Daily Papers

Federated learning (FL) has garnered significant attention as a prominent privacy-preserving Machine Learning (ML) paradigm. Decentralized FL (DFL) eschews traditional FL's centralized server architecture, enhancing the system's robustness and scalability. However, these advantages of DFL also create new vulnerabilities for malicious participants to execute adversarial attacks, especially model poisoning attacks. In model poisoning attacks, malicious participants aim to diminish the performance of benign models by creating and disseminating the compromised model. Existing research on model poisoning attacks has predominantly concentrated on undermining global models within the Centralized FL (CFL) paradigm, while there needs to be more research in DFL. To fill the research gap, this paper proposes an innovative model poisoning attack called DMPA. This attack calculates the differential characteristics of multiple malicious client models and obtains the most effective poisoning strategy, thereby orchestrating a collusive attack by multiple participants. The effectiveness of this attack is validated across multiple datasets, with results indicating that the DMPA approach consistently surpasses existing state-of-the-art FL model poisoning attack strategies.

Multi-agent systems, where LLM agents communicate through free-form language, enable sophisticated coordination for solving complex cooperative tasks. This surfaces a unique safety problem when individual agents form a coalition and collude to pursue secondary goals and degrade the joint objective. In this paper, we present Colosseum, a framework for auditing LLM agents' collusive behavior in multi-agent settings. We ground how agents cooperate through a Distributed Constraint Optimization Problem (DCOP) and measure collusion via regret relative to the cooperative optimum. Colosseum tests each LLM for collusion under different objectives, persuasion tactics, and network topologies. Through our audit, we show that most out-of-the-box models exhibited a propensity to collude when a secret communication channel was artificially formed. Furthermore, we discover ``collusion on paper'' when agents plan to collude in text but would often pick non-collusive actions, thus providing little effect on the joint task. Colosseum provides a new way to study collusion by measuring communications and actions in rich yet verifiable environments.

An important function of collaborative network intrusion detection is to analyze the network logs of the collaborators for joint IP addresses. However, sharing IP addresses in plain is sensitive and may be even subject to privacy legislation as it is personally identifiable information. In this paper, we present the privacy-preserving collection of IP addresses. We propose a single collector, over-threshold private set intersection protocol. In this protocol N participants identify the IP addresses that appear in at least t participant's sets without revealing any information about other IP addresses. Using a novel hashing scheme, we reduce the computational complexity of the previous state-of-the-art solution from O(M(N M/t)^{2t}) to O(t^2MN{t}), where M denotes the dataset size. This reduction makes it practically feasible to apply our protocol to real network logs. We test our protocol using joint networks logs of multiple institutions. Additionally, we present two deployment options: a collusion-safe deployment, which provides stronger security guarantees at the cost of increased communication overhead, and a non-interactive deployment, which assumes a non-colluding collector but offers significantly lower communication costs and applicable to many use cases of collaborative network intrusion detection similar to ours.

As LLM agents are increasingly deployed in multi-agent systems, they introduce risks of covert coordination that may evade standard forms of human oversight. While linear probes on model activations have shown promise for detecting deception in single-agent settings, collusion is inherently a multi-agent phenomenon, and the use of internal representations for detecting collusion between agents remains unexplored. We introduce NARCBench, a benchmark for evaluating collusion detection under environment distribution shift, and propose five probing techniques that aggregate per-agent deception scores to classify scenarios at the group level. Our probes achieve 1.00 AUROC in-distribution and 0.60--0.86 AUROC when transferred zero-shot to structurally different multi-agent scenarios and a steganographic blackjack card-counting task. We find that no single probing technique dominates across all collusion types, suggesting that different forms of collusion manifest differently in activation space. We also find preliminary evidence that this signal is localised at the token level, with the colluding agent's activations spiking specifically when processing the encoded parts of their partner's message. This work takes a step toward multi-agent interpretability: extending white-box inspection from single models to multi-agent contexts, where detection requires aggregating signals across agents. These results suggest that model internals provide a complementary signal to text-level monitoring for detecting multi-agent collusion, particularly for organisations with access to model activations. Code and data are available at https://github.com/aaronrose227/narcbench.

