Abstract:

In today's interconnected systems, the ability to monitor session data in real-time is important for securing systems, performance, and identifying breaches. Applications with a data focus now have to handle large quantities of user interactions, making it essential to identify anomalies associated with data transfers that could indicate system or data abuse. In this paper, we present a session data tracker based on finite automata that can continuously monitor the total session size and transition between defined states based on threshold values. For instance, based on the data size the automata would categorize usage between 0-25MB, 25-50MB, 50-75MB, and >100MB. Once the total data size exceeded 100MB, the automata would move into a warning state where alerts and possible mitigation actions could be automated. Our tracker will keep these transitions deterministic with minimal computational overheads, and provide near real-time readings, where machine learning and/or statistical designs do not fit. In practice, the automata is evaluated across several experiments, is able to 100% detect anomalies with low latency. The simulation with practice design showed that the automata would be effective in network data translates into cloud infrastructure and IoT systems.