“AI solutions are secure by design, really?
Usually, hackers access traditional software systems to steal data. They hack industrial control systems and misguide them to do the wrong action(s). However, the core of AI systems has mainly algorithms and not much data. This created the illusion of absolute security by nature among some people, as nothing inside to steal. However, instead of stealing data, cyber attackers can feed AI systems with wrong data to manipulate their ability to take the right decisions. For example, attackers could access Electronic Medical Records (EMR) to add or remove medical conditions in MRI scans which will lead to the wrong diagnosis by ML algorithms. Same could happen to financial data or the operational data of critical equipment in a Nuclear Power Plant (NPP) or a smart grid.”