arXiv subjects

Lipeng He

Publications and source records attributed to Lipeng He.

11 recordsLinked to original sources

arXiv2026

Defending against Adaptive Prompt Injection Attacks via Reasoning-enabled Task Alignment

Indirect prompt injection attacks hijack LLM-based agents by embedding malicious instructions in third-party data that the agent retrieves during task execution. Existing defenses report near-zero attack success rate on static benchmarks, yet recent adaptive evaluations show that these results collapse once the attacker is allowed to optimize against the deployed defense. In this work, we trace this collapse to two failure modes. First, existing defense methods are confined to recognizing specific attack patterns, rather than assessing whether the intent of every embedded instruction is relevant to the user task. Second, training-based defenses, which otherwise offer the strongest safety-utility trade-off, assemble their adversarial examples from a handful of hand-crafted templates, and the resulting defender fails to generalize outside that narrow strategy distribution. To address these gaps, we propose RETA, a training-based method that grounds defense decisions on the user tasks rather than attacker-controlled data. At each tool-output step, the defender undertakes chain-of-thought reasoning verifying that its actions are consistent with the user task. Leveraging red-teaming, a simulated attacker synthesizes adversarial training data and receives a dictionary-learning diversity reward, achieving broad coverage of injection-reformulation strategies. Together, these allow the defender to be optimized via multi-objective reinforcement learning and achieve better safety-utility trade-off. Across six black-box adaptive attacks, RETA keeps every per-attack ASR below 10%, with average ASR of 2.92% and 3.75% on the two target models, while preserving most utility under attack and on clean inputs.

cs.CR↗

arXiv2026

SoK: Colluding Adversaries in Machine Learning Pipelines

Machine learning (ML) models are susceptible to various security, privacy, and fairness risks. Adversaries with different characteristics (i.e., objectives, knowledge, and capabilities) can collude by executing one attack to amplify others. Existing work lacks a systematic framework to explore collusion among adversaries, and to study the implications of the adversaries' characteristics. We present a framework covering collusion (a) between train- and inference-time adversaries, and (b) among inference-time adversaries. Our framework accounts for factors enabling collusion between adversaries. We propose a guideline to conjecture about the potential for collusion using enabling factors. We use it to explain prior work, conjecture about unexplored collusions, and empirically validate five such cases. Finally, we discuss how adversaries' characteristics influence the potential for collusion.

cs.CR↗

arXiv2026

Beyond Similarity: Trustworthy Memory Search for Personal AI Agents

Personal AI agents increasingly rely on long-term memory to provide persistent personalization across sessions. However, existing memory pipelines are largely driven by semantic similarity: memory data close to the current query is retrieved and injected into the model context. This creates a critical trustworthiness gap, since a semantically related memory may still be contextually inappropriate, leading to threats such as cross-domain leakage, sycophancy, tool-call drift, or memory-induced jailbreaks. In this paper, we study memory search as a trust boundary in personal AI agents. We evaluate representative agentic memory frameworks, including A-Mem, Mem0, and MemOS, together with OpenClaw, a real-world personal-agent environment with persistent state and tool-use capability. Our results show that long-term memory is not merely a utility layer, but a durable control channel that can reshape how agents interpret tasks and execute actions, leaving them highly susceptible to the aforementioned threats. To mitigate these vulnerabilities, we propose MemGate, a lightweight and deployable memory plug-in for trustworthy memory search, with only 9M parameters and a 35.1MB footprint. MemGate is inserted between the vector memory store and the backbone LLM, requiring no LLM modification, memory-database rewriting, or inference-time LLM judge. It applies a query-conditioned neural gate to candidate memory representations, turning raw similarity search into task-conditioned memory admission. Across multiple mainstream memory frameworks, real-world agent settings, and diverse LLM backbones, MemGate reduces memory-induced threats while preserving long-term memory utility.

cs.AI↗

arXiv2026

Backdooring Bias in Large Language Models

Large language models (LLMs) are increasingly deployed in settings where inducing a bias toward a certain topic can have significant consequences, and backdoor attacks can be used to produce such models. Prior work on backdoor attacks has largely focused on a black-box threat model, with an adversary targeting the model builder's LLM. However, in the bias manipulation setting, the model builder themselves could be the adversary, warranting a white-box threat model where the attacker's ability to poison, and manipulate the poisoned data is substantially increased. Furthermore, despite growing research in semantically-triggered backdoors, most studies have limited themselves to syntactically-triggered attacks. Motivated by these limitations, we conduct an analysis consisting of over 1000 evaluations using higher poisoning ratios and greater data augmentation to gain a better understanding of the potential of syntactically- and semantically-triggered backdoor attacks in a white-box setting. In addition, we study whether two representative defense paradigms, model-intrinsic and model-extrinsic backdoor removal, are able to mitigate these attacks. Our analysis reveals numerous new findings. We discover that while both syntactically- and semantically-triggered attacks can effectively induce the target behaviour, and largely preserve utility, semantically-triggered attacks are generally more effective in inducing negative biases, while both backdoor types struggle with causing positive biases. Furthermore, while both defense types are able to mitigate these backdoors, they either result in a substantial drop in utility, or require high computational overhead.

