|dc.description.abstract||Web logs are a set of recorded events between clients and web servers. Information provided by these events is valuable to computer system administrators, digital forensic investigators and system security personnel during digital investigations. It is important for these entities to understand when certain system events were initiated and by whom. To achieve this, it is fundamental to gather related evidence to the crime from log files. These forensic procedures however pose a major challenge due to large sizes of the web log files, difficulty in understanding and correlating to attack patterns associated to digital crimes. The connections of events that are remotely positioned in the large log files require extensive computational manpower.
This dissertation proposes the design, implementation and evaluation of a web log analysis system based on temporal logic and reconstruction. The case study will be on web server misuse. Temporal Logic operators represent system changes over time. The reconstruction of records in web server log files as streams will enable the implementation of temporal logic on the streaming data. The web server attack patterns established will be described by a special subset of temporal logic known as MSFOMTL (Many Sorted First Order Metric Temporal Logic). The attack patterns will be written in a special EPL (Event Processing Language) as queries and be parsed through Esper, a Complex Event Processing (CEP) engine. To ensure the proposed system increases the quality of log analysis process, log analysis will be performed based on a time window mechanism on sorted log files.||en_US