This report presents RACR, a reference attribute grammar library for the programming language Scheme. RACR supports incremental attribute evaluation in the presence of abstract syntax tree rewrites. It provides a set of functions that can be used to specify abstract syntax tree schemes and their attribution and construct respective trees, query their attributes and node information and annotate and rewrite them. Thereby, both, reference attribute grammars and rewriting, are seamlessly integrated, such that rewrites can reuse attributes and attribute values change depending on performed rewrites – a technique we call Reference Attribute Grammar Controlled Rewriting. To reevaluate attributes influenced by abstract syntax tree rewrites, a demand-driven, incremental evaluation strategy, which incorporates the actual execution paths selected at runtime for control-flows within attribute equations, is used. To realize this strategy, a dynamic attribute dependency graph is constructed throughout attribute evaluation – a technique we call Dynamic Attribute Dependency Analyses. The report illustrates RACR's motivation, features, instantiation and usage. In particular its application programming interface is documented and exemplified. The report is a reference manual for RACR developers. Further, it presents RACR’s complete implementation and therefore provides a good foundation for readers interested into the details of reference attribute grammar controlled rewriting and dynamic attribute dependency analyses.:1. Introduction 1.1. RACR is Expressive, Elegant, Ecient, Flexible and Reliable 1.2. Structure of the Manual 2. Library Overview 2.1. Architecture 2.2. Instantiation 2.3. API 3. Abstract Syntax Trees 3.1. Specification 3.2. Construction 3.3. Traversal 3.4. Node Information 4. Attribution 4.1. Specification 4.2. Evaluation and Querying 5. Rewriting 5.1. Primitive Rewrite Functions 5.2. Rewrite Strategies 6. AST Annotations 6.1. Attachment 6.2. Querying 7. Support API A. RACR Source Code B. MIT License API Index

在住宅市場中，「空屋率」一直是備受注目的焦點，在高空屋率的住宅市場之下，其不但對整體社會資源造成浪費外，亦使整體住宅市場機制之扭曲，故探討高空屋率之成因是非常重要的。本文以2000年台灣359區市鄉鎮空屋率來觀察高空屋率之成因，並分析都市發展程度與住宅存量屬性對空屋率之影響。 研究發現，在區市鄉鎮空屋率之成因分析中，透過實證模型檢驗出不同都市發展程度、人口淨遷徙率及住宅存量屬性對空屋率皆有顯著之影響。都市發展處於鄉鎮階段時，住宅存量屬性的變異對於空屋率的影響程度相當高，不論是小坪數、大坪數或高樓層住宅比例的增加，皆會提高空屋率的比率。反之，對於都市發展處於快速成長階段時，住宅空屋率會比鄉鎮地區來得高，但有趣的是，住宅存量變異對空屋率的影響反而是呈現下降的情形。至於當都市發展至成熟階段時，除了小坪數住宅比例外，其餘變數都不會顯著影響到空屋率的增減，而呈現相對穩定的供需狀態。 空屋率政策應該配合都市發展的階段性，以及住宅產品變異的交互影響效果，才能因地制宜解決台灣高空屋率現象或降低投資風險。例如處於都市發展處於鄉鎮階段時，住宅產品應朝向同質與典型的產品規劃，但是當都市快速發展時，適當地放寬高樓層與大面積住宅產品反而可以降低空屋現象。

本研究以民國86年至90年之54檔投資台灣開放式一般股票型之共同基金為研究對象，探討財務資訊與股票型基金持股間之關聯。主要實證議題包括：（一）在不同的產業之下，共同基金經理人從事持股決策所重視的財務屬性是否有差異及（二）公司之財務表現與共同基金對該公司平均持股比例之關聯性。 本研究分別針對建築、食品、塑化、紡織、電機、資訊電子及服務銷售類等七個產業，首先利用因素分析將16項財務比率萃取出數個具代表性之財務屬性，並加入股票週轉率、淨值市價比、本益比與每股股價等控制變數進行迴歸分析，以探討財務屬性與共同基金平均持股比例之關聯性。 實證結果發現，償債能力、經營能力、投資報酬、經營績效、固定資產資金之安全與效率和現金支應能力六個財務屬性最能代表建築產業之財務表現；償債能力、經營能力、應收帳款週轉率和獲利能力四個財務屬性最能代表食品產業之財務表現；償債能力、利息保障倍數、經營能力、和獲利能力四個財務屬性最能代表塑化產業之財務表現;償債能力、利息保障倍數、經營能力、獲利能力和財務結構五個財務屬性最能代表紡織產業之財務表現;償債能力、資金適足率、經營能力、獲利能力和財務結構五個財務屬性最能代表電機產業之財務表現；償債能力、經營能力和獲利能力三個財務屬性最能代表資訊電子產業之財務表現；而償債能力、經營能力、獲利能力、負債比率和固定資產資金之安全與效率五個財務屬性最能代表服務銷售產業之財務表現。各個產業所萃取出之財務屬性因素不盡相同，顯示彙總公司財務狀況及經營績效的財務屬性確有產業別之差異存在。在所探討的財務屬性構面中，償債能力、經營能力、獲利能力為各產業共同強調之財務屬性。 就基金經理人所做的持股策略來看，不論針對何種產業，償債能力、經營能力及獲利能力均是基金經理人最為重視的財務屬性指標。研究結果另外也顯示，除了服務銷售業、食品業與建築業外，其他產業之償債能力屬性表現越好，基金持股比例愈高。塑化業、紡織業、電機業、資訊電子業和服務銷售業之經營能力屬性表現愈好，基金持股比例也相對愈高。至於獲利能力屬性則在紡織業、電機業、資訊電子業和服務銷售業顯著地對基金持股比例有正面的影響。 本研究之發現顯示，不同的產業在不同的經營環境下，反映公司狀況的財務資訊會因產業別而有所不同，即使在相同的財務比率下，彙總公司財務狀況及經營結果之財務因素屬性仍有所差異，並進而影響基金經理人的持股決策，尤其是償債能力、經營能力和獲利能力，更是投資各產業時廣泛參考的基本財務屬性指標。 