cs.CR↗

arXiv2026

Understanding and Preserving Safety in Fine-Tuned LLMs

Fine-tuning is an essential and pervasive functionality for applying large language models (LLMs) to downstream tasks. However, it has the potential to substantially degrade safety alignment, e.g., by greatly increasing susceptibility to jailbreak attacks, even when the fine-tuning data is entirely harmless. Despite garnering growing attention in defense efforts during the fine-tuning stage, existing methods struggle with a persistent safety-utility dilemma: emphasizing safety compromises task performance, whereas prioritizing utility typically requires deep fine-tuning that inevitably leads to steep safety declination. In this work, we address this dilemma by shedding new light on the geometric interaction between safety- and utility-oriented gradients in safety-aligned LLMs. Through systematic empirical analysis, we uncover three key insights: (I) safety gradients lie in a low-rank subspace, while utility gradients span a broader high-dimensional space; (II) these subspaces are often negatively correlated, causing directional conflicts during fine-tuning; and (III) the dominant safety direction can be efficiently estimated from a single sample. Building upon these novel insights, we propose safety-preserving fine-tuning (SPF), a lightweight approach that explicitly removes gradient components conflicting with the low-rank safety subspace. Theoretically, we show that SPF guarantees utility convergence while bounding safety drift. Empirically, SPF consistently maintains downstream task performance and recovers nearly all pre-trained safety alignment, even under adversarial fine-tuning scenarios. Furthermore, SPF exhibits robust resistance to both deep fine-tuning and dynamic jailbreak attacks. Together, our findings provide new mechanistic understanding and practical guidance toward always-aligned LLM fine-tuning.

cs.LG↗

arXiv2026

Safety at One Shot: Patching Fine-Tuned LLMs with A Single Instance

Fine-tuning safety-aligned large language models (LLMs) can substantially compromise their safety. Previous approaches require many safety samples or calibration sets, which not only incur significant computational overhead during realignment but also lead to noticeable degradation in model utility. Contrary to this belief, we show that safety alignment can be fully recovered with only a single safety example, without sacrificing utility and at minimal cost. Remarkably, this recovery is effective regardless of the number of harmful examples used in fine-tuning or the size of the underlying model, and convergence is achieved within just a few epochs. Furthermore, we uncover the low-rank structure of the safety gradient, which explains why such efficient correction is possible. We validate our findings across five safety-aligned LLMs and multiple datasets, demonstrating the generality of our approach.

cs.LG↗

arXiv2025

Locket: Robust Feature-Locking Technique for Language Models

Chatbot service providers (e.g., OpenAI) rely on tiered subscription plans to generate revenue, offering black-box access to basic models for free users and advanced models to paying subscribers. However, this approach is unprofitable and inflexible. A pay-to-unlock scheme for premium features (e.g., math, coding) offers a more sustainable alternative. Enabling such a scheme requires a feature-locking technique (FLoTE) that is (i) effective in refusing locked features, (ii) utility-preserving for unlocked features, (iii) robust against evasion or unauthorized credential sharing, and (iv) scalable to multiple features and clients. Existing FLoTEs (e.g., password-locked models) fail to meet these criteria. To fill this gap, we present Locket, the first robust and scalable FLoTE to enable pay-to-unlock schemes. We develop a framework for adversarial training and merging of feature-locking adapters, which enables Locket to selectively disable specific features of a model. Evaluation shows that Locket is effective ($100$% refusal rate), utility-preserving ($\leq 7$% utility degradation), robust ($\leq 5$% attack success rate), and scalable to multiple features and clients.

