Evaluation of the privacy sisk for online search and social tagging systems

[ANGLÈS] The lack of privacy is nowadays a serious security issue for users on the Internet, since personalized information systems are in fact collecting an incredible amount of information, to the point where it appears they know us better than ourselves. Every single step we take on the Web can b...

Descripción completa

Detalles Bibliográficos
Autores: Rodríguez Hoyos, Ana Fernanda, Estrada Jiménez, José Antonio
Tipo de recurso: tesis de maestría
Fecha de publicación:2013
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2099.1/19550
Acceso en línea:https://hdl.handle.net/2099.1/19550
Access Level:acceso abierto
Palabra clave:Computer security
Internet -- Security measures
browser extension
profiling
privacy metrics
social tagging
user’s privacy
query obfuscation
extensión de navegador
obtención de perfiles
métricas de privacidad
etiquetado social
privacidad de usuario
ofuscación de consultas
Seguretat informàtica
Internet -- Mesures de seguretat
Àrees temàtiques de la UPC::Informàtica::Seguretat informàtica
Descripción
Sumario:[ANGLÈS] The lack of privacy is nowadays a serious security issue for users on the Internet, since personalized information systems are in fact collecting an incredible amount of information, to the point where it appears they know us better than ourselves. Every single step we take on the Web can be detected by the same companies that are providing the most popular services, such as search engines and social networks. Search queries, tags, clicks, e-mails, tweets can be used to get an accurate profile of our activity. Even when the profiling of this information is a huge privacy risk, only a few tools propose a protection mechanism taking this into consideration. TrackMeNot is one of these tools, which implements perturbation of a user’s queries, by generating fake ones, to obfuscate the user’s profile , although, similarly to other tools, it is not immediate to evaluate to what extent it is protecting users privacy. We have firstly contributed to the another work, where the user’s privacy is measured, interpreted and showed for the user in the Firefox web browser. Privacy is computed based on the search queries of the user. But, since social tags are an important component of users’ data, we have developed some modules to take these into consideration in the process of measuring privacy, and not only the search queries. Additionally, we also evaluated the obfuscation mechanism offered by TrackMeNot to protect user’s privacy. By implementing justified privacy metrics we measured the efficiency of TrackMeNot to enhance the user’s privacy. We found that for identifying attacks TrackMeNot importantly improved the user’s privacy. However, for more sophisticated attacks, such as classification attacks, the obfuscation mechanism was not successful enough. The way of generating fake queries is crucial to efficiently obfuscate the user’s profile against classification attacks. Finally, TrackMeNot was integrated with PrivMeter in such a way that the privacy metrics are showed both for the real and obfuscated user’s profiles.