Seminario di Informatica: Privacy Preserving Machine Learning Techniques for Data Classification

Mercoledì 24 Luglio, ore 12:00 - Aula 1BC50 - Yogachandran Rahulamathavan



Mercoledì 24 Luglio 2013 alle ore 12:00 in aula 1BC50 della Torre Archimede, Yogachandran Rahulamathavan (City University London, UK) terrà un seminario dal titolo "Privacy Preserving Machine Learning Techniques for Data Classification".

The recent trend towards remote outsourcing and cloud computing can be exploited in many data classification tasks such as remote diagnostic and affect detection as cloud services. These services could be utilized by any users in a flexible manner such as on-demand or pay-per use. This new notion overcomes the difficulties that would be faced by the users such as having to collect a large number of samples (i.e. rich data set), and requiring high computational and storage resources to build their own data classification tool. However, there is now a risk that the cloud servers are potentially untrusted servers. Hence, releasing the data samples owned by the user to the untrusted server raises privacy concerns. This drawback can affect the adoption of outsourcing techniques. Furthermore, the server may not wish to disclose the features of the data classification tool even if it offers the service to the users. In this talk I will go through how to modify the well-known machine learning tools such as support vector machine in order to preserve the privacy of both the user data and the cloud data.

Download Seminari di Informatica