zhou hongyi talks about 360's big security model: it is equivalent to having l4 level "autonomous driving" capabilities
2024-09-05
한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina
from september 3rd to 5th, the second cyberspace security (tianjin) forum with the theme of "building cybersecurity together, governing cyberspace together" was held in tianjin binhai new area. zhou hongyi, founder of 360 group, attended the forum and delivered a keynote speech entitled "building a security big model, reshaping security with ai". zhou hongyi shared his understanding of the cybersecurity situation in the era of artificial intelligence big models from the perspective of the front line of the cybersecurity industry, and introduced the specific practices and experiences of 360 group.
zhou hongyi introduced that 360 group has invested nearly 30 billion yuan in the cybersecurity industry in the past 10 years, which is more than the total investment of the second to tenth companies in the industry. it has created the world's largest cybersecurity big data and has become the leader in defending national cybersecurity. "despite the huge investment, 360 has no hesitation and will continue to increase investment to contribute to national cybersecurity."
"in an era where big models reshape all industries and products, ai will definitely become a 'tool for bad guys', and the security industry will upgrade from 'human-to-human confrontation' to 'human-to-machine confrontation'," zhou hongyi said. in addition, the characteristics of big models will also lead to issues including ai attack and defense security, ai's own security, and ai ethical security. therefore, ai must be used to respond to ai attacks, that is, "fighting models with models" and defeating magic with "modeling methods."
zhou hongyi proposed that the key to reshaping safety with ai is whether safety can be "autonomous driving" and form a complete closed loop of safety operations. 360 has developed a large safety model based on artificial intelligence technology to achieve the transformation from "rule-driven" to "learning-driven". based on the five levels of autonomous driving from l1 to l5, 360 is equivalent to having l4 "autonomous driving" capabilities.
regarding the difficulty of building a big security model, zhou hongyi pointed out that it is necessary to have rich experience in building big models, rich training data, and deep integration with security business. in the final analysis, it is necessary to have a set of effective professional big model methodology. this set of methodologies mainly includes five steps, namely finding star scenes, knowledge management, expert collaboration, building intelligent bodies, and finally business integration. zhou hongyi introduced that under the guidance of this set of methodologies, 360 has built a big security model, which includes multiple expert models such as attack detection, operation disposal, tracking and tracing, knowledge management, data protection, code security, and content security. after comparison and testing, the 360 security big model has surpassed gpt-4 in security capabilities.
"i have always said that we should bring big models down from the altar and make them affordable and useful to every enterprise." zhou hongyi introduced that 360's full line of security products have integrated big security model capabilities, and all users who purchase 360 standard products will be provided with big model standard capabilities free of charge. he said that in the future, 360 group will carry out the empowerment and upgrading of big security models, provide big security model capabilities based on the existing security construction achievements, and create new quality productivity and combat effectiveness of security.
