On May 18, 2024, the 2024 West Lake Sword · The special conference on AI leading the new wave of digital security was successfully held. Focusing on AI+ security, industry elites will discuss how AI will lead the development of new quality productivity of digital security, and unlock the infinite mystery of the future.
Lu Wei, Vice Chairman of China Cyberspace Security Association and Director of Artificial Intelligence Security Governance Professional Committee of China Cyberspace Security Association, Wang Juhong, Member of the Party Group and Vice Chairman of Hangzhou Science and Technology Association of Zhejiang Province, Wang Xiaozheng, General Manager of Information Technology and Data Management Department of Zhejiang Mobile and Chief Expert of China Mobile Group, Wang Yanchang, General Manager of Information Security Department of Zhejiang Mobile, Yu Yongping, Director of Management Department (Science and Technology Innovation Department) of Hangzhou Association for Science and Technology, He Wentian, Rotating President of Hailiang Group, Fan Yuan, Chairman of Anheng Information, Jin Ling, Deputy Director of the Infrastructure Department of Shanghai Big Data Center, Yu Nenghai, Executive Dean of the School of Cyberspace Security, University of Science and Technology of China, Yu Ruiguo, Director of the Information and Network Center of Tianjin University, Zhao Xinglun, Secretary-General of Hangzhou Artificial Intelligence Association, and relevant industry leaders and enterprise experts attended the forum.
Hu Ruimin, Chairman of the Cyberspace Security Working Committee of the New Engineering Alliance of the Ministry of Education and Foreign Academician of the Russian National Academy of Engineering, presided over the forum
Lu Wei, vice chairman of the China Cyberspace Security Association and director of the Artificial Intelligence Security Governance Professional Committee of the China Cyberspace Security Association, attended the meeting
AI leads to building a new quality of productivity for digital security
Since 2023, artificial intelligence has been booming, bringing a new ecological environment and industrial pattern to the field of digital security. While embracing AI, we must also guard against the risks that AI may bring, such as data leakage and deepfakes. Using AI to AI and AI to manage AI has become an inevitable choice to promote the deep integration of artificial intelligence and network security technology, build a new quality productivity of digital security, and provide a strong guarantee for economic and social development and national security.
Yu Rui, executive deputy director of the CCF Computer Security Professional Committee and deputy director of the First Research Institute of the Ministry of Public Security, said in a video speech that the rapid development of artificial intelligence and the wide application of AI technology have brought new challenges and opportunities to digital security, and issues such as data security, algorithm security, and privacy protection have become increasingly prominent. He put forward four suggestions on how to build a more solid digital security barrier, realize the revolutionary upgrade of traditional network and data security, and create a new AI+ network and data security system while enjoying technological dividends: first, strengthen collaborative innovation and build a security ecosystem; the second is to carry out technological innovation to drive safety upgrades; the third is to increase the guidance of laws and regulations to ensure healthy development; Fourth, strengthen personnel training and strengthen the foundation of safety.
Director Yu Rui delivered a video speech at the forum
Professor Yu Nenghai pointed out in his keynote speech on "Generative AI Security and New Quality Digital Assets" that artificial intelligence has made a leap to generative and comprehensively improved its understanding ability, AIGC has continuously made breakthroughs, and multi-modal large models have developed rapidly, and the guarantee of AI security will become an important foundation for the reliable and controllable application of artificial intelligence. In order to avoid the risk of AIGC, it is necessary not to rely solely on intelligent algorithms, but to effectively combine human creativity with AI's generative capabilities. In this regard, Professor Yu suggested that it should be carried out from three dimensions; The first is to use AIGC to synthesize content identification to deal with the malicious application of AI; The second is to improve the security capability of AI through large-scale model correction; The third is to protect AI-enabled systems with the help of generative model watermarks.
Yu Nenghai delivered a keynote speech on "Generative Artificial Intelligence Security and New Quality Digital Assets".
Wang Jian, Partner of Ernst & Young (China) Advisory Co., Ltd., pointed out in his keynote speech on "Evolution and Application of AI Large Model Technology" that the future evolution route of AI large model technology may develop along six dimensions: first, the growth of model scale and the surge in the number of large model parameters; the second is multimodal learning, which can handle applications and scenarios of multiple data types; the third is to improve the efficiency and sustainability of model training and inference; Fourth, improve the transparency and explainability of the model decision-making process and reduce the "black box" phenomenon; Fifth, the model has the ability to learn continuously, avoid catastrophic forgetting, and adapt to new data and changes in the dynamic environment; Sixth, ensure the stability and security of the model under malicious attacks, and enhance robustness.
Wang Jian delivered a keynote speech on "Evolution and Application of AI Large Model Technology".
Al+ Security, exploring the innovative future of the security industry
In the current security industry, there are problems such as security balance and capability gap from a technical perspective, and security issues such as data security, content security, business security, and supply chain security in the AI era are particularly prominent. In this context, Wang Xin, deputy secretary-general of the Hangzhou Artificial Intelligence Society and president of the Anheng Information Research Institute, proposed in his keynote speech "Application Exploration of Al in the Security Industry" that AI+ security must be a huge direction for the future development of security.
