![\"Masked](\"http:\/\/cdn.defenseone.com\/media\/img\/upload\/2017\/09\/07\/090517protestersNG\/defense-large.jpg\")
Artificial intelligence is giving rise to\u00a0<\/strong>unprecedented capabilities for surveillance<\/a>, from\u00a0facial recognition at bridge crossings<\/a>\u00a0to\u00a0the ability to identify thousands of people at once<\/a>. Now,\u00a0new research<\/a>\u00a0suggests that\u00a0AI<\/span>\u00a0could potentially be used to identify people who have taken steps to conceal their identities by wearing hats, sunglasses, or scarves over their\u00a0faces.<\/p>\nThe paper, accepted to appear in a\u00a0computer vision conference workshop<\/a>\u00a0next month and detailed in\u00a0Jack Clark\u2019s ImportAI newsletter<\/a>, shows that identifying people covering their faces\u00a0is<\/em>\u00a0possible, but there\u2019s a long way to go before it\u2019s accurate enough to be relied upon. Researchers used a deep-learning algorithm\u2014a flavor of artificial intelligence that detects patterns within massive amounts of data\u2014to find specific points on a person\u2019s face and analyze the distance between those points. When asked to compare a face concealed by a hat or scarf against photos of five people, the algorithm was able to correctly identify the person 56% of the time. If the face was also wearing glasses, that number dropped to\u00a043%.<\/p>\n But those imperfect results don\u2019t mean the paper should be ignored. The team, with members from the University of Cambridge, India\u2019s National Institute of Technology, and the Indian Institute of Science, also released two datasets of disguised and undisguised faces for others to test and improve the technology. (Data has been\u00a0shown to be a key component<\/a>\u00a0for driving progress in the field of\u00a0AI<\/span>; when deep-learning algorithms have more data to analyze, they can identify patterns in the data with greater\u00a0accuracy.)<\/p>\n The algorithm for identifying disguised faces maps 14 points on a person\u2019s face, and then uses the distance between those points to identify them again in other images. The points mainly focus on the area around the eyes, but also designate the tip of the nose and corners of the mouth. (The placement of these markers indicates that large glasses alone might confuse the algorithm, but the paper\u2019s authors don\u2019t disclose the accuracy of a glasses-only\u00a0test.)<\/p>\n Glasses have been researched before as a way to evade facial recognition. A team of Carnegie Mellon researchers\u00a0printed glasses<\/a>\u00a0with a pattern custom-built to fool facial-recognition algorithms into misidentifying the wearer as someone else, such as a celebrity. Other attempts at beating identifying algorithms have included custom scarves with patterns that\u00a0look, to machines, like human faces<\/a>.<\/p>\n But faces aren\u2019t the only way to identify a person\u2014other\u00a0AI<\/span>\u00a0research indicates that the way a person walks is\u00a0almost like a fingerprint<\/a>. Researchers achieved more than 99% accuracy when using a deep-learning algorithm to identify one person\u2019s gait from 154\u00a0others.<\/p>\n The new research skips past\u00a0generic fear-mongering<\/a>\u00a0about artificial intelligence\u00a0to get more directly at the realistic implications of\u00a0AI<\/span>\u00a0systems being developed and used by unchecked or authoritarian government powers. Facial-recognition technology that could bypass disguises, for example, would be immensely useful for identifying political dissidents or\u00a0protestors.<\/p>\n As technology writer and New York Times opinion contributor Zeynep Tufekci\u00a0tweeted<\/a>, \u201cToo many worry about what\u00a0AI<\/span>\u2014as if some independent entity\u2014will do to us. Too few people worry what *power* will do *with*\u00a0AI<\/span>.\u201d<\/p>\n