搜索结果：Injection

We present a reproducible failure mode of safety training in RAG-based LLM recommendation -- the Injection Paradox -- in which prompt injections embedded in retrieved documents backfire against the attacker, suppressing the target brand below the injection-free baseline. In safety-trained Claude models, documents containing prompt injections suffer a sharp drop in recommendation rate, and this suppression propagates beyond the injected document to unmodified documents of the same brand. In Claude Opus 4.6, the target brand drops from a 54% baseline to zero top-2 recommendations across all 50 trials, even though only 1 of 4 brand documents in the corpus contains an injection. The directional pattern is reproduced in counterfactual experiments and across three brands. A contrasting result across the GPT models tested, where the same injection instead increases recommendations, suggests model-family differences in how injection-like context affects recommendation behavior. These findings raise the technical possibility of a reverse-attack scenario in which an adversary embeds injections in a competitor's documents to suppress the competitor's brand via safety-sensitive model behavior.

Modern agentic systems transform LLMs from session-bounded assistants into stateful systems that persist and evolve shared world state across sessions through memories, filesystems, tools, and other long-lived contextual artifacts. This shift fundamentally expands the attack surface of prompt injection. However, prior works on prompt injection have largely focused on model-level threats within a single session, overlooking how cross-session persistent system state fundamentally changes the system-level risk of agentic systems. Inspired by stored cross-site scripting in web systems, we introduce cross-session stored prompt injection, where a successful injection can persist within agentic system state and silently influence future executions long after the original attacker interaction has ended. To systematically study this threat, we formalize stored prompt injection and develop a taxonomy of how adversarial content persists and affects agentic systems across sessions. We further develop a benchmark and sandbox toolkit to evaluate the risks of stored prompt injection, enabling quantitative analysis of attack success across different models, attack goals, and persistence channels.

LLMs are vulnerable to prompt injection attacks. However, this vulnerability has been primarily demonstrated conceptually in academic studies or through a few anecdotal case studies. Its prevalence and impact in real-world LLM-based applications are largely unexplored. In this work, we present the first systematic study of prompt-injection attacks in a widely used application: LLM-based resume screening. Our analysis is based on approximately 200K real-world resumes collected over multiple years by hireEZ. We first design tailored methods to detect prompt injection in resumes. Manual validation on a small-scale dataset demonstrates that our detectors achieve high precision and outperform state-of-the-art general-purpose detectors. We then apply our detector to the full resume dataset and conduct a comprehensive measurement study of real-world prompt injection attacks. Our analysis reveals several intriguing findings: approximately 1% of resumes contain hidden prompt injections; the prevalence of such injected resumes has increased noticeably over the past one to two years; and more than 90% of injected prompts do not use explicit instructions. These results provide the first eviden

As large language models (LLMs) become integrated into various sensitive applications, prompt injection, the use of prompting to induce harmful behaviors from LLMs, poses an ever increasing risk. Prompt injection attacks can cause LLMs to leak sensitive data, spread misinformation, and exhibit harmful behaviors. To defend against these attacks, we propose CourtGuard, a locally-runnable, multiagent prompt injection classifier. In it, prompts are evaluated in a court-like multiagent LLM system, where a "defense attorney" model argues the prompt is benign, a "prosecution attorney" model argues the prompt is a prompt injection, and a "judge" model gives the final classification. CourtGuard has a lower false positive rate than the Direct Detector, an LLM as-a-judge. However, CourtGuard is generally a worse prompt injection detector. Nevertheless, this lower false positive rate highlights the importance of considering both adversarial and benign scenarios for the classification of a prompt. Additionally, the relative performance of CourtGuard in comparison to other prompt injection classifiers advances the use of multiagent systems as a defense against prompt injection attacks. The imple

