Sharingindividually contributed data is a cornerstone ofmodern artificial intelligence. For example, large languagemodels(LLMs)were trained on hundreds of billion tokensscraped from the internet.

The digital asset is a non-monetary asset in the form of data owned andcontrolled by individuals,enterprises, governmentsor variousorganizations . The digital assets include currently very popularcryptocurrencies and NFT tokens

在《联邦参与者贡献评估》报告中，刘泽雷博士会分享联邦学习中参与者贡献评估的意义、方法和应用，包括贡献评估的目的、基于Shapley Value的计算方法、与联邦学习聚合的结合、进行激励和知识贡献选择等应用。系统地阐述联邦学习中贡献评估的框架和运用。

In this talk we consider a defending problem on a network. In the model, thedefender holds a total defending resource of R, which can be distributed to thenodes of the network.

Cyber-Physical Systems (CPS) are the results of the inevitable merge ofcomputers with other domains of physical-world practices. A key challenge toCPS is software

隐私保护和拜占庭容错性是联邦学习系统中两个不可或缺的要求。尽管在隐私和拜占庭安全方面已经进行了广泛研究，但同时考虑这两个要求的解决方案仍然很少。这是由于在保护隐私和拜占庭容错算法之间进行调和的困难所致。

机密计算是一种新兴的安全技术，它利用可信执行环境对计算过程中的数据进行加密和保护，实现数据的“可用而不可见”。在我国数据安全技术创新的战略背景下，机密计算技术具有非常广阔应用前景。

Artificial intelligence-based speech recognition technology is now widelyused, and intelligent voice devices such as Amazon Echo and XiaoMi Xiaoaihave entered thousands of households. However, studies have shown thatspeech recognition systems are threatened by audio adversarial sampleattacks,

This talk will first illustrate how to use Al techniques to enhance cyber security ofvarious systems. There are several ways to apply Al to cyber security areas. This talkwill use prediction-based Al technics to enhance the total security of the V2X (Vehicle.to -Everything)communication system.

Aquila is the first practically usable verification system for production-scaleprogrammable data planes. lt addresses four challenges in building apractically usable verification: (1)specification complexity; (2) verificationscalability; (3) bug localization, and