The in-depth development and rapid popularization of AIGC have opened up the vast imagination space of the intelligent world, but at the time of the emergence of technological achievements, the ensuing model security, data security, and content security issues have become increasingly prominent.
Recently, at the 2024 media and analyst communication meeting of H3C Group, a subsidiary of Tsinghua Unigroup, Sun Songer, senior vice president of H3C Group and president of H3C Information Security Technology Co., Ltd., was interviewed by a number of media and shared in depth the evolution of H3C's active security 3.0 in the three dimensions of comprehensive intelligent upgrade, new paradigm of innovation scenarios and comprehensive localization, which brought new solutions to strengthen AIGC data security governance, implement AIGC application specifications, and promote business security protection.
Sun Song'er, Senior Vice President of New H3C Group and President of New H3C Information Security Technology Co., Ltd
Security × AI and improve security capabilities in a positive cycle
For the large-scale model technology represented by AIGC and its application, development and security are a unified whole with one body and two wings, which must be planned and promoted simultaneously. Sun Songer said that the AIGC from the perspective of security of New H3C Group is based on the concept of "active security", which first meets the new security needs of large model applications, and at the same time uses AIGC as a new tool and new capability to empower the improvement of security capabilities, that is, "security is the core cornerstone of ensuring the actual combat of AIGC, and AIGC is the best means to improve security capabilities". In order to realize the positive cycle of AIGC development and security capability improvement, based on the AI capabilities of the Lingxi model, Xinhua Sanhei has created the Lingxi assistant to inject deeper intelligence into the construction of the security system, the improvement of operational efficiency, and the sustainable development of business, and put forward a new idea of security compliance and protection of the large model - to realize the security protection of the large model with the Lingxi Guard, and to realize the security of the large model with the Lingxi assistant.
The security protection of large models runs through the whole process from the training stage to the operation stage, and it is necessary to solve many problems such as the difficulty of training large models in vertical domains, the risk of network attacks in the large model environment, and the compliance risk of large model generated content. Among them, content security compliance is the first priority in the development of large models, and the importance of security compliance is emphasized in the Administrative Measures for Generative AI Services (Draft for Comments) issued by the Cyberspace Administration of China. In the face of AIGC content security requirements, Sun Songer believes that illegal information should be filtered and avoided from three levels: the input side, the model side, and the output side. To this end, New H3C Group has built a complete AIGC content security protection system, through Prompt review and rewriting, generated content filtering, redline knowledge base, etc., Lingxi Guardian can monitor user input and generated content, identify various violation risks, and comprehensively ensure content compliance, and the overall content risk can be reduced by nearly 80%.
In terms of large models empowering security, Sun Songer pointed out that large security models are triggering an "efficiency revolution" and bringing leaps and bounds to security operations. The Lingxi assistant built with large model technology is now fully embedded in security products, management platforms, and various O&M tools, and through the security enablement platform, the large model and business scenarios are connected, and all-round operational efficiency improvement is achieved in attack research and judgment, asset combing, compliance inspection, and intelligent prediction. For example, in the research and judgment of the collapse event, the research and judgment efficiency can be increased by 50% through AIGC-assisted analysis, and in the risk response and disposal, AIGC can directly give accurate disposal suggestions based on the training experience, and quickly generate a disposal report, so as to achieve twice the result with half the effort and liberate manpower from complicated basic work.
Create a new paradigm of scenario-based empowerment of scenario-based implementation of network security
From technology research and development to practical application, how to cross the gap between the laboratory and the industrial chain and achieve substantial value improvement in specific industries and scenarios is the focus of the current industry. In this regard, Sun Songer said that the security requirements of vertical industries are quite different, and it is necessary to establish scenario-based thinking to capture the actual security needs of different industries and build a new scenario-based paradigm based on the customer's business itself.
In terms of scenario-based exploration, New H3C Group is deeply engaged in key industry scenarios such as the Internet of Vehicles, digital government, and operators, and creates on-demand computing, network, and security integration solutions for multi-scenario services, and realizes the end-to-end business chain connection and security orchestration of network and security services by building a distributed security resource pool and leveraging the advantages of SRv6 elastic programmability, measurable performance, and isolation and guarantee, so as to ensure the safe and stable operation of data and services. In the field of IoV security, H3C has created IoV security solutions for vehicle-road collaboration, IoV operation and IoV talent training scenarios, so as to fully ensure the security of the IoV end-side, roadside and cloud. Focusing on the whole life cycle of data circulation, H3C has built an AI-based data security governance and operation platform, which empowers data classification and grading, data dynamic graphing and data security management with large model technology, so as to realize the sensibility, knowing, management and controllability of data circulation.
Comprehensive localization Establish a new benchmark for localization safety
In addition, New H3C Group has comprehensively deepened the security product system and localization layout, launched the industry's first flagship firewall product embedded with AIGC, built a security operation center, data security governance center and cloud security platform that fully support localization, set a new benchmark for localized security capabilities, and fully adapted 40+ security atomic capabilities to localization, including mainstream localized operating systems and software and hardware, to fully protect terminals, borders, applications and data security.
As a representative of new quality productivity, AIGC is the core engine of a new round of industrial transformation, and cyber security is the cornerstone of escorting AIGC's steady and long-term development. As a leading enterprise in the field of cyber security, New H3C Group will uphold the concept of "intensive and pragmatic cultivation and wisdom for the times", continue to promote the technological innovation and scenario-based application of "security × AI", and help all industries move forward on the road of digital transformation with the two-way empowerment of cyber security and AIGC.