New adversarial techniques developed by engineers at Southwest Research Institute 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. According to the researchers, deep-learning neural networks are highly effective at many tasks, but they were adopted so quickly that the security implications of the algorithms weren't fully considered.
Deep-learning algorithms excel at using shapes and color to recognize the differences between humans and animals or cars and trucks, for example. These systems reliably detect objects under an array of conditions and are, therefore, used in myriad applications and industries, often for safety-critical uses. The automotive industry uses deep-learning object detection systems on roadways for lane-assist, lane-departure, and collision-avoidance technologies. Vehicles rely on cameras to detect potentially hazardous objects around them. While the image processing systems are vital for protecting lives and property, the algorithms can be deceived by parties intent on causing harm.
Security researchers working in “adversarial learning” are finding and documenting vulnerabilities in deep- and other machine-learning algorithms. They have developed what look like futuristic patterns. When worn by a person or mounted on a vehicle, the patterns trick object detection cameras into thinking the objects aren't there, that they're something else, or that they're in another location. Malicious parties could place these patterns near roadways, potentially creating chaos for vehicles equipped with object detectors. The patterns cause the algorithms in the camera to either misclassify or mislocate objects, creating a vulnerability. The researchers call these patterns ‘perception invariant’ adversarial examples because they don't need to cover the entire object or be parallel to the camera to trick the algorithm. The algorithms can misclassify the object as long as they sense some part of the pattern.
While they might look like unique and colorful displays of art to the human eye, the patterns are designed in such a way that object-detection camera systems see them very specifically. A pattern disguised as an advertisement on the back of a stopped bus could make a collision-avoidance system think it sees a harmless shopping bag instead of the bus. If the vehicle's camera fails to detect the true object, it could continue moving forward and hit the bus, causing a potentially serious collision.
The first step to resolving the danger is to test the deep-learning algorithms. The team has created a framework capable of repeatedly testing these attacks against a variety of deep-learning detection programs, which will be extremely useful for testing solutions.
The researchers continue to evaluate how much, or how little, of the pattern is needed to misclassify or mislocate an object. Working with clients, this research will allow the team to test object detection systems and ultimately improve the security of deep-learning algorithms.
For more information, contact Maria Stothoff at 210-522-3305.