The spread of spam mails have become a serious threat in the Internet. In addition to commercial messages, some malicious messages such as phishing, pornography messages, fraudulent messages and malicious codes are spread via spam. A practical server-side anti-spam system should have ability to (1) filter out growing volume of spam mails correctly; (2) recognize new type of spam mails and (3) manage the increasing spam rules automatically. Most work only focused on single aspect (especially for spam rule generation) to prevent spam mail. However, in real world, spam prevention is not just applying data mining algorithm for rule generation. To filter out spam mails correctly and efficiently in a real world, there are still many issues should be considered in addition to spam rule generation. In this research, we propose and integrate three sub-systems to form a practical anti-spam system, the sub-systems are spam rule generation sub-system, spam rule sharing sub-system and spam rule management sub-system. In this research, rule-based data mining approach is used to generate manageable and shareable spam rules. The latest spam rules are shared through machine-readable XML format. Spam rules stored in mail servers are managed based on statistical testing approach. The Rule management sub-system can automatically enable high performance rules and disable out-of-date rules to improve the miss rate and efficiency of spam filter. This research will develop and integrate the three sub-systems to achieve the goal of spam prevention.

Pseudorandom generators belong to the primary focus of cryptology. The key to every cipher has to be generated at random, otherwise the security of the whole cipher is threatened. Another point of importance is the pseudorandom generators' close relationship to the stream ciphers. In this work, we first introduce statistical theory related to randomness testing. Then, we describe 8 classical statistical tests. We introduce a concept of next bit testing and derive variants of previous tests. Moreover, with this new battery of tests we examine the randomness of SHA-3 second round candidates and present the results. Also a sensitivity of tests is investigated and several useful transformations are shown. Powered by TCPDF (www.tcpdf.org)

Submitted by Daniella Santos (daniella.santos@fgv.br) on 2009-08-07T12:21:28Z No. of bitstreams: 1 Dissertação_João_Marco.pdf: 182972 bytes, checksum: 496cda0ecb8e5b1f2b520e21b3b169cf (MD5) / Approved for entry into archive by Antoanne Pontes(antoanne.pontes@fgv.br) on 2009-08-07T17:33:07Z (GMT) No. of bitstreams: 1 Dissertação_João_Marco.pdf: 182972 bytes, checksum: 496cda0ecb8e5b1f2b520e21b3b169cf (MD5) / Made available in DSpace on 2009-08-07T17:33:07Z (GMT). No. of bitstreams: 1 Dissertação_João_Marco.pdf: 182972 bytes, checksum: 496cda0ecb8e5b1f2b520e21b3b169cf (MD5) / The existence and the sign of the volatility premium has been causing controversies in the specialized literature. This work proposed, criticized and applied a novel methodology, aiming to test statistically the existence of a premium for volatility, with the advantages of testing for a set of equities jointly, not for individual series, and independent of any specific functional form for the relationship between the expected return and volatility. The results obtained on the application with a set of selected equities from Bovespa were favorable to the existence of the premium. / A existência e o sinal do prêmio de volatilidade têm causado controvérsias dentro da literatura especializada. Este trabalho propôs, criticou e aplicou uma nova metodologia com a natalidade de testar estatisticamente a existência do prêmio de volatilidade, com as vantagens de testar para um conjunto de ações, e não para séries individuais, e de não depender de uma forma funcional específica para e relação entre o retorno e a volatilidade esperados. Os resultados da aplicação para um conjunto selecionado de ações negociadas na Bovespa foram favoráveis à existência do prêmio.

In this thesis an algorithm (MLOPT) for mixed logit models is proposed. Mixed logit models are flexible discrete choice models, but their estimation with large datasets involves the solution of a nonlinear optimization problem with a high dimensional integral in the objective function, which is the log-likelihood function. This complex structure is a general problem that occurs in statistics and optimization. MLOPT uses sampling from the dataset of individuals to generate a data sample. In addition to this, Monte Carlo samples are used to generate an integration sample to estimate the choice probabilities. MLOPT estimates the log-likelihood function values for each individual in the dataset by controlling and adaptively changing the data sample and the size of the integration sample at each iteration. Furthermore, MLOPT incorporates statistical testing for the quality of the solution obtained within the optimization problem. MLOPT is tested with a benchmark study from the literature (AMLET) and further applied to real-life applications in the automotive industry by predicting market shares in the Low Segment of the new car market. The automotive industry is particularly interesting in that understanding the behavior of buyers and how rebates affect their preferences is very important for revenue management. Real transaction data is used to generate and test the mixed logit models developed in this study. Another new aspect of this study is that the sales transactions are differentiated with respect to the transaction type of the purchases made. These mixed logit models are used to estimate demand and analyze market share changes under different what-if scenarios. An analysis and discussion of the results obtained are also presented.

