paperbot · PL 论文追踪

全部论文 待读 收藏

P2 A Hybrid, Dynamic Logic for Hybrid-Dynamic Information Flow

一种面向混合动态信息流的混合动态逻辑
LICS Proceedings of the 33rd Annual ACM/IEEE Symposium on Logic in Computer Science2018匹配度 2/10 · 论文属于形式化方法与信息流安全,提出扩展微分动态逻辑的 dHL 来验证 CPS 混合动态信息流,并给出可靠性与电网案例,但未涉及 LLM/Agent、ML 系统静态分析、浮点程序分析或贝叶斯程序分析等用户核心方向,仅与“形式化验证”有宽泛关联。引用 23 inbox
Rose Bohrer, André Platzer
crossref: 2026-09-12

AI 速览

一句话:提出首个验证信息物理系统中混合动态信息流安全性的逻辑 dHL,并给出可靠证明演算与电网案例。

CPS 的信息可经离散网络和连续物理信道流动,形成混合动态信息流(HDIF),威胁隐私与物理安全。本文提出 dHL,首个在 CPS 混合动态模型中验证 HDIF 的逻辑。dHL 扩展微分动态逻辑(dL),加入混合逻辑特性以显式表示程序状态,并支持信息流论证所需的关系推理,可在攻击者连续观测时间与物理值的强攻击模型下保证安全。作者给出 Hilbert 风格证明演算并证明其可靠性,比较了 dHL 与 dL 的表达力。基于智能电网 FREEDM 模型,他们证明朴素模型存在未知信息流漏洞,并验证修订模型已修复,这是 HDIF 及一般混合动态模型上的首个信息流证明。

问题
CPS 中信息既经离散网络也经连续物理信道流动,形成 HDIF;在攻击者可连续观测时间和物理值的强攻击模型下,如何验证其信息流安全。
方法
提出 dHL,扩展微分动态逻辑 dL,加入混合逻辑特性与关系推理,并给出 Hilbert 风格证明演算,证明可靠性并分析表达力。
评估
在基于智能电网 FREEDM 的混合系统模型上展示 dHL,证明朴素模型存在此前未知的信息流漏洞,并验证修订模型可修复该漏洞。
原文摘要(Abstract)

Information-flow security is important to the safety and privacy of cyber-physical systems (CPSs) across many domains: information leakage can both violate user privacy and reveal vulnerabilities to physical attacks. CPSs face the challenge that information can flow both in discrete cyber channels and in continuous real-valued physical channels ranging from time to motion to electrical currents. We call these hybrid-dynamic information flows (HDIFs) and introduce dHL, the first logic for verifying HDIFs in hybrid-dynamical models of CPSs. Our logic extends differential dynamic logic (dL) for hybrid-dynamical systems with hybrid-logical features for explicit program state representation, supporting relational reasoning used for information flow arguments. By verifying HDIFs, we ensure security even under a strong attacker model wherein an attacker can observe time and physical values continuously. We present a Hilbert-style proof calculus for dHL, prove it sound, and compare the expressive power of dHL with dL. We develop a hybrid system model based on the smart electrical grid FREEDM, with which we showcase dHL. We prove that the naive model has a previously unknown information flow vulnerability, which we verify is resolved in a revised model. This is the first information flow proof both for HDIFs and for a hybrid-dynamical model in general.

链接与引用

DOI 原文 ·

BibTeX
@inproceedings{paperbot4346,
  title = {A Hybrid, Dynamic Logic for Hybrid-Dynamic Information Flow},
  author = {Rose Bohrer and André Platzer},
  booktitle = {LICS},
  year = {2018},
  doi = {10.1145/3209108.3209151}
}