關鍵字：股票型基金、財務屬性、基金持股比例 / The primary object of this research is to exam the relationship between financial information of a firm and its share held by mutual funds, based on 54 mutual funds of general stock investments in Taiwan over the period of 1997 to 2001. Two major research issues are investigated in this thesis. The study first explores whether the managers of mutual funds emphasize on the same attributes abstracted from financial statements across different industries for their share-holding decisions. The study then examines the relationship between financial performance of a company and its average invested percentage by mutual funds. This thesis focuses on architecture, foods, plastic, textile, electrical machinery, information electronics, as well as service and sales industries. Based on a factor analysis approach, the thesis is able to obtain several representative dimensions of financial attributes out of 16 financial ratios. Then, the use of multiple regression models enables the study to investigate the relationship between financial attributes and average shareholding percentage of the mutual funds across industries. The empirical results indicate that financial attributes extracted are not exactly the same across industries. It implies that industry differences do exist in financial attributes that summarize the financial conditions and operating performances of companies. Among others, attributes that characterize the performance of liquidity, operating efficiency, and profitability of a firm are found important for share-holding percentage across all industries examined. The research findings show that with exceptions in the service and sales, foods, and architecture industries, liquidity has found to have a significant and positive relation with the share-holding percentage of mutual funds in all industries examined. Operating efficiency has also found to be significantly and positively associated with the share-holding percentage in the plastic, textile, electrical machinery, information electronics, as well as service and sales industries. In addition, a significant and positive relation is found between profitability and the share-holding percentage in the electrical machinery, information electronics, textile, as well as service and sales industries. The finding of this study implies that difference in financial conditions of firms may result in difference in financial attributes across industries; even the analysis is based on the same financial ratios. Further, the liquidity, operating efficiency and profitability are the most essential indicators shared by mutual funds managers in making investment decisions. Keywords: stock-type mutual fund, financial attribute, share-holding ratio

廣告的目的，是用來與「目標消費者」進行訊息溝通與傳達，而不是自我藝術美感的創作與表達。廣告的內容，是將「廣告訊息」，藉由編碼，正確的傳遞給目標消費者（閱聽人）。而有說服力的廣告，就會是可以幫助銷售的好廣告。在消費者對於廣告解讀的論述中，出現了許多影響訊息說服與溝通的閱聽人的心理意識，包括了消費者的特質、生活型態、價值觀，文化、社會認同、社會地位、群體同儕、品味、成就、以及夢想等。越能以這些因素，引起閱聽人共鳴，就能夠越容易完成訊息傳遞的目的。 而這些心理意識，就是「內團體意識」中，區分內、外團體的歸因要素。有了情感上的歸屬，訊息傳遞就可以比較容易。電視廣告在「內團體意識」的包裝下，消費者（想要）及（接受）的心理因素不斷的被提到，被群體「認同」的需求，一再的被滿足，期望一再的被塑造。隨著「內團體意識」的表現，在經歷了認知、情感、行為三個階段後，溝通目的才開始清楚的完成。 　　為瞭解「內團體意識」如何建構電視廣告訊息？有何表現上的可歸納的操作指標，本研究以內容分析法，選擇自1980年～現今2009年之間，具有「內團體意識」表現方式的電視廣告影片110支，以年代別及FCB模式商品分類別兩種方式，進行包括表達元素、表達形式、背景歸因、社會歸因、文化歸因、生活風格歸因、個體歸因等相關要素的表現分析與研究。 研究整理後發現，「內團體意識」在電視廣告訊息的表現中，以「情感認同」為核心，藉由「滿足需求」、「消費描述」及「建構想像」做為基本創意概念。另外，電視廣告片中大多必要且慣用「內團體意識」，甚至在許多的電視廣告中，會有一個或多個內團體意識的歸因元素同時存在。部份內團體意識，只是創意與橋段的表現，與原本要訴求對象的「團體意識」並不會衝突。最後也發現，似乎越是特殊的商品或勞務，就越需要使用「內團體意識」，來做為電視廣告中重要溝通核心。 