lone10
@lone10
Profile
Registered: 1 year, 4 months ago
Li Weibin: AI security is an arms race that must be invested immediately Looking back on 2021, in addition to the epidemic, chips and supply chain, information security is also the focus of the industry. On the one hand, the network systems of many well-known enterprises have been hacked or kidnapped, causing damage in different scopes; on the other hand, the government has also begun to actively regulate. Securities firms, futures, investment trust and investment consultants, and banking industries must set up chief security officers at or above the vice president level. The importance of information security (Cyber Security) is already a consensus, so what about AI that is gradually moving towards practical applications? How to ensure information security when using AI to avoid harm to life, property or the whole society? Li Weibin, chairman of the Artificial Intelligence Technology Foundation and executive director of Hon Hai Research Institute, is also a well-known information security expert in Taiwan. He pointed out that the security of AI itself is a challenge that various industries must face as soon as possible. Of course, it is also an important opportunity for Taiwan. AI information security is more complicated and difficult to solve Li Weibin said that due to the opacity, complexity and unexplainability of AI technology, it is difficult to implement information security measures. "Attacks may be carried out through original data, models, or even noise interference," he said. For example, if there is a pattern on the clothes that cannot be recognized by the human eye but is enough to interfere with the interpretation of the camera, it is likely that the self-driving car will misjudge the person wearing the clothes. "not human". Or because of the misjudgment of AI, the results of the health check will be wrong, which will lead to many personal and family problems. In recent years, AI has been widely used in web search, social software, natural language processing, self-driving cars, automatic decision-making systems, etc., all of which may become targets of attack. However, what is worrying is that the information security standards for AI systems have not yet appeared in Taiwan, and have not even begun to take off. "Many people discuss AI for cybersecurity (the application of artificial intelligence in information security), but the Security of AI (the security of artificial intelligence itself) has not received enough attention," Li Weibin said, according to the definition of the US National Science and Technology Council NSTC , AI Security should include the following items: Specification verification of systems to ensure functionality, safety, robustness and fairness Decisions made by AI can be trusted Detect and reduce adversarial inputs (Adversarial Inputs) AI-enabled systems are trusted Information security is the key to AI policies in various countries These items actually include consensus in various fields such as technology, patent standards, certification, laws and regulations, and even culture. It is impossible for a certain government unit or a single enterprise to create such a system. But unlike the network information security that everyone knows now, AI has almost become a part of everyone's life because of its wide range of applications and the need to collect a large amount of behavioral data. "It directly involves everyone's property, health and even life safety," Li Weibin said, so how to establish AI information security certification standards is a key issue that must be started as soon as possible.
Website: https://techtuba.com/
Forums
Topics Started: 0
Replies Created: 0
Forum Role: Participant