arXiv subjects

Libo Chen

Publications and source records attributed to Libo Chen.

10 recordsLinked to original sources

arXiv2026

A Plug-in Interpretation of Conditioning in Score-Based Diffusion Models

We propose a conditioning mechanism for diffusion models based on multi-speed joint diffusion of the target and the condition. The mechanism learns an unconditional joint score network and enforces conditioning at inference via a plug-in correction term. The plug-in term separates the conditioning contribution from the learned unconditional dynamics, offering a transparent view of how the condition steers generation of the target distribution. Building on this, we derive explicit conditional reverse-time SDEs and approximate probability-flow ODEs, enabling principled and directly comparable conditional samplers. To reduce the induced ODE--SDE discrepancy, we introduce a log-Fokker--Planck residual regularization that improves ODE sampling quality. Experiments on conditional image generation tasks demonstrate competitive performance and support the effectiveness of the plug-in conditioning view. Additional ODE--SDE comparison experiments show that the log-Fokker--Planck residual regularization improves deterministic ODE sampling.

cs.CV↗

arXiv2025

Parasites in the Toolchain: A Large-Scale Analysis of Attacks on the MCP Ecosystem

Large language models(LLMs) are increasingly integrated with external systems through the Model Context Protocol(MCP),which standardizes tool invocation and has rapidly become a backbone for LLM-powered applications. While this paradigm enhances functionality,it also introduces a fundamental security shift:LLMs transition from passive information processors to autonomous orchestrators of task-oriented toolchains,expanding the attack surface,elevating adversarial goals from manipulating single outputs to hijacking entire execution flows. In this paper,we identify and characterize a systematic privacy-leakage attack pattern,termed Parasitic Toolchain Attacks,instantiated as MCP Unintended Privacy Disclosure(MCP-UPD). These attacks require no direct victim interaction;instead,adversaries embed malicious instructions into external data sources that LLMs access during legitimate tasks. Unlike traditional prompt injection and tool poisoning attacks,our attack targets the interconnected toolchain itself,assembling multiple legitimate tools into a coordinated workflow whose combined behavior accomplishes malicious objectives. In MCP-UPD,the malicious logic infiltrates the toolchain and unfolds in three phases:Parasitic Ingestion,Privacy Collection,and Privacy Disclosure,culminating in stealthy exfiltration of private data. Our root cause analysis reveals that MCP lacks both context-tool isolation and least-privilege enforcement,enabling adversarial instructions to propagate unchecked into sensitive tool invocations. To assess the severity,we design MCP-SEC and conduct the first large-scale security census of the MCP ecosystem,analyzing 12230 tools across 1360 servers. Our findings show that the MCP ecosystem is rife with real-world exploitable gadgets and diverse attack methods,underscoring systemic risks in MCP platforms and the urgent need for defense mechanisms in LLM-integrated environments.

cs.CR↗

arXiv2025

A.S.E: A Repository-Level Benchmark for Evaluating Security in AI-Generated Code

The increasing adoption of large language models (LLMs) in software engineering necessitates rigorous security evaluation of their generated code. However, existing benchmarks often lack relevance to real-world AI-assisted programming scenarios, making them inadequate for assessing the practical security risks associated with AI-generated code in production environments. To address this gap, we introduce A.S.E (AI Code Generation Security Evaluation), a repository-level evaluation benchmark designed to closely mirror real-world AI programming tasks, offering a comprehensive and reliable framework for assessing the security of AI-generated code. Our evaluation of leading LLMs on A.S.E reveals several key findings. In particular, current LLMs still struggle with secure coding. The complexity in repository-level scenarios presents challenges for LLMs that typically perform well on snippet-level tasks. Moreover, a larger reasoning budget does not necessarily lead to better code generation. These observations offer valuable insights into the current state of AI code generation and help developers identify the most suitable models for practical tasks. They also lay the groundwork for refining LLMs to generate secure and efficient code in real-world applications.

cs.SE↗

arXiv2024

Preparing Spin Squeezed States via Adaptive Genetic Algorithm

We introduce a novel strategy employing an adaptive genetic algorithm (GA) for iterative optimization of control sequences to generate quantum nonclassical states. Its efficacy is demonstrated by preparing spin-squeezed states in an open collective spin model governed by a linear control field. Inspired by Darwinian evolution, the algorithm iteratively refines control sequences using crossover, mutation, and elimination strategies, starting from a coherent spin state within a dissipative and dephasing environment. We rigorously benchmark our method against constant control protocols and reinforcement learning, demonstrating competitive and robust performance. Furthermore, we showcase the GA's versatility by directly optimizing for metrologically relevant squeezing, achieving scalable performance, even in the presence of dissipation and thermal noise. The proposed strategy demonstrates a high state-preparation fidelity, exceeding 0.99, and provides a long time window for maintaining the spin squeezed state, even under dissipative conditions. We discuss feasible experimental implementations and potential extensions to alternative quantum systems, and the adaptability of the GA module. This research establishes the foundation for utilizing GA-like strategies in controlling quantum systems and achieving desired nonclassical states.

quant-ph↗

arXiv2024

An Intrinsically Knowledge-Transferring Developmental Spiking Neural Network for Tactile Classification

Gradient descent computed by backpropagation (BP) is a widely used learning method for training artificial neural networks but has several limitations: it is computationally demanding, requires frequent manual tuning of the network architecture, and is prone to catastrophic forgetting when learning incrementally. To address these issues, we introduce a brain-mimetic developmental spiking neural network (BDNN) that mimics the postnatal development of neural circuits. We validate its performance through a neuromorphic tactile system capable of learning to recognize objects through grasping. Unlike traditional BP-based methods, BDNN exhibits strong knowledge transfer, supporting efficient incremental learning of new tactile information. It requires no hyperparameter tuning and dynamically adapts to incoming data. Moreover, compared to the BP-based counterpart, it achieves classification accuracy on par with BP while learning over ten times faster in ideal conditions and up to two or three orders of magnitude faster in practical settings. These features make BDNN well-suited for fast data processing on edge devices.

