Peixuan Li

Haodong Zhao, Wei Du, Fangqi Li et al.

Federated learning (FL) has enabled global model training on decentralized data in a privacy-preserving way by aggregating model updates. However, for many natural language processing (NLP) tasks that utilize pre-trained language models (PLMs) with large numbers of parameters, there are considerable communication costs associated with FL. Recently, prompt tuning, which tunes some soft prompts without modifying PLMs, has achieved excellent performance as a new learning paradigm. Therefore we want to combine the two methods and explore the effect of prompt tuning under FL. In this paper, we propose "FedPrompt" to study prompt tuning in a model split aggregation way using FL, and prove that split aggregation greatly reduces the communication cost, only 0.01% of the PLMs' parameters, with little decrease on accuracy both on IID and Non-IID data distribution. This improves the efficiency of FL method while also protecting the data privacy in prompt tuning. In addition, like PLMs, prompts are uploaded and downloaded between public platforms and personal users, so we try to figure out whether there is still a backdoor threat using only soft prompts in FL scenarios. We further conduct backdoor attacks by data poisoning on FedPrompt. Our experiments show that normal backdoor attack can not achieve a high attack success rate, proving the robustness of FedPrompt. We hope this work can promote the application of prompt in FL and raise the awareness of the possible security threats.

Peixuan Li, Danfeng Zhang

Noninterference offers a rigorous end-to-end guarantee for secure propagation of information. However, real-world systems almost always involve security requirements that change during program execution, making noninterference inapplicable. Prior works alleviate the limitation to some extent, but even for a veteran in information flow security, understanding the subtleties in the syntax and semantics of each policy is challenging, largely due to very different policy specification languages, and more fundamentally, semantic requirements of each policy. We take a top-down approach and present a novel information flow policy, called Dynamic Release, which allows information flow restrictions to downgrade and upgrade in arbitrary ways. Dynamic Release is formalized on a novel framework that, for the first time, allows us to compare and contrast various dynamic policies in the literature. We show that Dynamic Release generalizes declassification, erasure, delegation and revocation. Moreover, it is the only dynamic policy that is both applicable and correct on a benchmark of tests with dynamic policy.

Peixuan Li, Danfeng Zhang

This paper investigates a flow- and path-sensitive static information flow analysis. Compared with security type systems with fixed labels, it has been shown that flow-sensitive type systems accept more secure programs. We show that an information flow analysis with fixed labels can be both flow- and path-sensitive. The novel analysis has two major components: 1) a general-purpose program transformation that removes false dataflow dependencies in a program that confuse a fixed-label type system, and 2) a fixed-label type system that allows security types to depend on path conditions. We formally prove that the proposed analysis enforces a rigorous security property: noninterference. Moreover, we show that the analysis is strictly more precise than a classic flow-sensitive type system, and it allows sound control of information flow in the presence of mutable variables without resorting to run-time mechanisms.