New adversarial techniques developed by engineers can make objects 'invisible' to image detection systems that use deep-learning algorithms. These techniques can also trick systems into thinking they see another object or can change the location of objects. The technique mitigates the risk for compromise in automated image processing systems.