LLM based agents are increasingly deployed in high stakes settings where they process external data sources such as emails, documents, and code repositories. This creates exposure to indirect prompt injection attacks, where adversarial instructions embedded in external content manipulate agent behavior without user awareness. A critical but underexplored dimension of this threat is concealment: since users tend to observe only an agent's final response, an attack can conceal its existence by presenting no clue of compromise in the final user facing response while successfully executing harmful actions. This leaves users unaware of the manipulation and likely to accept harmful outcomes as legitimate. We present findings from a large scale public red teaming competition evaluating this dual objective across three agent settings: tool calling, coding, and computer use. The competition attracted 464 participants who submitted 272000 attack attempts against 13 frontier models, yielding 8648 successful attacks across 41 scenarios. All models proved vulnerable, with attack success rates ranging from 0.5% (Claude Opus 4.5) to 8.5% (Gemini 2.5 Pro). We identify universal attack strategies that transfer across 21 of 41 behaviors and multiple model families, suggesting fundamental weaknesses in instruction following architectures. Capability and robustness showed weak correlation, with Gemini 2.5 Pro exhibiting both high capability and high vulnerability. To address benchmark saturation and obsoleteness, we will endeavor to deliver quarterly updates through continued red teaming competitions. We open source the competition environment for use in evaluations, along with 95 successful attacks against Qwen that did not transfer to any closed source model. We share model-specific attack data with respective frontier labs and the full dataset with the UK AISI and US CAISI to support robustness research.

Retrieval-Augmented Generation (RAG) systems, widely used to improve the factual grounding of large language models (LLMs), are increasingly vulnerable to poisoning attacks, where adversaries inject manipulated content into the retriever's corpus. While prior research has predominantly focused on single-attacker settings, real-world scenarios often involve multiple, competing attackers with conflicting objectives. In this work, we introduce PoisonArena, the first benchmark to systematically study and evaluate competing poisoning attacks in RAG. We formalize the multi-attacker threat model, where attackers vie to control the answer to the same query using mutually exclusive misinformation. PoisonArena leverages the Bradley-Terry model to quantify each method's competitive effectiveness in such adversarial environments. Through extensive experiments on the Natural Questions and MS MARCO datasets, we demonstrate that many attack strategies successful in isolation fail under competitive pressure. Our findings highlight the limitations of conventional evaluation metrics like Attack Success Rate (ASR) and F1 score and underscore the need for competitive evaluation to assess real-world attack robustness. PoisonArena provides a standardized framework to benchmark and develop future attack and defense strategies under more realistic, multi-adversary conditions.

This study probes the vulnerabilities of cooperative multi-agent reinforcement learning (c-MARL) under adversarial attacks, a critical determinant of c-MARL's worst-case performance prior to real-world implementation. Current observation-based attacks, constrained by white-box assumptions, overlook c-MARL's complex multi-agent interactions and cooperative objectives, resulting in impractical and limited attack capabilities. To address these shortcomes, we propose Adversarial Minority Influence (AMI), a practical and strong for c-MARL. AMI is a practical black-box attack and can be launched without knowing victim parameters. AMI is also strong by considering the complex multi-agent interaction and the cooperative goal of agents, enabling a single adversarial agent to unilaterally misleads majority victims to form targeted worst-case cooperation. This mirrors minority influence phenomena in social psychology. To achieve maximum deviation in victim policies under complex agent-wise interactions, our unilateral attack aims to characterize and maximize the impact of the adversary on the victims. This is achieved by adapting a unilateral agent-wise relation metric derived from mutual information, thereby mitigating the adverse effects of victim influence on the adversary. To lead the victims into a jointly detrimental scenario, our targeted attack deceives victims into a long-term, cooperatively harmful situation by guiding each victim towards a specific target, determined through a trial-and-error process executed by a reinforcement learning agent. Through AMI, we achieve the first successful attack against real-world robot swarms and effectively fool agents in simulated environments into collectively worst-case scenarios, including Starcraft II and Multi-agent Mujoco. The source code and demonstrations can be found at: https://github.com/DIG-Beihang/AMI.