Existing red-teaming studies on GUI agents have important limitations. Adversarial perturbations typically require white-box access, which is unavailable for commercial systems, while prompt injection is increasingly mitigated by stronger safety alignment. To study robustness under a more practical threat model, we propose Semantic-level UI Element Injection, a red-teaming setting that overlays safety-aligned and harmless UI elements onto screenshots to misdirect the agent's visual grounding. Our method uses a modular Editor-Overlapper-Victim pipeline and an iterative search procedure that samples multiple candidate edits, keeps the best cumulative overlay, and adapts future prompt strategies based on previous failures. Across five victim models, our optimized attacks improve attack success rate by up to 4.4x over random injection on the strongest victims. Moreover, elements optimized on one source model transfer effectively to other target models, indicating model-agnostic vulnerabilities. After the first successful attack, the victim still clicks the attacker-controlled element in more than 15% of later independent trials, versus below 1% for random injection, showing that the in

Web application firewall (WAF) plays an integral role nowadays to protect web applications from various malicious injection attacks such as SQL injection, XML injection, and PHP injection, to name a few. However, given the evolving sophistication of injection attacks and the increasing complexity of tuning a WAF, it is challenging to ensure that the WAF is free of injection vulnerabilities such that it will block all malicious injection attacks without wrongly affecting the legitimate message. Automatically testing the WAF is, therefore, a timely and essential task. In this paper, we propose DaNuoYi, an automatic injection testing tool that simultaneously generates test inputs for multiple types of injection attacks on a WAF. Our basic idea derives from the cross-lingual translation in the natural language processing domain. In particular, test inputs for different types of injection attacks are syntactically different but may be semantically similar. Sharing semantic knowledge across multiple programming languages can thus stimulate the generation of more sophisticated test inputs and discovering injection vulnerabilities of the WAF that are otherwise difficult to find. To this en

Current open-source prompt-injection detectors converge on two architectural choices: regular-expression pattern matching and fine-tuned transformer classifiers. Both share failure modes that recent work has made concrete. Regular expressions miss paraphrased attacks. Fine-tuned classifiers are vulnerable to adaptive adversaries: a 2025 NAACL Findings study reported that eight published indirect-injection defenses were bypassed with greater than fifty percent attack success rates under adaptive attacks. This work proposes seven detection techniques that each port a specific mechanism from a discipline outside large-language-model security: forensic linguistics, materials-science fatigue analysis, deception technology from network security, local-sequence alignment from bioinformatics, mechanism design from economics, spectral signal analysis from epidemiology, and taint tracking from compiler theory. Three of the seven techniques are implemented in the prompt-shield v0.4.1 release (Apache 2.0) and evaluated in a four-configuration ablation across six datasets including deepset/prompt-injections, NotInject, LLMail-Inject, AgentHarm, and AgentDojo. The local-alignment detector lifts

Prompt injection is a critical vulnerability in LLM agents, yet the strongest methods still rely on human red-teamers and hand-crafted prompts. Adapting automated jailbreak optimizers does not close this gap: jailbreaks shape models toward generic compliance, while prompt injection requires emitting specific tool calls with correct parameters. The success signal is binary, and randomly sampled suffixes almost never trigger it, so standard optimizers have no gradient to follow. We present AutoInject, a black-box reinforcement learning (RL) framework that learns adversarial suffixes for prompt injection. A learned comparison-based reward scores each candidate against the best suffix seen so far, turning the binary signal into a dense reward suitable for RL optimization. The framework supports both online query-based attacks and offline-trained transferable suffixes that need no utility access at deployment, and incorporates a utility objective when task-completion feedback is available. On AgentDojo, AutoInject outperforms template attacks, GCG, TAP, and adaptive attack across production models, with statistically significant improvements under McNemar's test with p<0.05. Suffixes

We theoretically investigate optical injection processes, including one- and two-photon carrier injection and two-color coherent current injection, in twisted bilayer graphene with moderate angles. The electronic states are described by a continuum model, and the spectra of injection coefficients are numerically calculated for different chemical potentials and twist angles, where the transitions between different bands are understood by the electron energy resolved injection coefficients. The comparison with the injection in monolayer graphene shows the significance of the interlayer coupling in the injection processes. For undoped twisted bilayer graphene, all spectra of injection coefficients can be divided into three energy regimes, which vary with the twist angle. For very low photon energies in the linear dispersion regime, the injection is similar to graphene with a renormalized Fermi velocity determined by the twist angle; for very high photon energies where the interlayer coupling is negligible, the injection is the same as that of graphene; and in the middle regime around the transition energy of the Van Hove singularity, the injection shows fruitful fine structures. Furth

Prompt injection attacks can compromise the security and stability of critical systems, from infrastructure to large web applications. This work curates and augments a prompt injection dataset based on the HackAPrompt Playground Submissions corpus and trains several classifiers, including LSTM, feed forward neural networks, Random Forest, and Naive Bayes, to detect malicious prompts in LLM integrated web applications. The proposed approach improves prompt injection detection and mitigation, helping protect targeted applications and systems.