關鍵字：內團體意識、廣告表現、團體認同、廣告訴求、團體偏私、核心概念、訊息解讀、情感認同、歸因要素、背景歸因、社會歸因、文化歸因、生活型態、個體歸因、廣告創意、訊息策略、說服理論。 / The intention of advertisement is to provide a message to the targeted consumer and communicate rather than a presentation of individual expression and creativity. The core meaning of an advertisement is to transfer a key message embedded with advertising means successfully to its target receiver; furthermore, an outstanding advertisement will be able to fulfill the intention of marketing and sales. An exposition regarding consumer advertisement decoding denoted numerous variables that potentially alter the protocol of message receiver and his way of interpreting a message. This includes an individual consumer’s characteristic, lifestyle, value, culture, social acceptance, social status, social circle, taste, accomplishment, and vision. Messages embedded with such variables tend to create sympathy between the message and the receiver; as a result, creates a psychological bonding in which grows more powerful whenever a new variable reaches its echo with the receiver. Such psychological behaviors are the essential attributes differentiating the in-group consciousness from out-group consciousness of an individual. With a sentimental bonding, messages can be transmitted more accurately and efficiently. Television advertisement exercise constant implicit of this in-group conscious behavior, by repeating intimation of “desire” and “acceptance” to the receiver, continuous fulfillment of ones the necessitate aspiration, and constant creation of expectation. As the realization of the in-group consciousness continues, the purpose of communication is only fully accomplished after an individual have gone through acknowledgement, emotion, and into action. To understand how in-group consciousness establishes and executes television ad messages and its protocols. This research suggested using content analysis; by selecting 110 Television ad clips that possesses in-group consciousness implementation in between 1980 to 2009 and analyze them in two ways–era differentiation and product type under FCB protocol. This study focused on observing and analyzing the performance of variables such as expressing element, expressing form, background, social, cultural, lifestyle, individual attributes, and other related elements. 　　After the compilation of results, this research suggests that most television ads exercise in-group consciousness intentionally to create a common attribute among its listeners. By establishing sensational acceptance as the core criteria, advertisements then elaborates on other attributes such as satisfying needs, consuming behavior, and image establishment to create the basic concept of creativity. Furthermore, this research shows that a large number of advertisements exercise one or more in-group consciousness attribute within them. Some of the in-group consciousness exercised in ads is portrayed simply due to creative and plotting needs and doesn’t actually create a conflict with the central in-group consciously. This research also found that the more unique or special a product or service is, the more dependant its advertisement will be on in-group consciousness. Keywords: in-group consciousness, advertisement performance, group acceptance, advertisement requirements, group favoritism, core concept, message interpretation, emotional acceptance, attributes factor, background attribute, social attribute, cultural attribute, lifestyle, individual attribute, advertisement creativity, message tactic, persuading theory.