Multi-agent systems composed of large generative models are rapidly moving from laboratory prototypes to real-world deployments, where they jointly plan, negotiate, and allocate shared resources to solve complex tasks. While such systems promise unprecedented scalability and autonomy, their collective interaction also gives rise to failure modes that cannot be reduced to individual agents. Understanding these emergent risks is therefore critical. Here, we present a pioneer study of such emergent multi-agent risk in workflows that involve competition over shared resources (e.g., computing resources or market share), sequential handoff collaboration (where downstream agents see only predecessor outputs), collective decision aggregation, and others. Across these settings, we observe that such group behaviors arise frequently across repeated trials and a wide range of interaction conditions, rather than as rare or pathological cases. In particular, phenomena such as collusion-like coordination and conformity emerge with non-trivial frequency under realistic resource constraints, communication protocols, and role assignments, mirroring well-known pathologies in human societies despite no explicit instruction. Moreover, these risks cannot be prevented by existing agent-level safeguards alone. These findings expose the dark side of intelligent multi-agent systems: a social intelligence risk where agent collectives, despite no instruction to do so, spontaneously reproduce familiar failure patterns from human societies.

As audio-visual multi-modal large language models (MLLMs) are increasingly deployed in safety-critical applications, understanding their vulnerabilities is crucial. To this end, we introduce Multi-Modal Typography, a systematic study examining how typographic attacks across multiple modalities adversely influence MLLMs. While prior work focuses narrowly on unimodal attacks, we expose the cross-modal fragility of MLLMs. We analyze the interactions between audio, visual, and text perturbations and reveal that coordinated multi-modal attack creates a significantly more potent threat than single-modality attacks (attack success rate = 83.43% vs 34.93%).Our findings across multiple frontier MLLMs, tasks, and common-sense reasoning and content moderation benchmarks establishes multi-modal typography as a critical and underexplored attack strategy in multi-modal reasoning. Code and data will be publicly available.

Engaging in the deliberate generation of abnormal outputs from Large Language Models (LLMs) by attacking them is a novel human activity. This paper presents a thorough exposition of how and why people perform such attacks, defining LLM red-teaming based on extensive and diverse evidence. Using a formal qualitative methodology, we interviewed dozens of practitioners from a broad range of backgrounds, all contributors to this novel work of attempting to cause LLMs to fail. We focused on the research questions of defining LLM red teaming, uncovering the motivations and goals for performing the activity, and characterizing the strategies people use when attacking LLMs. Based on the data, LLM red teaming is defined as a limit-seeking, non-malicious, manual activity, which depends highly on a team-effort and an alchemist mindset. It is highly intrinsically motivated by curiosity, fun, and to some degrees by concerns for various harms of deploying LLMs. We identify a taxonomy of 12 strategies and 35 different techniques of attacking LLMs. These findings are presented as a comprehensive grounded theory of how and why people attack large language models: LLM red teaming.

Red-teaming, or identifying prompts that elicit harmful responses, is a critical step in ensuring the safe and responsible deployment of large language models (LLMs). Developing effective protection against many modes of attack prompts requires discovering diverse attacks. Automated red-teaming typically uses reinforcement learning to fine-tune an attacker language model to generate prompts that elicit undesirable responses from a target LLM, as measured, for example, by an auxiliary toxicity classifier. We show that even with explicit regularization to favor novelty and diversity, existing approaches suffer from mode collapse or fail to generate effective attacks. As a flexible and probabilistically principled alternative, we propose to use GFlowNet fine-tuning, followed by a secondary smoothing phase, to train the attacker model to generate diverse and effective attack prompts. We find that the attacks generated by our method are effective against a wide range of target LLMs, both with and without safety tuning, and transfer well between target LLMs. Finally, we demonstrate that models safety-tuned using a dataset of red-teaming prompts generated by our method are robust to attacks from other RL-based red-teaming approaches.

