Ali Security Researcher Lu Quan: How to use AI to combat data pollution
Big data, calculations and algorithms are the three major elements of artificial intelligence. How to ensure data security and cleanliness of data is a key factor in improving the security of smart products. Recently, Internet companies such as BAT gathered in Shenzhen to conduct in-depth discussions on smart security at the 2018 Global Artificial Intelligence and Robotics Summit (CCF-GAIR).
As an invited guest, Lu Quan, senior director of Alibaba’s Security Department’s algorithm, pointed out that artificial intelligence brings more possibilities and brings many security risks. In particular, black and gray production is using "data pollution" and "data poisoning". The new complication means a fierce confrontation with the positive development of the artificial intelligence industry.
He cited the safety of artificial intelligence vehicles as an example. In addition to "using the intelligent algorithm itself to change traffic signs" or "traditional hacking infiltration to control other civilian cars", etc., you can control attacks outside the smart car, by moving to the smart device cloud. Or the system inputs pollution and “poisoning” data, which can also be used to attack artificial intelligence systems and devices.
The attack principle is that artificial intelligence is based on big data algorithms to achieve precise calculations, and ultimately produces actual behavior. It is this feature that allows the countermeasures to input some contaminated data to “feed” and “train” artificial intelligence systems. It deviates from the normal computing goal and eventually produces a defocusing operation, which leads to a series of security risks.
recommend