Single-mode semiconductor lasers subject to optical injection have been shown to trigger a wide range of dynamical behavior from injection locking to chaos. Multi-wavelength lasers add even more degrees of freedom and complexity to the dynamical repertoire potentially unlocking new functionalities for applications ranging from THz generation and processing to all-optical memories. In particular, leveraging the inherent mode coupling in multi-wavelength lasers, spectral multiplication over a THz range of an injected optical signal has been shown. While most of the research on optical injection has been focused on single-mode semiconductor lasers, the dynamical behavior of multi-wavelength lasers, particularly when subjected to injection of amplitude-modulated signals remains vastly unexplored. In this work, we numerically and experimentally investigate the response of an on-chip dual-wavelength laser subject to the optical injection of a single-sideband signal around one of the modes of the laser. Our findings show an asymmetric power evolution of the sidebands appearing around both the injected and un-injected modes with respect to the modulation frequency. The power and bandwidth

Large language models (LLMs) are increasingly used in interactive and retrieval-augmented systems, but they remain vulnerable to prompt injection attacks, where injected secondary prompts force the model to deviate from the user's instructions to execute a potentially malicious task defined by the adversary. Recent work shows that ML models trained on activation shifts from LLMs' hidden layers can detect such drift. In this paper, we demonstrate that these detectors are not robust to adaptive adversaries. We propose a multi-probe evasion attack that appends an adversarially optimised suffix to poisoned inputs, jointly optimising a universal suffix to simultaneously fool all layer-wise drift detectors while preserving the effectiveness of the underlying injection. Using a modified Greedy Coordinate Gradient (GCG) approach, we generate universal suffixes that make prompt injections consistently evasive across multiple probes simultaneously. On Phi-3 3.8B and Llama-3 8B, a single suffix achieves attack success rates of 93.91% and 99.63% in successfully evading all detectors simultaneously. These results show that activation-based task drift detectors are highly vulnerable to adaptive

Large language models (LLMs) are typically personalized via prompt engineering or parameter-efficient fine-tuning such as LoRA. However, writing style can be difficult to distill into a single prompt, and LoRA fine-tuning requires computationally intensive training and infrastructure. We investigate a possible lightweight alternative: steering a frozen LLM with n-gram style priors injected in logit space at decoding time. We train an n-gram model on stylistically distinct corpora -- including Don Quixote, CNN/DailyMail news headlines, and arXiv abstracts -- constructing an interpolated 1-to-3-gram prior over next-token probabilities. During generation we modify the LLM's logits by adding a weighted sum of style log-probabilities from each n-gram order that matches the current context, scaled by a control parameter lambda in [0, 1]. We sweep lambda and style corpora and report style perplexity under the n-gram model, base-model perplexity as a proxy for fluency, Jensen-Shannon (JS) divergence between the original and steered token distributions, and token-overlap statistics. On TinyLlama-1.1B we identify a single narrow regime (for the Don Quixote corpus at lambda=0.1) where style p

Large Language Models (LLMs) excel in processing and generating human language, powered by their ability to interpret and follow instructions. However, their capabilities can be exploited through prompt injection attacks. These attacks manipulate LLM-integrated applications into producing responses aligned with the attacker's injected content, deviating from the user's actual requests. The substantial risks posed by these attacks underscore the need for a thorough understanding of the threats. Yet, research in this area faces challenges due to the lack of a unified goal for such attacks and their reliance on manually crafted prompts, complicating comprehensive assessments of prompt injection robustness. We introduce a unified framework for understanding the objectives of prompt injection attacks and present an automated gradient-based method for generating highly effective and universal prompt injection data, even in the face of defensive measures. With only five training samples (0.3% relative to the test data), our attack can achieve superior performance compared with baselines. Our findings emphasize the importance of gradient-based testing, which can avoid overestimation of rob

Multi-phase injection-locked ring oscillators (MP-ILROs) are widely used for multi-phase clock generation, with their phase accuracy primarily determined by the inherent accuracy of the oscillator itself, due to the suppression of input signal errors. However, a quantitative analysis of the oscillator's sensitivity to input errors remains largely unexplored. This paper presents a phasor-based analysis of injection locking, revealing that the phase error sensitivity is influenced by factors such as injection strength and the free-running frequency of the oscillator. Simulation results align closely with theoretical calculations, validating the effectiveness of the proposed method.