Wang Xin pointed out that at the beginning of its establishment, artificial intelligence is a technical discipline that simulates, extends and expands the theory, method, technology and application system of human intelligence. Compared with the previous artificial intelligence technology, the emergence ability of large models has been enhanced in four dimensions: understanding instructions, understanding examples, chain of thought ability and rapid learning ability, and sharing the practical experience of security large models in multiple scenarios such as data security and security operation.
Wang Xin believes that the real AI capability of AI for security is only less than 5%, and he has made eight predictions on the application trend of AI for security: First, AI agents with large models as the core will become the standard configuration of the security industry, and more and more AI native tools will become the "hands and feet" of Agents. Second, under the influence of AI agents, the adaptive security architecture will be upgraded. Third, under the influence of AI agents, robotic process automation will be combined with many security businesses. Fourth, AI agents will reshape the way security work is collaborative. Fifth, with the continuous evolution of AI multimodal technology, public opinion warfare, information warfare, and new types of cyber-related crimes will take the lead in applying AI technology to obtain technological "dividends", and supervision will face unprecedented challenges. Sixth, security attack and defense will enter a new stage, intelligent attack and intelligent defense will become a new field of technology game, and AI-native worms may sweep the world. Seventh, security issues such as data security, content security, business security, and supply chain security in the AI era are particularly prominent, which will bring challenges to the application and development of the AI industry. Eighth, the AI era will bring comprehensive upgrades and innovations to security products, and the atomization of capabilities and the standardization of interfaces will be rapidly promoted.
Wang Xin delivered a keynote speech on "Application Exploration of Al in the Security Industry".
Practice first, proofing for the intelligent transformation of the industry
Yu Ruiguo, Director of the Information Network Center of Tianjin University, introduced the successful experience of Tianjin University in response to the national policy guidance of vigorously promoting the construction of supercomputing and artificial intelligence platforms, building a university-level computing resource sharing platform, effectively supporting the development and innovation of major projects or topics, and serving AI for Science.
Yu Ruiguo delivered a keynote speech on "Integration Construction of Computing Power and Network Security in Tianjin University".
Tao Lifeng, General Manager of Anheng Information Privacy Computing Division, delivered a keynote speech on "Large Model Training and Inference Data Fencing Based on Confidential Computing", introducing how to solve the contradictory needs of data utilization and protection, as well as the dual problems of protecting model security, in large model training and inference. On the basis of privacy computing, ANHENG Information supports the training and inference capabilities of large models to ensure that the training data of large models can be trained in a safer, more credible and reliable environment, and the models can be deployed and applied in a safer environment. Tao Lifeng said that by combining the trusted execution environment, confidential computing and large model training framework, building a data fence based on confidential computing for large model training and inference, combined with the whole process of encryption, authorization, auditing, approval and other control methods, it is possible to achieve scenario-based minimized data authorization, while ensuring that data will not be abused in a controlled environment.
Tao Lifeng delivered a keynote speech on "Large Model Training and Inference Data Fence Based on Confidential Computing".
In response to the challenges of large-scale model development and large-scale AI computing systems, ANHENG Information officially released the Ascend-Hengnao inference all-in-one machine to accelerate the innovation and implementation of large models in the whole process.
In the unremitting pursuit of artificial intelligence high-performance computing, ANHENG Information and Huawei's Ascend team have experienced long-term and in-depth cooperation and co-creation. Based on the Ascend software and hardware ecosystem, ANHENG Brain has made a series of important achievements, and at the same time, it has also taken the lead in completing the hardware adaptation of multiple generations of Ascend. The all-in-one machine integrates the Constant Brain large model independently developed by ANHENG Information and Huawei's industry-leading AI computing capabilities, which not only accelerates the innovative application of large models in the whole process, but also demonstrates the continuous innovation capabilities of both parties in the field of artificial intelligence.
Wu Weijing, General Manager of Anheng Information Ecosystem Cooperation Department, and Yu Zhuo, Vice President of Huawei's Manufacturing and Large Enterprise Corps Solutions, attended the ceremony. In the future, the two parties will continue to strengthen cooperation under the framework of the ANHENG-Huawei strategic cooperation, and develop and innovate in the fields of large-scale model technology, software and hardware ecosystem, and cloud services, so as to jointly promote the progress of intelligent digital security.
Ascend-Constant Brain Reasoning All-in-One Machine Release Ceremony
Yao Longfei, general manager of the innovation business department of Anheng Information Research Institute, proposed in his keynote speech on "Hengbrain 2.0 Intelligent Twin Core and Automation Innovation" that the second half of the large model should be the explosion of agents. He took the large model of Hengnao safety vertical domain as an example to introduce the exploration and application of the combination of products and business driven by ANHENG Information with agents as the core. At present, ANHENG Information has fully understood the idea of RPA and integrated it into the design of the agent framework, and has been widely used in ANHENG's data classification and grading products, Alpha alarm research and judgment, DLP, API, SOAR and other products, which has greatly improved the ability in different scenarios.
