Search
On March 4, Deng Zhonghan, a member of the National Committee of the Chinese People's Political Consultative Conference (CPPCC) and an academician at the Chinese Academy of Engineering, underscored the perils of data breaches, which can facilitate identity theft, incur property losses, and even pose significant threats to national security. He highlighted that the training of large models is heavily reliant on extensive datasets, and any tampering or injection of false information into these datasets could lead to deviations in model outputs, thereby eroding social trust. This practice, known as "data poisoning," involves attackers deliberately introducing incorrect or malicious samples into training data with the aim of disrupting the model's decision-making logic.
-
C114 Communication Network -
Communication Home
