Big data analysis for social network

For years, social networks have been responsible for enormous production of data exchanged on the network. The number of people connected to social networks has changed over the years and their habits connected to them as well. The Polytechnic of Turin used passive probes to capture internet traffic for 5 years, from 2013 to 2018, collecting 218TB of compressed TCP log data relating to connection flows to a central node of the network in Italy. The main topic of this thesis work is to isolate and analyze this logs to find out how to define a heuristic able to group user flows to Facebook in sessions and then analyze these sessions to understand how the amounts of data and habits of users towards Facebook have changed over the years. The initial logs were filtered to isolate data related to interactions with Facebook. Later the logs were filtered to eliminate the interactions caused by social buttons, which do not correspond to users' active sessions on Facebook. A heuristic has been studied and applied on the filtered data to group user flows into sessions. Finally, after validating the data obtained on different samples and comparing them with historical data in the literature, a quantitative analysis was made of the results obtained to see how the behavior of Facebook users has changed over the years. The results obtained show a tendency to anticipate the first connection to the platform, a longer time window for the most intense sessions and, few changes for the seasonal pattern of use. From the point of view of the visit frequency, the results show an increase in the average number of daily visits per user and a decrease in the average time between two consecutive sessions. These latest results, combined with the previous ones, confirm a hypothesis of compulsiveness of users towards the use of the Facebook platform.