🔴 AI in Digital Forensics: When Can We Trust Automatic Classification?

Artificial intelligence can dramatically reduce the time needed to process large amounts of images and videos. But good accuracy doesn't automatically mean a system is reliable for a specific forensic task. The most sensitive issue may not be what the model misclassifies before our eyes, but what it leaves out without ever drawing our attention.

For this reason, before entrusting a classifier with a part of the research, we should understand what errors it can make, on which data we actually know its performance, and what consequences that error can have on our analysis.