Talking with one of our new business partners who recently joined forces with to bring radio based automation to an industry that is currently underserved, the traveling carnival. Many is the late night tripping incident down at the Stampede grounds, while some may blame an excess of Stampede spirits, the big power cables all over the grounds don’t make it any easier. Our new partnership is developing solutions for controlling amusement park rides centrally, eliminating wires and ensuring rider safety and experience don’t vary based on the individual operator.
We develop custom protocols here at YMMV, and of course the most important part of the protocol is controlling the motor/sensor/device as intended but there is also a lot of validation that needs to be done to ensure unplanned inputs don’t cause the whole system to crash. I had a tech buddy that had an uncanny knack of always inputting some weird value in my programs that would cause them to crash. While annoying, this is a valuable service because cyber security incidents often occur from miss-handled input, but not everyone has a Henri to do their testing.
This led to the AI discussion with Carousail‘s founder about programming an AI engine to generate thousands of variations on input to make sure your application fails safely. This isn’t a slight against the developers either, even the most secure code is still running on a web server the person didn’t write, which is running on an operating system the web server person didn’t write, which is being accessed by a network stack the OS person didn’t write and so on & so on. If Spectre and Solarwinds have taught us one thing it’s stack to supply chain is an almost infinitely divisible series of interrelationships that we’ll struggle to track despite well intentioned initiatives like software bill of materials.
So maybe AI making a million tries at random input won’t prove things are completely bullet proof but may give us more confidence even the most persistent attacker isn’t likely to get too far.