| Resumo: | Nowadays, the World Wide Web (WWW) is simultaneously an accumulator and a provider of huge amounts of information, which is delivered to users through news, blogs, social networks, etc. The exponential growth of information is a major challenge for the community in general, since the frequent demand and correlation of news becomes a repetitive task, potentially tedious and prone to errors. Although information scrutiny is still performed manually and on a regular basis by most people, the emergence of Open-Source Intelligence (OSINT) systems in recent years for monitoring, selection and extraction of textual information from social networks and the Web promise to change the life of some of them. These systems are now very popular and useful tools for professionals from different areas, such as the cyber-security community, where being updated with the latest news and trends can lead to a direct impact on threat response. This work aims to address the previously motivated problem through the implementation of a dynamic OSINT system. For this system, two algorithms were developed: one to dynamically add, remove and rate user accounts with relevant tweets in the computer security area; and another one to classify the publications of those users. The relevance of a user depends not only on how frequently he publishes, but also on his importance (status) in the social network, as well as on the relevance of the information published by him. Text mining functions are proposed herein to achieve the objective of measuring the relevance of text segments. The proposed approach is innovative, involving dynamic management of the relevance of users and their publications, thus ensuring a more reliable and important source of information framework. Apart from the algorithms and functions on which they were build (which were also proposed in the scope of this work), this dissertation describes several experiments and tests used in their evaluation. The qualitative results are very interesting and demonstrate the practical usefulness of the approach. In terms of human-machine interface, a mural of information, generated dynamically and automatically from the social network Twitter, is provided to the end-user. In the current version of the system, the mural is presented in the form of a web page, highlighting the news by its relevancy (red for high relevance, yellow for moderate relevance, and green for low relevance). The main contributions of this work are the two proposed algorithms and their evaluation. A fully working prototype of a system with their implementation, along with a mural for showing selected news, is another important output of this work.
|
|---|