The aim of this thesis is to study how words denoting age are used in newspaper texts. I have chosen mainly to study the following Swedish focus words for females and males of different ages (flicka, kvinna, tjej, dam; pojke, man, kille and herre) as they relate to age. Furthermore, age lexemes – different variations of Swedish expressions of age – are investigated and finally also a number of new Swedish expressions that are used to describe age – kids, förtis, grups, tweenie, mappie, senior and what I call “plus gradings” – e.g. 50 +. The study is particularly concerned with where and how age is referred to and how the chosen focus words relate in different ways to age and sex – and finally also with how these things have changed over time. I have examined newspaper corpora from three separate years – 1965, 1987 and 2000. My theoretical perspective is social constructivist, age being understood as something that is constructed and negotiated in language use. As for method, I adopt a corpus linguistic approach – large corpora and quantitative language patterns in frequencies and lexical content. Thus, my investigation is aimed at the focus words, the age lexemes and the terms in their linguistic contexts, primarily as regards adjective attributes, lexical relations and collocations. My conclusion is that the examined expressions occur in different contexts and are used in different ways, but are not always related to age or age differentiation. I have noticed a couple of tendencies; one relates to vagueness when it comes to age and one relates to a semantic division of labour. The focus words are both used as synonyms and as oppositions, and they refer to referents belonging to different age groups. At the same time there is a clear division of labour between them in that variation in many contexts relates more to subject and contextual pattern than to different chronological ages. Ihave also noticed that chronological age is generally used when something is unusual, deviant or when the context has to do with crime, sports, (change of or debut in) a profession, birthday announcements etc. / Den här avhandlingen undersöker hur ålder används språkligt i tidningstext. Utgångspunkten är hur ålder har förändrats. Framförallt är det fokusorden flicka, pojke, man, kvinna, kille, tjej, dam och herre som i relation till ålder undersöks i avhandlingen. Därutöver studeras ålderslexem - olika variationer av de språkliga uttrycken ålder och år och slutligen även ett antal nya begrepp som förekommer i relation till ålder – kids, förtis, grups, tweenie, mappie, senior och det jag kallar för plusgraderingar – t ex 50 +. Frågor jag ställer mig har med var och hur ålder förekommer och hur de utvalda fokusorden relateras på olika sätt till ålder och till kön – och slutligen hur detta har förändrats över tid. Jag har studerat tidningskorpusar från tre olika årtal – 1965, 1987 och 2000. Teoretiskt utgår jag från ett socialkonstruktivistiskt perspektiv där ålder ses som någonting som konstrueras och förhandlas i språkanvändningen. Metodologiskt utgår jag från ett korpuslingvistiskt synsätt - stora textkorpusar och kvantitativa språkliga mönster i frekvenser och lexikalt innehåll. Jag studerar alltså fokusorden, ålderslexemen och de nya begreppen i sina språkliga kontexter, och då framförallt adjektivattribut, lexikala relationer och kollokationer. Jag kommer fram till att de olika språkliga uttrycken förekommer i olika typer av kontexter och på olika sätt, men att det långt ifrån alltid har med ålder eller med åldersdifferentiering att göra. Två olika tendenser är tydliga som har med åldersvaghet och arbetsfördelning att göra. De olika fokusorden både likställs och differentieras, och de används om referenter i olika ålderskategorier. Samtidigt är det en tydlig arbetsfördelning mellan dem där variationen i många sammanhang har mer med ämnen och innehållsliga mönster att göra, än med olika kronologiska åldrar. Jag har också kunnat konstatera att kronologisk ålder generellt används när någonting är ovanligt eller avvikande utifrån en åldersnorm eller när kontexterna har med idrott, debut eller byte av profession, födelse- och dödsannonser etc.

