Leveraging machine learning at ESC Minneapolis 2017
I find it difficult to believe that we are now only five weeks away from ESC Minneapolis (I'm still recovering from the excitement of ESC Boston earlier this year).
Do you remember the original Doctor Dolittle film from 1967 starring Rex Harrison? (You can call me "old fashioned" if you will, but I much prefer the original to the 1997 remake.) When wishing to travel, the Doctor selected his destination at random (I seem to recall him closing his eyes, opening an atlas, and plonking his finger down on the map).
Well, that's pretty much the way I feel with regards to the conference topics at ESC Minneapolis. I could close my eyes, scroll through the conference schedule, place my finger at a random location on the screen, then look to discover something interesting.
I just tried doing this. The first session I landed on was How to Select the Right Connector for Your Application by Randy Schueller, Senior Member of the Technical Staff at DfR Solutions. Funnily enough, I was pondering the topic of choosing the right connector for the task at hand only a couple of weeks ago (see Prognostication Connections).
The next session I ran across using my "eyes-closed" technique was Single Event Effects & their Impact on Safety Critical Systems by Scott Meyers, Staff Software Engineer at UTC Aerospace Systems. Once again, things like single event upsets (SEUs) and single event transients (SETs) are effects in which I am very much interested.
"What the heck," I thought, "let's spin the wheel again." I'm really interested in the advanced technologies that are appearing in today's embedded systems -- things like deep learning and artificial intelligence and augmented reality (see also my own presentation: Advanced Technologies for 21st Century Embedded Systems) -- so you can only imagine my surprise to see the session Leveraging Machine Learning to Test the Robustness of Your Protocols by Sagar Patel, Security Software Engineer at the Battelle Memorial Institute. The description of Sagar's session reads as follows:
Fuzz testing is a technique to automatically provide random, malformed, and unexpected data to a system in order to test its robustness. Fuzzing can be an incredibly powerful technique in detecting flaws and weaknesses in the implementations of communication protocols. Using commercially available tools, or developing your own is usually communication protocol specific, which can be cumbersome and rigid. This talk discusses how machine learning algorithms can be leveraged to rapidly and automatically reverse engineer and fuzz any communication protocol. This proposed approach would result in an easier to build and more flexible to use fuzz testing tool.
I tell you. It seems like I'm introduced to new applications for machine learning every day. It also seems like I'm constantly saying to myself, "Ooh, I would never have thought of that!" As far as I'm concerned, this presentation is definitely on my "must see" list.
Are you planning on attending ESC Minneapolis? If so, keep an eye open and say "Hi" if you see me wandering around. I'll be the one in the Hawaiian shirt. As always, all you have to do is shout "Max, Beer!" or "Max, Bacon!" to be assured of my undivided attention.