abstract: The present thesis explores how statistical methods are conceptualized, used, and interpreted in quantitative Hispanic sociolinguistics in light of the group of statistical methods espoused by Kline (2013) and named by Cumming (2012) as the “new statistics.” The new statistics, as a conceptual framework, repudiates null hypothesis statistical testing (NHST) and replaces it with the ESCI method, or Effect Sizes and Confidence Intervals, as well as meta-analytic thinking. In this thesis, a descriptive review of 44 studies found in three academic journals over the last decade (2005 – 2015), NHST was found to have a tight grip on most researchers. NHST, much discredited outside of linguistics, confused authors who conflated the theories of Fisher and Neyman-Pearson, who themselves battled acrimoniously until the end of their publishing lives. Within the studies reviewed, with exceptions, dichotomous thinking ruled the quantitative approach, and binary reporting ruled the results and discussions. In addition, this thesis revealed that sociolinguistics, at least within the studies reviewed, is not exactly a “statistical monoculture” as suspected by Gorman and Johnson (2013), rather ANOVAs have joined Goldvarb’s logistic regression in its dominance. As described insightfully by Plonsky (2015), these two methods are exposed as extensions of the dichotomous thinking that attaches itself to NHST. Further, little evidence was found that the methods of the new statistics were being implemented in a coordinated fashion, including far too few meta-analyses. As such, quantitative Hispanic sociolinguistics, and linguistics in general, were shown to be vulnerable to problems with reliable quantitative theory building. / Dissertation/Thesis / Masters Thesis Spanish 2015

Redundancy in plaintext is a fertile source of attack in any encryption system. Compression before encryption reduces the redundancy in the plaintext, but this does not make a cipher more secure. The cipher text is still susceptible to known-plaintext and chosen-plaintext attacks. The aim of homophonic coding is to convert a plaintext source into a random sequence by randomly mapping each source symbol into one of a set of homophones. Each homophone is then encoded by a source coder after which it can be encrypted with a cryptographic system. The security of homophonic coding falls into the class of unconditionally secure ciphers. The main advantage of homophonic coding over pure source coding is that it provides security both against known-plaintext and chosen-plaintext attacks, whereas source coding merely protects against a ciphertext-only attack. The aim of this dissertation is to investigate the implementation of an adaptive homophonic coder based on an arithmetic coder. This type of homophonic coding is termed universal, as it is not dependent on the source statistics. / Computer Science / M.Sc. (Computer Science)

The thesis "Comparison of knowledge of electromagnetic field theory of the laity and experts in the context of civil emergency preparedness" to put three goals: 1. The formation of the structure of an electromagnetic field for experts. 2. The reaching of the comparison of knowledge among experts and laymen. 3. Statistical processing of the results. The author has set the following hypotheses: H1. Theoretical distribution of knowledge in a sample of the general public will have a normal distribution . H2. Theoretical distribution of knowledge in a sample of professional community will not have a normal distribution. H3. The comparison of knowledge among the experts and the laymen will lead to an alternative hypothesis. The thesis was based on the knowledge of the theory curricular process. On the basis of this theory was made up not only the structure of the electromagnetic field, but also the questionnaire. An important step in this thesis was the creating a model structure of electromagnetic field . The structure was based on an analysis of the scientific system - the system of educational programs in the field of civil protection.The same structure was applied to the general public. An important step was to compare the knowledge of protect the population from experts and laymen. This issue has not been investigated in detail and it did not compare the knowledge of laymen and experts in the studied physics. The idea came from the possibility of extraordinarily events where respondents can meet with electromagnetic fields and will need the relevant theoretical knowledge. The aim was to the statistical evaluate of the applied questionnaires. There were applied nonparametric and parametric testing as the verification methods. The theoretical division of knowledge of experts is supposed Poisson distribution, on the contrary, the theoretical division of the general public should have a normal distribution. There was also compared the difference between knowledge of laymen and professionals. The using of the statistical methods have been received and confirmed the hypothesis and the thesis goals were fulfilled.