Yao Longfei delivered a keynote speech on "Constant Brain 2.0 Intelligent Body Core and Automation Innovation".
Sun Wen, Director of the Digital Innovation Department of Hailiang Group, delivered a keynote speech on "Exploration and Practice of Hailiang Group's Large Model Application in Data Security", sharing that Hailiang Group is exploring the application of AI large models in different business scenarios, especially in the field of data security. She said that as enterprise boundaries continue to expand and scenarios become more complex, traditional security protections are gradually becoming ineffective. At the same time, enterprise data is becoming more and more valuable with the deepening of digital transformation, and the risk of data breaches is increasing day by day. To this end, Hailiang Group has built a "113N" security capacity building framework to do end-to-end protection from "before-during-after", so as to achieve the goal of external attack and internal leakage prevention, and escort digital transformation with security. Sun Wen said that Hailiang Group will work with Anheng Information to train more agents suitable for different enterprises and focus on different industries based on the professional capabilities of Anheng Brain, Hailiang's rich business scenarios and multi-field industry data precipitation.
Sun Wen delivered a keynote speech on "Exploration and Practice of Hailiang Group's Large Model Application to Data Security".
At the subsequent unveiling ceremony of the Hailiang-Anheng Joint Innovation Laboratory, He Wentian, Rotating President of Hailiang Group, Fan Yuan, Chairman of Anheng Information, Wang Xin, President of Anheng Information Research Institute, and Sun Wen, Director of the Digital Innovation Department of Hailiang Group, witnessed the unveiling of the joint innovation laboratory. Explore the application of large model + data security, create an advanced benchmark case of domestic large model + data security, and form an industrial synergy.
The unveiling ceremony of Hailiang-Anheng Joint Innovation Laboratory
Guo Jianxin, head of the R&D of Volcano Engine large model security products, summarized the four risk points of large model application in the "Internal Practice Sharing of Large Model Security", which are model output anomaly, dialogue mechanism, data security and system security. In this regard, the Volcano Engine Ark platform takes business security and base security as the core, and builds a solution of "one spear", "one shield" and "one sandbox" to ensure the security of large model applications.
Guo Jianxin's keynote speech on "Internal Practice Sharing of Large Model Security".
AI is a sword to discuss the balance between security and development
With the vigorous development of artificial intelligence technology, which industries or fields will have significant effects and bright development prospects after the introduction of large model technology? How to balance security and development? What new application scenarios or business models may emerge in the future?
Under the chairmanship of Wang Yaqi, Deputy Secretary-General of Hangzhou Artificial Intelligence Society and Lecturer of the School of Media Engineering of Zhejiang University of Media and Communication, Hu Ruimin, Chairman of the Cyberspace Security Working Committee of the New Engineering Alliance of the Ministry of Education and Foreign Academician of the Russian National Academy of Engineering, Wang Zheng, Deputy Director of the Data Economy Research Center of Ali Research Institute, Yu Zhuo, Vice President of Huawei Manufacturing and Large Enterprise Corps Solutions, and Shui Xuefei, Vice President of Anheng Information Research Institute, had a fierce collision on "the application potential and security impact of large models in various industries".
Panel Discussion: AI on the Sword - The Application Potential of Large Models in Various Industries for Security Impact
Hu Ruimin summarized the three elements of the large model landing scenario: data resources are abundant enough, high-end talents are extremely scarce, and there is an urgent need to improve efficiency; Industries and fields that meet these three elements are suitable for applying large models for empowerment and enrichment. Yu shared Huawei's own experience, saying that large models have been applied to improve quality and efficiency in the To B scenario, and it is necessary to seize the appropriate scenarios to improve productivity. Wang Zheng believes that AI applications have changed from Microsoft's AI assistant to more AI agents, reflecting the improvement of machines' planning, decision-making and execution capabilities. In the medium and long term, the development trend of large-scale model technology is to improve the autonomy of machines under the premise of human control. Shui Xuefei believes that from the recent development of the world's mainstream large models, the large models have reached a period of rapid ramp-up of application landing; In the field of security, the confrontation between the attacker and the defender on the application of large models has begun, and ANHENG is also promoting the application of large models and agents to build a new generation of security defense line.
Finally, Huang Lianjin, Vice President of CSA Greater China Research Institute, delivered a live speech on "Security Challenges of Generative AI Applications", pointing out that the security challenges of generative AI mainly include shadow AI, automated and scalable threats and zero-day vulnerabilities, security tool integration, malicious tools, data leakage caused by large model aggregation, and the use of AI to make the original permission policy issues more prominent. The overall survey of the 2024 AI security industry map was introduced.
Huang Lianjin's live speech on "Security Challenges of Generative AI Applications".
The "Special Conference on AI Leading the New Wave of Digital Security" was guided by the Zhejiang Association for Science and Technology, hosted by the Hangzhou Association for Science and Technology, and organized by the Hangzhou Artificial Intelligence Society, the Computer Security Professional Committee of the China Computer Federation, the Cloud Security Alliance Greater China, the Hangzhou Research Institute of Xidian, and Anheng Information.