This study is about the nature of persons and personal identity. It belongs to a tradition that maintains that in order to understand what it is to be a person we must clarify what personal identity consists in. In this pursuit, I differentiate between the problems (i) How do persons persist? and (ii) What facts, if any, does personal identity consist in? Concerning the first question, I argue that persons persist three-dimensionally (the endurance view), and not four-dimensionally (the perdurarne view), on the ground that objects must always fall under some substance sortal concept S (the sortal dependency of individuation), and that the concept person entails that objects falling under it are three-dimensional. Concerning the second question, I differentiate between Criterianists, who maintain that it is possible to specify a non-circular and informative criterion for personal identity, and Non-Criterianists, who deny that such a specification is possible. I argue against Criterianist accounts of personal identity on the ground that they are either (i) circular, (ii) violate the intrinsicality of identity or (iii) do not adequately represent what we are essentially. I further criticise three Psychological Non-Criterianist accounts of personal identity on the ground that they wrongly assume that 'person' refers to mental entities. Instead I formulate the Revised Animal Attribute View where person is understood as a basic sortal concept which picks out a biological sort of enduring animals. In this, I claim that the real essence of a person is determined by the real essence of the kind of animal he is, without thereby denying that persons have a real essence as persons. / digitalisering@umu

Ces dernières années, la conception de produit vise de plus en plus à remettre le client aucentre du processus de développement. De nouvelles méthodes et outils ont permis de formaliser nonseulement l’identification des besoins et des attentes du client mais également leur transformation enexigences. Cependant, malgré les progrès récemment apparus dans ce domaine, la notion de valeurperçue par le client qui est associée au produit reste faiblement considérée durant le développement duproduit alors que la perception de cette valeur par le client va jouer un rôle clé au moment du choix duproduit.Dans cette thèse, nous proposons une approche normative visant l’intégration de la notion devaleur du produit dans le processus d’établissement des exigences. Un état de l’art présente lesconcepts et les pratiques impliqués dans le déploiement de la fonction de qualité, étroitement liée auprocessus de spécification des exigences basé sur la valeur. Ces concepts sont relatifs aux notionsd’objectifs, de modèles de la valeur, de fonctions d’utilité multi attributs et de hiérarchie et de réseauxd’objectifs.Notre approche se déroule en 4 étapes : (1) identifier et structurer les objectifs à partir del’expression des attentes du client, (2) spécifier les attributs de la valeur perçue par le client etconstruire un modèle de la valeur client, (3) transformer les objectifs en exigences pour construire unmodèle global de la valeur produit et (4) dériver ce modèle global en un modèle de la valeur pourchaque composant du produit. Cette approche permet ainsi de prendre en compte explicitement lavaleur produit perçue par le client en l’intégrant dans les phases de développement, ce qui favorise uneconception proactive dirigée par la valeur.Nous l’illustrons sur un exemple d’avion de ligne développé en Vanguard Studio. Cet exemplea été élaboré par l’ensemble des partenaires au cours du projet Européen IST CRESCENDO et sert decas-test pour ce projet / Recently customer-based product development is becoming a popular paradigm. Customer needsand expectations are identified and transformed into