Searcharxiv⌕ Search
Home/Archive/arXiv

arXiv · 2610.35451

Then there is prompt injection, a malicious email, ...: The Wild Wide West of OpenClaw Agentic System Security

Mostafijur Rahman Akhond·Mahmuda Khatun·Gias Uddin
View original source ↗Read source document ↗Export metadata ↓

Abstract

AI-powered personal assistant frameworks such as OpenClaw combine LLM reasoning, persistent memory, tool invocation, local execution, and web-based workflows. These capabilities enable useful automation, but they also create security risks when untrusted inputs, model reasoning, tools, memory, and deployment configuration interact without explicit trust boundaries. This paper presents an empirical study of reported security concerns in OpenClaw using two complementary data sources: Reddit discussions and security-tagged GitHub issues. We analyze more than 25K Reddit posts from 13 OpenClaw-related subreddits and identify 162 security-relevant posts with 1,787 associated comments. We further collect 358 security-tagged GitHub issues and retain 180 issues describing concrete security concerns. Using qualitative analysis and evidence-grounded CWE mapping, we identify 521 CWE instances across 59 distinct weakness types, concentrated in access-control failures, sensitive-data exposure, injection, and unsafe execution. We also extract 543 practitioner mitigation suggestions and consolidate them into seven recurring strategies. To complement the empirical analysis, we present an evidence-grounded illustrative reconstruction of a representative sensitive-data exposure pathway using dummy secret values. Our findings show that OpenClaw security issues are not isolated implementation bugs; they reflect recurring trust-boundary failures across the agent execution pipeline. We discuss how these weaknesses map to both traditional software-security categories and agent-specific failure mechanisms, and derive implications for safer agentic systems, including explicit trust boundaries, least-privilege tool access, execution containment, runtime oversight, and defense-in-depth deployment.

Explore related subjects

cs.CRcs.SE

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Mostafijur Rahman Akhond, Mahmuda Khatun, Gias Uddin. 2026-10-26. Then there is prompt injection, a malicious email, ...: The Wild Wide West of OpenClaw Agentic System Security. https://arxiv.org/abs/2610.35451

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

What's in a Package? Getting Visibility Into Dependencies Using Security-Sensitive API Calls

The Log4Shell vulnerability occurred because the JNDI (Java Naming and Directory Interface) lookup feature was used, even though it was not needed for log4j's core functionality. This case illustrates why developers may want visibility into the sensitive capabilities available through a dependency. The presence of a call alone does not establish that the call is unnecessary or risky. Security-sensitive APIs provide access to security-sensitive resources, such as filesystem, network, or process-related resources. If developers know which types of sensitive resources a dependency can access, they can control their dependencies at a finer grain, and they can consider an alternative dependency when a dependency makes unnecessary or suspicious security-sensitive API calls. However, prior work has not provided a documented, repeatable procedure for deriving an ecosystem-specific security-sensitive API list, so developers lack systematic visibility into the sensitive capabilities of their dependencies. The goal of this study is to aid developers in understanding sensitive resources accessible to their dependencies by identifying security-sensitive APIs in packages through an empirical study of Java packages. In this study, we present a novel and repeatable methodology to construct a security-sensitive API list for an ecosystem, and we apply it to Java. We then compare the security-sensitive API calls of functionally similar Java packages. The number of security-sensitive API calls (call level) ranges from 0 to 368 in one capability and from 0 to 429 in total among packages with similar functionality. These packages also differ in capability breadth, the number of the 15 capabilities that a package reaches. For example, capability breadth ranges from 0 to 10 in the "Dependency Injection" category and from 0 to 13 in the "Web Frameworks" category.

cs.CR↗
arXiv2026

WADBERT: Dual-Channel Web Attack Detection via Permutation-Invariant Parameter Modeling

Web attack detection is the first line of defense for securing web applications, designed to preemptively identify malicious activities. Deep learning-based approaches are increasingly popular for their advantages: automatically learning complex patterns and extracting semantic features from HTTP requests to achieve superior detection performance. However, existing methods are less effective at representing irregular HTTP requests and introduce unnecessary sensitivity to payload parameter order. To tackle this issue, in this paper, we propose WADBERT, a structure-aware dual-channel framework for web attack detection built on two complementary designs. First, we introduce \textit{Hybrid Granularity Embedding (HGE)}, which combines WordPiece embeddings with token-aligned character representations to capture fine-grained patterns in symbol-dense URLs and payload parameters. Second, we develop \textit{structure-aware feature extraction}, with separate branches for the sequential URL and the unordered collection of payload parameters. Each parameter is encoded independently, and position-free multi-head attention followed by mean pooling produces a payload representation invariant to parameter reordering. The URL and payload features are fused for request classification, while the learned parameter-level attention weights support inspection of potentially malicious parameters. Experiments on CSIC2010 and SR-BH2020 validate the effectiveness of our method: WADBERT achieves F1-scores of 99.64\% and 99.59\%, respectively, outperforming the evaluated state-of-the-art baselines.

cs.CR↗
arXiv2026

A Cross-Modal Prompt Injection Attack against Large Vision-Language Models with Image-Only Perturbation

Large vision-language models (LVLMs) have emerged as a powerful paradigm for multimodal intelligence, but their growing deployment also expands the attack surface of prompt injection. Despite this growing concern, existing attacks still suffer from a critical limitation: the injected prompt for one modality only steers the model's interpretation of that singular input. Alternatively, these attacks remain multimodal but fail to achieve cross-modal prompt perturbation. To bridge this gap, we introduce a novel cross-modal prompt injection attack CrossMPI, which can steer the model's interpretation of both textual and visual inputs via image-only prompt injection. Our design is underpinned by the following key breakthroughs. First, we turn the focus of the injected prompt perturbation optimization from the visual embedding space (typically with only $10^5$ parameters) to the model hidden state space (for multimodal information integration and with $10^7$ parameters). Then, two strategies are adopted to mitigate the optimization challenges posed by the larger parameter space. To constrain the optimized model parameter space, we introduce a layer selection strategy that identifies the layers most critical to multimodal integration. Interestingly, deviating from the past experience, our analysis reveals that the optimal layers for LVLM prompt perturbation reside in the middle of the model rather than the last. To constrain the image perturbation space, we propose a new distance-decremental perturbation budget assignment strategy that allocates budgets decrementally as the pixel distance to semantic-critical regions increases. Extensive experiments across multiple LVLMs and datasets show that our method significantly outperforms baseline approaches.

cs.CR↗