Submitted by Renata Lopes (renatasil82@gmail.com) on 2017-05-30T19:30:25Z No. of bitstreams: 1 rafaelbarradealmeida.pdf: 1586266 bytes, checksum: b824fb31f42c6ab41350af41be08ad11 (MD5) / Approved for entry into archive by Adriana Oliveira (adriana.oliveira@ufjf.edu.br) on 2017-06-01T11:36:00Z (GMT) No. of bitstreams: 1 rafaelbarradealmeida.pdf: 1586266 bytes, checksum: b824fb31f42c6ab41350af41be08ad11 (MD5) / Made available in DSpace on 2017-06-01T11:36:00Z (GMT). No. of bitstreams: 1 rafaelbarradealmeida.pdf: 1586266 bytes, checksum: b824fb31f42c6ab41350af41be08ad11 (MD5) Previous issue date: 2013-01-25 / FAPEMIG - Fundação de Amparo à Pesquisa do Estado de Minas Gerais / Aplicações de transmissão de vídeo ao vivo na Internet têm ganhado bastante populari dade nos últimos anos. A facilidade para se publicar e ter acesso a esse tipo de conteúdo tem atraído grande atenção. A arquitetura P2P ganhou lugar de destaque neste contexto, principalmente, por ser mais tolerante a falhas e superar os problemas de escalabilidade presentes no modelo cliente-servidor. No entanto, devido às suas características, como ausência de controle centralizado, as redes P2P podem ser suscetíveis a ataques e com portamentos maliciosos. Este trabalho analisa o impacto causado por ataques de poluição e whitewashing em sistemas P2P de transmissão de vídeo ao vivo. Para combater estes tipos de ataques, me canismos simples e descentralizados de reputação são propostos e implementados através de um protótipo de aplicação executado em um ambiente de rede real, con gurado no PlanetLab. Os resultados mostram que ataques de poluição são bastante prejudiciais a sistemas P2P de transmissão de vídeo ao vivo. Quando apenas 10% dos peers agem de maneira maliciosa, os demais participantes do sistema recebem mais de 90% de dados poluídos. Além disso, a sobrecarga média na banda de rede chega a 230% em momentos de pico, forçando os participantes a utilizar 3 vezes mais banda de rede do que seria necessário em um sistema sem ataques de poluição. Os mecanismos de reputação propostos, testados no ambiente do PlanetLab, bloqueiam ataques de poluição rapidamente, reduzindo a porcentagem de dados poluídos recebidos a 6% e a sobrecarga a 5%. Para o caso de ataque combinado de poluição e whitewashing, o mecanismo de reputação proposto diminui a sobrecarga no sistema de 112% para 20% e a porcentagem de dados poluídos de 70% para 19%. / Live streaming video applications on the Internet are becoming more popular in the last years. The ease of publishing and accessing content through such applications has at tracted great attention from users and researchers. The P2P architecture has obtained a prominent place in this context, mainly because it is more fault tolerant and overcome the scalability issues present in the client-server model. However, due to its distributed characteristics, the P2P networks can be more susceptible to attacks and malicious be havior. In this work, we analyze the impact of pollution attacks and whitewashing in P2P live streaming systems. To combat these attacks, simple mechanisms and decentralized reputation are proposed and implemented by a prototype application running on a real network environment, set in PlanetLab. Our results show that pollution attacks are harmful to P2P live streaming systems. When 10% of peers are malicious, the remaining peers in the system receive more than 90% of data polluted. Furthermore, the average overhead in network bandwidth reaches 230% at peak times, forcing the peers to use three times the bandwidth required in a system without pollution attacks. The proposed reputation mechanism, tested on PlanetLab testbed, quickly blocks isolated pollution attacks, reducing the amount of polluted data received to 6% and the overhead to 5%. In the case of pollution attack combined with whitewashing, the proposed reputation mechanism reduces the overhead in the system from 112% to 20% and the amount of polluted data from 70% to 19%.

Whitewashing entails the practice of whitening or altering an historical or fictional characters of colours with the casting of white actors. The lack of diversity in today’s media landscape highlight the importance of problematizing such representational practice usage and effects. This study therefore examines the discourses and meanings found in whitewashed characters from selected scenes in the films Dr Strange and Prince of Persia. The study used the theoretical approach of Richard Dyers “Whiteness”, Edwards Saids “Orientalism and Stuart Halls “Representation and Stereotype “to gain a diverse perspective on the practices of representation. The applied methodological approach of Fairclough Critical Discourse Analysis enabled hidden and unspoken messages to be brought forth and discussed accordingly to the theoretical frames. The study found that the “whitewashing” is an excursion of symbolic power in which it is used to ultimately maintain and produce whiteness as the norm.

Whitewashing and colourism is prevalent in the Hollywood film industry due to decades of institutional racism, unconscious and implicit bias of professionals in the industry combined with the misconceptions of the success factors in film. This study will explore the capabilities of artificial intelligence helping the industry professionals of Hollywood film fight against whitewashing. The research of AI and AI-human interactions are a relatively young phenomenon, let alone combining artificial intelligence with the creative process of creating a cast. This means that the whole research is a relatively new territory and will be examined from an exploratory point of view. The findings of this paper suggest that AI certainly has the capabilities to help Hollywood make more diverse casts with more equal representation of people of colour, but technology needs to face and overcome the biases of its own first. / Whitewashing och colourism är utbredda fenomen i filmbranschen i Hollywood. Detta är en konsekvens av årtionden av institutionell rasism, omedvetna och indirekta fördomar hos branschfolk, i kombination med missuppfattningar om framgångsfaktorerna i filmer. Denna studie kommer att undersöka möjligheten att använda artificiell intelligens som ett hjälpmedel för professionella inom filmindustrin i Hollywood i kampen mot whitewashing. Forskningen av AI (artificiell intelligens) och AI-mänskliga interaktioner är ett relativt nytt fenomen, likaså att kombinera artificiell intelligens med den kreativa processen av att skapa en gjutning. Detta innebär att forskningen kring ämnet är ett relativt nytt område och kommer att undersökas ur en utforskande synvinkel. Resultatet av denna avhandling tyder på att användningen av AI som ett hjälpmedel kan resultera i mer varierande rollbesättningar, med bredare representation av färgade personer, men tekniken måste till att börja med möta och övervinna sina egna fördomar.

Colorblind TV: Primetime Politics of Race in Television Casting posits that in our current racially colorblind society, oftentimes actors of color are cast to prove that multiculturalism is important. However, the characters often have little cultural specificity and are only different in terms of skin tone. While this type of sameness on the surface may appear to offer a sense of racial parity, it actually encourages the opposite. Colorblindness works to make race immutable and objective, which inevitably disallows difference and instead outputs “whiteness” as the normative standard. Through a series of interviews with casting directors and actors guild diversity representatives as well as an ethnographic account of an actual casting audition, the dissertation argues that for the media industry, colorblindness is both a way of avoiding the messiness of race and of denying actors of color the ability to be culturally specific. This denial occurs because ultimately the desire to work supersedes the desire to reject role universality. As a case study, I examine the hit ABC primetime show Grey’s Anatomy (2005- ) as a way of illustrating how the blindcasting process became a part of public discourse that the television industry both praised and disowned. / text

No description available.