requirements in systematic processes for productdesign with the help of various methods and tools. However, in many cases, these approaches fail tofocus on the perceived value that is crucially important when customers make the decision ofpurchasing a product. The requirements specification derived from these approaches are typicallyvalue-implicit.In this thesis, a prescriptive approach to derive value-based requirements specification is proposedby integrating the concept of value into the house of quality of quality function deployment. Anintegrated set of theories, methods and concepts is introduced in order to mitigate the sevenmethodological problems of house of quality regarding to establishing appropriate value-basedrequirements specification. The foundations of the approach include concepts of objective, valuemodel and consequence model, methods of means-ends objectives network, fundamental objectiveshierarchy, response surface methodology and value-driven design, and theory of multi-attribute utilitytheory. The procedure of the approach is a four-step process: (1) identify and structure objectives frominitial customer statements of expectations, (2) specify attributes and construct customer value model,(3) transform fundamental objectives into engineering characteristics to construct system value model,and (4) derive component value models from system value model. Through this procedure, initialcustomer statements can be reasonably derived into customer value model, system value model andcomponent value model. The benefits of the approach are that it enables (1) reasonably qualifying andquantifying customer value, and performing value modeling and simulation, (2) perceived customervalue being subsequently used reactively for design evaluation, and proactively for value-drivendesign.The approach is applied in the context of a European Community’s R&D project CRESCENDO tohelp constructing airlines’ group value model for commercial aircraft development. This applicationfocuses on the first two-steps of the approach, and the value models are implemented in businesssoftware Vanguard Studio

[pt] A melhoria nas técnicas de obtenção de dados sísmicos fez com que estes dados sejam da ordem de até terabytes, o que gera um grande problema no armazenamento e transmissão dos mesmos. No que se refere ao problema de armazenamento em disco, apesar da queda no preço do disco rígido, o avanço das técnicas de obtenção do dado sísmico e o barateamento destas técnicas de obtenção do dado fazem com que a demanda por espaço em disco cresça mais rápido do que a queda no preço do hardware. Outro fator importante é a limitação de espaço físico para armazenar todos os discos rígidos de forma a garantir a segurança da informação destes discos. Este trabalho tem como objetivo apresentar uma nova técnica de compactação para dados sísmicos utilizando a transformada wavelet 3D. É feito um comparativo com as técnicas de compressão atuais, que utilizam a transformada 2D. / [en] Improvements made in techniques for seismic data acquisition have led to data that are of the order of terabytes, which in turn creates the problem of data storage and transmission of them through the network. In regards to the problem of disk storage, despite the hard disk decreasing price, the improvement and reduction cost in seismic data acquisition techniques has pushed the demand for disk space to grow faster than the reduction of the hardware cost. Another important factor is the space necessary to store all hard drives, to provide security for information from these records. In order to tackle these problems this thesis presents a new technique for seismic data compression using 3D wavelet transform. It is also presented in comparison with current compression techniques which use the 2D transform.

