Mobile Internet Traffic Analysis: Session Level Approach

The accessibility and usage of the Internet has tremendously increased in recent years. Smartphones and mobile networks played a critical role to enhance accessibility and usage of Internet. The increase of data and it’s usage on Internet has also created many opportunities to analyze the data to gain understanding regarding its usage. This thesis focuses on analysis of the mobile Internet traffic on session level. The data utilized in this research are mobile Internet traffic traces from Finland. Two different measurement points of mobile networks were used to measure data, and these two data sets were mapped together in this thesis to provide detailed information regarding session level usage characteristics. The mappings of TCP/IP flows to the sessions from RADIUS server data were based on matching the timestamps and IP addresses. These mappings resulted in more than 94% accuracy. The device identification was based on HTTP user agent based approach and for Internet protocol identification Tstat classifier was utilized along with port number analysis. The session level statistics show that session volumes and durations are widely spread with majority of sessions having short duration and low traffic volume. The number of sessions generated from handsets is higher when compared to other devices. The device type analysis also shows that personal computer sessions have higher volume and last longer than handset and tablet sessions. The protocol level analysis indicates that web (HTTP) sessions are longer and have higher volume than email (IMAP4, POP3), and secured (SSL) sessions. The analysis shows that mapping of RADIUS data and TCP/IP flows provides deeper understanding of session level traffic characteristics. These mapping results could be utilized for modeling user behavior on a session level. In future this session level analysis could be used to gain knowledge regarding user’s behavior to support different pricing schemes for different user groups