eess.SP↗

arXiv2024

Seismic fragility curves fitting revisited: ordinal regression models and their generalization

This study revisits the modeling of seismic fragility curves by applying ordinal regression models, offering an alternative to the commonly used log-normal distribution function. It compares various ordinal regression approaches, including Cumulative, Sequential, and Adjacent Category models, along with extensions that account for category-specific effects and variance heterogeneity. The methodologies are applied to bridge damage data from the 2008 Wenchuan earthquake, using both frequentist and Bayesian inference methods, with model diagnostics conducted using surrogate residuals. The analysis examines eleven models, from basic forms to those incorporating heteroscedastic extensions and category-specific effects. Based on leave-one-out cross-validation, the Sequential model with category-specific effects performs well compared to traditional Cumulative probit models. The results indicate differences in damage probability predictions between the models, suggesting the potential for more flexible fragility curve modeling techniques to improve seismic risk assessments. This study highlights the importance of continued evaluation of existing methods to enhance the predictive accuracy and applicability of seismic fragility models in performance-based earthquake engineering.

stat.AP↗

arXiv2024

SynGhost: Invisible and Universal Task-agnostic Backdoor Attack via Syntactic Transfer

Although pre-training achieves remarkable performance, it suffers from task-agnostic backdoor attacks due to vulnerabilities in data and training mechanisms. These attacks can transfer backdoors to various downstream tasks. In this paper, we introduce $\mathtt{maxEntropy}$, an entropy-based poisoning filter that mitigates such risks. To overcome the limitations of manual target setting and explicit triggers, we propose $\mathtt{SynGhost}$, an invisible and universal task-agnostic backdoor attack via syntactic transfer, further exposing vulnerabilities in pre-trained language models (PLMs). Specifically, $\mathtt{SynGhost}$ injects multiple syntactic backdoors into the pre-training space through corpus poisoning, while preserving the PLM's pre-training capabilities. Second, $\mathtt{SynGhost}$ adaptively selects optimal targets based on contrastive learning, creating a uniform distribution in the pre-training space. To identify syntactic differences, we also introduce an awareness module to minimize interference between backdoors. Experiments show that $\mathtt{SynGhost}$ poses significant threats and can transfer to various downstream tasks. Furthermore, $\mathtt{SynGhost}$ resists defenses based on perplexity, fine-pruning, and $\mathtt{maxEntropy}$. The code is available at https://github.com/Zhou-CyberSecurity-AI/SynGhost.

cs.CR↗

arXiv2022

Improved probabilistic seismic demand-intensity relationship: heteroskedastic approachs

As an integral part of assessing the seismic performance of structures, the probabilistic seismic demand-intensity relationship has been widely studied. In this study, the phenomenon of heteroscedasticity in probabilistic seismic demand models was systematically investigated. A brief review of the definition, diagnosis, and conventional treatment of heteroscedasticity is presented herein, and based on that, two more generalized methods for both univariate and multivariate cases are proposed. For a typical four-span simply supported girder bridge, a series of nonlinear time history analyses were performed through multiple stripe analysis to determine its seismic demand-intensity that can be employed as a sample set. For both univariate and multivariate cases, probabilistic seismic demand models were developed based on the two aforementioned methods under the Bayesian regression framework, and the fitted results were compared and analyzed with the conventional models using linear regression approaches. In the presence of probabilistic seismic demand considering heteroscedasticity, the patterns of non-constant variance or covariance can be characterized effectively, and a better-calibrated prediction region than that of homoscedastic models can be provided. The causes of the heteroscedasticity phenomenon and subsequent solutions are thoroughly discussed. The analysis procedures can be further embedded in seismic fragility and risk assessment, thus providing a more accurate basis for aseismic decision-making.

stat.AP↗

arXiv2018

Improving quantum parameter estimation by monitoring quantum trajectories

Quantum-enhanced parameter estimation has widespread applications in many fields. An important issue is to protect the estimation precision against the noise-induced decoherence. Here we develop a general theoretical framework for improving the precision for estimating an arbitrary parameter by monitoring the noise-induced quantum trajectorie (MQT) and establish its connections to the purification-based approach to quantum parameter estimation. MQT can be achieved in two ways: (i) Any quantum trajectories can be monitored by directly monitoring the environment, which is experimentally challenging for realistic noises; (ii) Certain quantum trajectories can also be monitored by frequently measuring the quantum probe alone via ancilla-assisted encoding and error detection. This establishes an interesting connection between MQT and the full quantum error correction protocol. Application of MQT to estimate the level splitting and decoherence rate of a spin-1/2 under typical decoherence channels demonstrate that it can avoid the long-time exponential loss of the estimation precision and, in special cases, recover the Heisenberg scaling.

quant-ph↗

arXiv2014

Optomechanical steady state entanglement induced by electrical interaction

We propose a scheme for generating remote continuous steady-state entanglement of output light leaked from optomechanical system, in which two mechanical oscillators are coupled through long-range Coulomb interaction. we show that the entanglement of output light is affected by the detuning and the strength of the Coulomb interaction. We also demonstrate that two movable mirrors and two light beams can be entangled in the steady state. We suggest an experimental readout scheme to fully verify the characteristic of entangled state.

quant-ph↗