A multi-agent system (MAS) powered by large language models (LLMs) can automate tedious user tasks such as meeting scheduling that requires inter-agent collaboration. LLMs enable nuanced protocols that account for unstructured private data, user constraints, and preferences. However, this design introduces new risks, including misalignment and attacks by malicious parties that compromise agents or steal user data. In this paper, we propose the Terrarium framework for fine-grained study on safety, privacy, and security in LLM-based MAS. We repurpose the blackboard design, an early approach in multi-agent systems, to create a modular, configurable testbed for multi-agent collaboration. We identify key attack vectors such as misalignment, malicious agents, compromised communication, and data poisoning. We implement three collaborative MAS scenarios with four representative attacks to demonstrate the framework's flexibility. By providing tools to rapidly prototype, evaluate, and iterate on defenses and designs, Terrarium aims to accelerate progress toward trustworthy multi-agent systems.

Large Language Model-based Multi-Agent Systems (LLM-MAS) are increasingly applied to complex collaborative scenarios. However, their collaborative mechanisms may cause minor inaccuracies to gradually solidify into system-level false consensus through iteration. Such risks are difficult to trace since errors can propagate and amplify through message dependencies. Existing protections often rely on single-agent validation or require modifications to the collaboration architecture, which can weaken effective information flow and may not align with natural collaboration processes in real tasks. To address this, we propose a propagation dynamics model tailored for LLM-MAS that abstracts collaboration as a directed dependency graph and provides an early-stage risk criterion to characterize amplification risk. Through experiments on six mainstream frameworks, we identify three vulnerability classes: cascade amplification, topological sensitivity, and consensus inertia. We further instantiate an attack where injecting just a single atomic error seed leads to widespread failure. In response, we introduce a genealogy-graph-based governance layer, implemented as a message-layer plugin, that suppresses both endogenous and exogenous error amplification without altering the collaboration architecture. Experiments show that this approach raises the defense success rate from a baseline of 0.32 to over 0.89 and significantly mitigates the cascading spread of minor errors.

Federated Learning (FL), a privacy-preserving decentralized machine learning framework, has been shown to be vulnerable to backdoor attacks. Current research primarily focuses on the Single-Label Backdoor Attack (SBA), wherein adversaries share a consistent target. However, a critical fact is overlooked: adversaries may be non-cooperative, have distinct targets, and operate independently, which exhibits a more practical scenario called Multi-Label Backdoor Attack (MBA). Unfortunately, prior works are ineffective in the MBA scenario since non-cooperative attackers exclude each other. In this work, we conduct an in-depth investigation to uncover the inherent constraints of the exclusion: similar backdoor mappings are constructed for different targets, resulting in conflicts among backdoor functions. To address this limitation, we propose Mirage, the first non-cooperative MBA strategy in FL that allows attackers to inject effective and persistent backdoors into the global model without collusion by constructing in-distribution (ID) backdoor mapping. Specifically, we introduce an adversarial adaptation method to bridge the backdoor features and the target distribution in an ID manner. Additionally, we further leverage a constrained optimization method to ensure the ID mapping survives in the global training dynamics. Extensive evaluations demonstrate that Mirage outperforms various state-of-the-art attacks and bypasses existing defenses, achieving an average ASR greater than 97\% and maintaining over 90\% after 900 rounds. This work aims to alert researchers to this potential threat and inspire the design of effective defense mechanisms. Code has been made open-source.