cs.CR↗

arXiv2025

StructEval: Benchmarking LLMs' Capabilities to Generate Structural Outputs

As Large Language Models (LLMs) become integral to software development workflows, their ability to generate structured outputs has become critically important. We introduce StructEval, a comprehensive benchmark for evaluating LLMs' capabilities in producing both non-renderable (JSON, YAML, CSV) and renderable (HTML, React, SVG) structured formats. Unlike prior benchmarks, StructEval systematically evaluates structural fidelity across diverse formats through two paradigms: 1) generation tasks, producing structured output from natural language prompts, and \textbf{2)} conversion tasks, translating between structured formats. Our benchmark encompasses 18 formats and 44 types of task, with novel metrics for format adherence and structural correctness. Results reveal significant performance gaps-even state-of-the-art models like o1-mini achieve only 75.58 average score, with open-source alternatives lagging approximately 10 points behind. We find generation tasks more challenging than conversion tasks, and producing correct visual content more difficult than generating text-only structures.

cs.SE↗

arXiv2025

Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense

Large Language Models (LLMs) have showcased remarkable capabilities across various domains. Accompanying the evolving capabilities and expanding deployment scenarios of LLMs, their deployment challenges escalate due to their sheer scale and the advanced yet complex activation designs prevalent in notable model series, such as Llama, Gemma, Mistral. These challenges have become particularly pronounced in resource-constrained deployment scenarios, where mitigating inference bottlenecks is imperative. Among various recent efforts, activation approximation has emerged as a promising avenue for pursuing inference efficiency, sometimes considered indispensable in applications such as private inference. Despite achieving substantial speedups with minimal impact on utility, even appearing sound and practical for real-world deployment, the safety implications of activation approximations remain unclear. In this work, we fill this critical gap in LLM safety by conducting the first systematic safety evaluation of activation approximations. Our safety vetting spans seven state-of-the-art techniques across three popular categories (activation polynomialization, activation sparsification, and activation quantization), revealing consistent safety degradation across ten safety-aligned LLMs. To overcome the hurdle of devising a unified defense accounting for diverse activation approximation methods, we perform an in-depth analysis of their shared error patterns and uncover three key findings. We propose QuadA, a novel safety enhancement method tailored to mitigate the safety compromises introduced by activation approximations. Extensive experiments and ablation studies corroborate QuadA's effectiveness in enhancing the safety capabilities of LLMs after activation approximations.

cs.CR↗

arXiv2024

A Comparative Examination of Network and Contract-Based Blockchain Storage Solutions for Decentralized Applications

Decentralized applications (DApps), which are innovative blockchain-powered software systems designed to serve as the fundamental building blocks for the next generation of Internet services, have witnessed exponential growth in recent years. This paper thoroughly compares and analyzes two blockchain-based decentralized storage networks (DSNs), which are crucial foundations for DApp and blockchain ecosystems. The study examines their respective mechanisms for data persistence, strategies for enforcing data retention, and token economics. In addition to delving into technical details, the suitability of each storage solution for decentralized application development is assessed, taking into consideration network performance, storage costs, and existing use cases. By evaluating these factors, the paper aims to provide insights into the effectiveness of these technologies in supporting the desirable properties of truly decentralized blockchain applications. In conclusion, the findings of this research are discussed and synthesized, offering valuable perspectives on the capabilities of these technologies. It sheds light on their potential to facilitate the development of DApps and provides an understanding of the ongoing trends in blockchain development.

cs.NI↗

arXiv2024

LookAhead: Preventing DeFi Attacks via Unveiling Adversarial Contracts

The exploitation of smart contract vulnerabilities in Decentralized Finance (DeFi) has resulted in financial losses exceeding 3 billion US dollars. Existing defense mechanisms primarily focus on detecting and reacting to adversarial transactions executed by attackers that target victim contracts. However, with the emergence of private transaction pools where transactions are sent directly to miners without first appearing in public mempools, current detection tools face significant challenges in identifying attack activities effectively. Based on the fact that most attack logic rely on deploying intermediate smart contracts as supporting components to the exploitation of victim contracts, novel detection methods have been proposed that focus on identifying these adversarial contracts instead of adversarial transactions. However, previous state-of-the-art approaches in this direction have failed to produce results satisfactory enough for real-world deployment. In this paper, we propose LookAhead, a new framework for detecting DeFi attacks via unveiling adversarial contracts. LookAhead leverages common attack patterns, code semantics and intrinsic characteristics found in adversarial smart contracts to train Machine Learning (ML)-based classifiers that can effectively distinguish adversarial contracts from benign ones and make timely predictions of different types of potential attacks. Experiments on our labeled datasets show that LookAhead achieves an F1-score as high as 0.8966, which represents an improvement of over 44.4% compared to the previous state-of-the-art solution, with a False Positive Rate (FPR) at only 0.16%.

cs.CR↗