Redundancy in plaintext is a fertile source of attack in any encryption system. Compression before encryption reduces the redundancy in the plaintext, but this does not make a cipher more secure. The cipher text is still susceptible to known-plaintext and chosen-plaintext attacks. The aim of homophonic coding is to convert a plaintext source into a random sequence by randomly mapping each source symbol into one of a set of homophones. Each homophone is then encoded by a source coder after which it can be encrypted with a cryptographic system. The security of homophonic coding falls into the class of unconditionally secure ciphers. The main advantage of homophonic coding over pure source coding is that it provides security both against known-plaintext and chosen-plaintext attacks, whereas source coding merely protects against a ciphertext-only attack. The aim of this dissertation is to investigate the implementation of an adaptive homophonic coder based on an arithmetic coder. This type of homophonic coding is termed universal, as it is not dependent on the source statistics. / Computer Science / M.Sc. (Computer Science)

Sign language is used widely around the world as a first language for those that are unable to use spoken language and by groups of people that have a disability which precludes them from using spoken language (such as a hearing impairment). The importance of effective learning of sign language and its applications in modern computer science has grown widely in the modern aged society and research around sign language recognition has sprouted in many different directions, some examples using hidden markov models (HMMs) to train models to recognize different sign language patterns (Swedish sign language, American sign language, Korean sign language, German sign language and so on).  This thesis project researches the assessment and skill efficiency of using a simple video game to learn Swedish sign language for children in the ages within the range of 10 to 11 with no learning disorders, or any health disorders. During the experimental testing, 38 children are divided into two equally sized groups of 19 where each group plays a sign language video game. The context of the video game is the same for both groups, where both listened to a 3D avatar speak to them using both spoken language and sign language. The first group played the game and answered questions given to them by using sign language, whereas the other group answered questions given to them by clicking on an alternative on the video game screen. A week after the children have played the video game, the sign language skills that they have acquired from playing the video game are assessed by simple questions where they are asked to provide some of the signs that they saw during the duration of the video game. The main hypothesis of the project is that the group of children that answered by signing outperforms the other group, in both remembering the signs and executing them correctly. A statistical null hypothesis test is performed on this hypothesis, in which the main hypothesis is confirmed. Lastly, discussions for future research within sign language assessment using video games is described in the final chapter of the thesis. / Teckenspråk används i stor grad runt om i världen som ett modersmål för dom som inte kan använda vardagligt talsspråk och utav grupper av personer som har en funktionsnedsättning (t.ex. en hörselskada). Betydelsen av effektivt lärande av teckenspråk och dess tillämpningar i modern datavetenskap har ökat i stor utsträckning i det moderna samhället, och forskning kring teckenspråklig igenkänning har spirat i många olika riktningar, ett exempel är med hjälp av statistika modeller såsom dolda markovmodeller (eng. Hidden markov models) för att träna modeller för att känna igen olika teckenspråksmönster (bland dessa ingår Svenskt teckenspråk, Amerikanskt teckenspråk, Koreanskt teckenspråk, Tyskt teckenspråk med flera). Denna rapport undersöker bedömningen och skickligheten av att använda ett enkelt teckenspråksspel som har utvecklats för att lära ut enkla Svenska teckenspråksmönster för barn i åldrarna 10 till 11 års ålder som inte har några inlärningssjukdomar eller några problem med allmän hälsa. Under projektets experiment delas 38 barn upp i två lika stora grupper om 19 i vardera grupp, där varje grupp kommer att få spela ett teckenspråksspel. Sammanhanget för spelet är detsamma för båda grupperna, där de får höra och se en tredimensionell figur (eng. 3D Avatar) tala till dom med både talsspråk och teckenspråk. Den första gruppen spelar spelet och svarar på frågor som ges till dem med hjälp av teckenspråk, medan den andra gruppen svarar på frågor som ges till dem genom att klicka på ett av fem alternativ som finns på spelets skärm. En vecka efter att barnen har utfört experimentet med teckenspråksspelet bedöms deras teckenspråkliga färdigheter som de har fått från spelet genom att de ombeds återuppge några av de tecknena som de såg under spelets varaktighet. Rapportens hypotes är att de barn som tillhör gruppen som fick ge teckenspråk som svar till frågorna som ställdes överträffar den andra gruppen, genom att både komma ihåg tecknena och återuppge dom på korrekt sätt. En statistisk hypotesprövning utförs på denna hypotes, där denna i sin tur bekräftas. Slutligen beskrivs det i rapportens sista kapitel om framtida forskning inom teckenspråksbedömning med tv spel och deras effektivitet.