Searcharxiv⌕ Search

arXiv subjects

Ronghao Cui

Publications and source records attributed to Ronghao Cui.

2 recordsLinked to original sources

A scaling law for numerical computation of fracture stress intensity factors

Numerical simulations of fracture propagation require accurate computation of stress intensity factors (SIFs) to determine whether a pre-existing fracture will propagate. Even though displacement solutions from well-established numerical methods generally converge with mesh refinement, the convergence behavior of SIF estimates remains less understood. In this work, we propose a scaling law to improve the performance of SIF extraction using two-point displacement correlation methods. This scaling law relates sampling distance from the fracture tip to mesh resolution. It is developed by balancing the asymptotic truncation error against the numerical discretization error. Three numerical discretization methods--standard quadratic finite elements, quadratic finite elements with quarter-point elements, and a multi-point stress approximation finite volume method--are employed to validate the proposed scaling law on fracture benchmark problems under pure-mode and mixed-mode loading. We find that the conventional nearest-tip sampling strategy fails to compute convergent SIF estimates. In contrast, the proposed scaling-based approach yields decreasing errors in the SIF estimates as the mesh is refined. Across all three discretization methods, the proposed scaling law captures the sampling locations required for convergent SIF extraction, suggesting a general numerical characteristic of near-tip SIF extraction.

math.NA↗

Towards Action Hijacking of Large Language Model-based Agent

Recently, applications powered by Large Language Models (LLMs) have made significant strides in tackling complex tasks. By harnessing the advanced reasoning capabilities and extensive knowledge embedded in LLMs, these applications can generate detailed action plans that are subsequently executed by external tools. Furthermore, the integration of retrieval-augmented generation (RAG) enhances performance by incorporating up-to-date, domain-specific knowledge into the planning and execution processes. This approach has seen widespread adoption across various sectors, including healthcare, finance, and software development. Meanwhile, there are also growing concerns regarding the security of LLM-based applications. Researchers have disclosed various attacks, represented by jailbreak and prompt injection, to hijack the output actions of these applications. Existing attacks mainly focus on crafting semantically harmful prompts, and their validity could diminish when security filters are employed. In this paper, we introduce AI$\mathbf{^2}$, a novel attack to manipulate the action plans of LLM-based applications. Different from existing solutions, the innovation of AI$\mathbf{^2}$ lies in leveraging the knowledge from the application's database to facilitate the construction of malicious but semantically-harmless prompts. To this end, it first collects action-aware knowledge from the victim application. Based on such knowledge, the attacker can generate misleading input, which can mislead the LLM to generate harmful action plans, while bypassing possible detection mechanisms easily. Our evaluations on three real-world applications demonstrate the effectiveness of AI$\mathbf{^2}$: it achieves an average attack success rate of 84.30\% with the best of 99.70\%. Besides, it gets an average bypass rate of 92.7\% against common safety filters and 59.45\% against dedicated defense.

cs.CR↗