Eliminating warpage in injection molded polymeric parts is one of the most important problems in the injection molding industry today. This situation is critical in geometries that are particularly susceptible to warping due to their geometric features, and this occurs with topologies of great length and slenderness with high changes in thickness. These features are, in these special geometries, impossible to manufacture with traditional technologies to meet the dimensional and sustainable requirements of the industry. This paper presents an innovative green conformal cooling system that is specifically designed for parts with slender geometric shapes that are highly susceptible to warping. Additionally, the work presented by the authors investigates the importance of using highly conductive inserts made of steel alloys in combination with the use of additively manufactured conformal channels for reducing influential parameters, such as warpage, cooling time, and residual stresses in the complex manufacturing of long and slender parts. The results of this real industrial case study indicated that the use of conformal cooling layouts decreased the cycle time by 175.1 s 66% below the

We examine the circular, self-similar expansion of frictional rupture due to fluid injected at a constant rate. Fluid migrates within a thin permeable layer parallel to and containing the fault plane. When the Poisson ratio $ν=0$, self-similarity of the fluid pressure implies fault slip also evolves in an axisymmetric, self-similar manner, reducing the three-dimensional problem for the evolution of fault slip to a single self-similar dimension. The rupture radius grows as $λ\sqrt{4α_{hy} t}$, where $t$ is time since the start of injection and $α_{hy}$ is the hydraulic diffusivity of the pore fluid pressure. The prefactor $λ$ is determined by a single parameter, $T$, which depends on the pre-injection stress state and injection conditions. The prefactor has the range $0<λ<\infty$, the lower and upper limits of which correspond to marginal pressurization of the fault and critically stressed conditions, in which the fault-resolved shear stress is close to the pre-injection fault strength. In both limits, we derive solutions for slip by perturbation expansion, to arbitrary order. In the marginally pressurized limit ($λ\rightarrow 0$), the perturbation is regular and the series ex

A prompt injection attack aims to inject malicious instruction/data into the input of an LLM-Integrated Application such that it produces results as an attacker desires. Existing works are limited to case studies. As a result, the literature lacks a systematic understanding of prompt injection attacks and their defenses. We aim to bridge the gap in this work. In particular, we propose a framework to formalize prompt injection attacks. Existing attacks are special cases in our framework. Moreover, based on our framework, we design a new attack by combining existing ones. Using our framework, we conduct a systematic evaluation on 5 prompt injection attacks and 10 defenses with 10 LLMs and 7 tasks. Our work provides a common benchmark for quantitatively evaluating future prompt injection attacks and defenses. To facilitate research on this topic, we make our platform public at https://github.com/liu00222/Open-Prompt-Injection.

Modern advanced driver assistance systems (ADAS) rely on deep neural networks (DNNs) for perception and planning. Since DNNs' parameters reside in DRAM during inference, bit flips caused by cosmic radiation or low-voltage operation may corrupt DNN computations, distort driving decisions, and lead to real-world incidents. This paper presents a SpatioTemporal-Aware Fault Injection (STAFI) framework to locate critical fault sites in DNNs for ADAS efficiently. Spatially, we propose a Progressive Metric-guided Bit Search (PMBS) that efficiently identifies critical network weight bits whose corruption causes the largest deviations in driving behavior (e.g., unintended acceleration or steering). Furthermore, we develop a Critical Fault Time Identification (CFTI) mechanism that determines when to trigger these faults, taking into account the context of real-time systems and environmental states, to maximize the safety impact. Experiments on DNNs for a production ADAS demonstrate that STAFI uncovers 29.56x more hazard-inducing critical faults than the strongest baseline.