Seminars

View all Seminars  |  Download ICal for this event

A Fuzz Tester for IEC 61850-based IED Devices using Reinforcement Learning

Series: M.Tech (Research) Colloquium

Speaker: KANMANI A . M.Tech (Research) student, Computer Science and Engg. under Computer Science and Automation

Date/Time: Jan 29 10:00:00

Location: CSA Auditorium, (Room No. 104, Ground Floor)

Faculty Advisor: Prof. Vinod Ganapathy

Abstract:
The goal of this research is to uncover potential vulnerabilities in Intelligent Electronic Devices (IEDs), which are critical components of power grid systems used in generation, transmission, and distribution of power. Communication between these devices is regu- lated by the IEC 61850 standard, which specifies the Manufacturing Message Specification (MMS) protocol over a TCP/IP network stack. To identify vulnerabilities, we employ fuzzing, a technique that involves sending inputs and analyzing outputs. We start by crafting a series of valid input requests targeting various data points within the IEDs. Using a black-box fuzzing approach, we transmit MMS IEC 61850 request packets to the IEDs and use the response packets as feedback for categorization. This approach helps us identify interest- ing requests that can be further probed and mutated to increase fuzz testing code coverage. Our methodology involves developing a re inforcement learning agent that is rewarded for exploring new responses and crashes, while being penalized for revisiting previously encountered responses. The agent learns the optimal sequence of mutations for any specific request packet to generate new responses and crashes. This fuzzing experiment is designed to identify issues in the communication module by traversing new paths that trigger unexplored responses.