At present, the digital economy has brought extensive and profound impacts on human production and life, but the consequent massive data security problems have become increasingly prominent. At the Zhongguancun Forum - Data Security Governance and Development Forum held today (April 26), experts and scholars at home and abroad carried out exchanges and discussions on the fields of intelligent networked vehicles and artificial intelligence to discuss the path of data security governance.
Zhongguancun Forum - Data Security Governance and Development Forum/Photo by reporters
In recent years, the data security of intelligent networked vehicles has become increasingly prominent. Taking the Beijing Economic and Technological Development Zone as an example, there are 838 connected cars and 367 intelligent networked intersections, adding more than 305TB of various types of data every day, including sensitive information such as license plates, human privacy behaviors, and vehicle tracks.
Xu Hongwei, executive deputy director of the Beijing High-level Autonomous Driving Demonstration Zone Office, introduced that the demonstration zone proposed the overall data-based governance for the first time. From the aspects of governance dimension and governance points, three systems are needed, one is the organizational system, the other is the institutional system, and the other is the technical system.
"In the key points of data governance, we put forward the security of the whole life cycle, that is, from the collection, transmission, storage, use, sharing, and finally how to destroy, from the vehicle end, the road end, the cloud, the vehicle transmission, the vehicle cloud transmission, the road cloud transmission, these aspects must be classified and classified. Xu Hongwei said that at the same time, communication protection and cloud protection from the end, according to the vehicle, road, and cloud, a technical system has been created from security governance and static desensitization. ”
Autonomous driving test/from the network
Since 2022, the Cyberspace Administration of Beijing Municipality has continuously carried out the review of annual reports on automotive data security management to guide car companies to carry out automotive data processing activities in compliance. Li Zhao, data protection officer of Audi China, said that in 2021, Audi China set up a data protection office, which adopts a very enhanced security and confidentiality agreement for all data during transmission, and at the same time adopts encryption measures to prevent data from being tampered with or leaked, so as to ensure the integrity of data.
"At the institutional level, we must ensure that the relevant management systems, such as the personal information protection policy or the data classification and grading system, meet the requirements of the latest national supervision, and at the technical level, we must also ensure that they are friendly during the life cycle of data processing. Li Zhao said.
Intelligent networked vehicles form a massive amount of data/from the network
With the rapid development of artificial intelligence, data security and privacy protection have also encountered major challenges. Artificial intelligence models rely on hundreds of millions of training data and reinforcement learning, how to ensure that personal privacy data and harmful data are filtered out in the process? Hu Songlin, a researcher at the Institute of Information Engineering of the Chinese Academy of Sciences, pointed out that while improving detection capabilities, security protection is increased in the whole process.
"How do we detect that this data is harmful, how do we detect that this data involves private data, how do we detect that this data conflicts with human values, these are all research points of our key technology. Hu Songlin said that it is also necessary to increase the ability to protect security, from the data stage to the pre-training stage, to the human values alignment stage, until the final reasoning stage, in fact, the current security is a series of interventions.
Artificial intelligence large model / from the network
In response to the value bias, privacy leakage, false information and other problems caused by the field of artificial intelligence, it is also possible to "defeat magic with magic". Sun Maosong, executive vice president of Tsinghua University's Institute of Artificial Intelligence, said that generative AI provides a new and more effective means of data security governance. For example, there are a lot of privacy issues in data today, but generative AI can be used to generate real-world data while avoiding privacy. This shows its power, which was not possible before without generative AI.
In addition, the current environment facing the management of artificial intelligence is very complex, some situations may not appear on the Internet, there is no kind of data, but it is predicted that this situation may happen, at this time to add certain conditions to the large model, generative artificial intelligence will generate some data of the scenarios that have not yet been encountered, so that human beings can make judgments in advance.
Reporter Wang Chenchen Editor Wang Chenchen Chief Editor Cheng Yan