstreams: 1 Almeida_TiagoAgostinhode_D.pdf: 1582584 bytes, checksum: 8a444adaf46219a5200a75deb26be781 (MD5) Previous issue date: 2010 / Resumo: Nos últimos anos, spams têm se tornado um importante problema com enorme impacto na sociedade. A filtragem automática de tais mensagens impõem um desafio especial em categorização de textos, no qual a característica mais marcante é que os filtros enfrentam um adversário ativo, que constantemente procura evadir as técnicas de filtragem. Esta tese apresenta um estudo abrangente sobre o problema do spamming. Dentre as contribuições oferecidas, destacam-se: o levantamento histórico e estatístico do fenômeno do spamming e as suas consequências, o estudo sobre a legalidade do spam e os recursos jurídicos adotados por alguns países, a análise de medidas de desempenho utilizadas na avaliação dos filtros de spams, o estudo dos métodos mais empregados para realizar a filtragem de spams, a proposta de melhorias dos filtros Bayesianos através da adoção de técnicas de redução de dimensionalidade e, principalmente, a proposta de um novo método de classificação baseado no princípio da descrição mais simples auxiliado por fatores de confidência. Vários experimentos são apresentados e os resultados indicam que a técnica proposta 'e superior aos melhores filtros anti-spams presentes tanto comercialmente quanto na literatura. / Abstract: Spam has become an increasingly important problem with a big economic impact in society. Spam filtering poses a special problem in text categorization, in which the defining characteristic is that filters face an active adversary, which constantly attempts to evade filtering. In this thesis, we present a comprehensive study of the spamming problem. Among many offered contributions we present: the statistical and historical survey of spamming and its consequences, a study regarding the legality of spams and the main juridic methods adopted by some countries, the study and proposal of new performance measures used for the evaluation of the spam classifiers, the proposals for improving the accuracy of Naive Bayes filters by using dimensionality reduction techniques and a novel approach to spam filtering based on the minimum description length principle and confidence factors. Furthermore, we have conducted an empirical experiments which indicate that the proposed classifier outperforms the state-of-the-art spam filters. / Doutorado / Automação / Doutor em Engenharia Elétrica

In January 2004, the United States Congress passed and put into effect the Controlling the Assault of Non-Solicited Pornography and Marketing Act (CAN SPAM). The Act was set forth to regulate bulk commercial email (spam) and set the limits for what was acceptable. Various sources have since investigated and speculated on the efficacy of the CAN SPAM Act, few of which report a desirable outcome for users of electronic mail. Despite the apparent consensus of anti-spam firms and the community of email users that the Act was less than effective, there is little to no research on the efficacy of the Act that utilizes any significant statistical rigor or accepted scientific practices. The present study seeks to determine what, if any, impact the CAN SPAM act had on spam messages, to identify areas of improvement to help fight spam that is both fraudulent and dangerous. The data consisted of 2,071,965 spam emails sent between February 1, 1998 and December 31, 2008. The data were aggregated by month and an interrupted time series design was chosen to assess the impact the CAN SPAM Act had on spam. Analyses revealed that the CAN SPAM Act had no observable impact on the amount of spam sent and received; no impact on two of three CAN SPAM laws complied with among spam emails, the remaining law of which there was a significant decrease in compliance after the Act; and no impact on the number of spam emails sent from within the United States. Implications of these findings and suggestions for policy are discussed.

Spammers are continually looking to circumvent counter-measures seeking to slow them down. An immense amount of time and money is currently devoted to hiding spam, but not enough is devoted to effectively preventing it. One approach for preventing spam is to force the spammer's machine to solve a computational problem of varying difficulty before granting access. The idea is that suspicious or problematic requests are given difficult problems to solve while legitimate requests are allowed through with minimal computation. Unfortunately, most systems that employ this model waste the computing resources being used, as they are directed towards solving cryptographic problems that provide no societal benefit. While systems such as reCAPTCHA and FoldIt have allowed users to contribute solutions to useful problems interactively, an analogous solution for non-interactive proof-of-work does not exist. Towards this end, this paper describes MetaCAPTCHA and reBOINC, an infrastructure for supporting useful proof-of-work that is integrated into a web spam throttling service. The infrastructure dynamically issues CAPTCHAs and proof-of-work puzzles while ensuring that malicious users solve challenging puzzles. Additionally, it provides a framework that enables the computational resources of spammers to be redirected towards meaningful research. To validate the efficacy of our approach, prototype implementations based on OpenCV and BOINC are described that demonstrate the ability to harvest spammer's resources for beneficial purposes.

Universiẗat, Diss.--Freiburg. i. Br., 2005. / Literaturverz. S. 274 - 285.

Thesis (doctoral)--Albert-Ludwigs-Universität Freiburg im Breisgau, 2007. / Includes bibliographical references (p. 271-284).