Given a very large dataset of moderate-to-high dimensionality, how to mine useful patterns from it? In such cases, dimensionality reduction is essential to overcome the well-known curse of dimensionality. Although there exist algorithms to reduce the dimensionality of Big Data, unfortunately, they all fail to identify/eliminate non-linear correlations that may occur between the attributes. This MSc work tackles the problem by exploring concepts of the Fractal Theory and massive parallel processing to present Curl-Remover, a novel dimensionality reduction technique for very large datasets. Our contributions are: (a) Curl-Remover eliminates linear and non-linear attribute correlations as well as irrelevant attributes; (b) it is unsupervised and suits for analytical tasks in general not only classification; (c) it presents linear scale-up on both the data size and the number of machines used; (d) it does not require the user to guess the number of attributes to be removed, and; (e) it preserves the attributes semantics by performing feature selection, not feature extraction. We executed experiments on synthetic and real data spanning up to 1.1 billion points, and report that our proposed Curl-Remover outperformed two PCA-based algorithms from the state-of-the-art, being in average up to 8% more accurate. / Dada uma grande base de dados de dimensionalidade moderada a alta, como identificar padrões úteis nos objetos de dados? Nesses casos, a redução de dimensionalidade é essencial para superar um fenômeno conhecido na literatura como a maldição da alta dimensionalidade. Embora existam algoritmos capazes de reduzir a dimensionalidade de conjuntos de dados na escala de Terabytes, infelizmente, todos falham em relação à identificação/eliminação de correlações não lineares entre os atributos. Este trabalho de Mestrado trata o problema explorando conceitos da Teoria de Fractais e processamento paralelo em massa para apresentar Curl-Remover, uma nova técnica de redução de dimensionalidade bem adequada ao pré-processamento de Big Data. Suas principais contribuições são: (a) Curl-Remover elimina correlações lineares e não lineares entre atributos, bem como atributos irrelevantes; (b) não depende de supervisão do usuário e é útil para tarefas analíticas em geral não apenas para a classificação; (c) apresenta escalabilidade linear tanto em relação ao número de objetos de dados quanto ao número de máquinas utilizadas; (d) não requer que o usuário sugira um número de atributos para serem removidos, e; (e) mantêm a semântica dos atributos por ser uma técnica de seleção de atributos, não de extração de atributos. Experimentos foram executados em conjuntos de dados sintéticos e reais contendo até 1,1 bilhões de pontos, e a nova técnica Curl-Remover apresentou desempenho superior comparada a dois algoritmos do estado da arte baseados em PCA, obtendo em média até 8% a mais em acurácia de resultados.

Cette thèse vise à comprendre le risque de fuite d’informations personnelles sur un réseau social. Nous étudions les violations potentielles de la vie privée, concevons des attaques, prouvons leur faisabilité et analysons leur précision. Cette approche nous aide à identifier l’origine des menaces et constitue un premier pas vers la conception de contre-mesures efficaces. Nous avons d’abord introduit une mesure de sensibilité des sujets à travers une enquête par questionnaire. Puis, nous avons conçu des attaques de divulgation (avec certitude) des liens d’amitié et des liens d’appartenance aux groupes sur “Facebook”. Ces attaques permettent de découvrir le réseau local d’une cible en utilisant uniquement des requêtes légitimes. Nous avons également conçu une technique d’échantillonnage pour collecter rapidement des données utiles autour d’une cible. Les données collectées sont ensuite représentées par des graphes et utilisées pour effectuer des inférences d’attributs (avec incertitude). Pour augmenter la précision des attaques, nous avons conçu des algorithmes de nettoyage. Ces algorithmes quantifient la corrélation entre les sujets, sélectionnent les plus pertinents et permettent de gérer la rareté (sparsity) des données. Enfin, nous avons utilisé un réseau de neurones pour classer les données et déduire les valeurs secrètes d’un attribut sensible d’une cible donnée avec une précision élevée mesurée par AUC sur des données réelles. Les algorithmes proposés dans ce travail sont inclus dans un système appelé SONSAI qui aide les utilisateurs finaux à contrôler la collecte d’informations sur leur vie privée / In this thesis we shed the light on the danger of privacy leakage on social network. We investigate privacy breaches, design attacks, show their feasibility and study their accuracies. This approach helps us to track the origin of threats and is a first step toward designing effective countermeasures. We have first introduced a subject sensitivity measure through a questionnaire survey. Then, we have designed on-line friendship and group membership link disclosure (with certainty) attacks on the largest social network “Facebook”. These attacks successfully uncover the local network of a target using only legitimate queries. We have also designed sampling techniques to rapidly collect useful data around a target. The collected data are represented by social-attribute networks and used to perform attribute inference (with uncertainty) attacks. To increase the accuracy of attacks, we have designed cleansing algorithms. These algorithms quantify the correlation between subjects, select the most relevant ones and combat data sparsity. Finally, we have used a shallow neural network to classify the data and infer the secret values of a sensitive attribute of a given target with high accuracy measured by AUC on real datasets. The proposed algorithms in this work are included in a system called SONSAI that can help end users analyzing their local network to take the hand over their privacy