The proliferation of agentic AI coding assistants, including Claude Code, GitHub Copilot, Cursor, and emerging skill-based architectures, has fundamentally transformed software development workflows. These systems leverage Large Language Models (LLMs) integrated with external tools, file systems, and shell access through protocols like the Model Context Protocol (MCP). However, this expanded capability surface introduces critical security vulnerabilities. In this Systematization of Knowledge (SoK) paper, we present a comprehensive analysis of prompt injection attacks targeting agentic coding assistants. We propose a novel three-dimensional taxonomy categorizing attacks across delivery vectors, attack modalities, and propagation behaviors. Our meta-analysis synthesizes findings from 78 recent studies (2021--2026), consolidating evidence that attack success rates against state-of-the-art defenses exceed 85\% when adaptive attack strategies are employed. We systematically catalog 42 distinct attack techniques spanning input manipulation, tool poisoning, protocol exploitation, multimodal injection, and cross-origin context poisoning. Through critical analysis of 18 defense mechanisms reported in prior work, we identify that most achieve less than 50\% mitigation against sophisticated adaptive attacks. We contribute: (1) a unified taxonomy bridging disparate attack classifications, (2) the first systematic analysis of skill-based architecture vulnerabilities with concrete exploit chains, and (3) a defense-in-depth framework grounded in the limitations we identify. Our findings indicate that the security community must treat prompt injection as a first-class vulnerability class requiring architectural-level mitigations rather than ad-hoc filtering approaches.

Large Language Model-based Multi-Agent Systems (LLM-MAS) have revolutionized complex problem-solving capability by enabling sophisticated agent collaboration through message-based communications. While the communication framework is crucial for agent coordination, it also introduces a critical yet unexplored security vulnerability. In this work, we introduce Agent-in-the-Middle (AiTM), a novel attack that exploits the fundamental communication mechanisms in LLM-MAS by intercepting and manipulating inter-agent messages. Unlike existing attacks that compromise individual agents, AiTM demonstrates how an adversary can compromise entire multi-agent systems by only manipulating the messages passing between agents. To enable the attack under the challenges of limited control and role-restricted communication format, we develop an LLM-powered adversarial agent with a reflection mechanism that generates contextually-aware malicious instructions. Our comprehensive evaluation across various frameworks, communication structures, and real-world applications demonstrates that LLM-MAS is vulnerable to communication-based attacks, highlighting the need for robust security measures in multi-agent systems.

Mixture of Experts (MoE) has become a key ingredient for scaling large foundation models while keeping inference costs steady. We show that expert routing strategies that have cross-batch dependencies are vulnerable to attacks. Malicious queries can be sent to a model and can affect a model's output on other benign queries if they are grouped in the same batch. We demonstrate this via a proof-of-concept attack in a toy experimental setting.

This study exposes the safety vulnerabilities of Large Language Models (LLMs) in multi-turn interactions, where malicious users can obscure harmful intents across several queries. We introduce ActorAttack, a novel multi-turn attack method inspired by actor-network theory, which models a network of semantically linked actors as attack clues to generate diverse and effective attack paths toward harmful targets. ActorAttack addresses two main challenges in multi-turn attacks: (1) concealing harmful intents by creating an innocuous conversation topic about the actor, and (2) uncovering diverse attack paths towards the same harmful target by leveraging LLMs' knowledge to specify the correlated actors as various attack clues. In this way, ActorAttack outperforms existing single-turn and multi-turn attack methods across advanced aligned LLMs, even for GPT-o1. We will publish a dataset called SafeMTData, which includes multi-turn adversarial prompts and safety alignment data, generated by ActorAttack. We demonstrate that models safety-tuned using our safety dataset are more robust to multi-turn attacks. Code is available at https://github.com/renqibing/ActorAttack.

To ensure AI safety, instruction-tuned Large Language Models (LLMs) are specifically trained to ensure alignment, which refers to making models behave in accordance with human intentions. While these models have demonstrated commendable results on various safety benchmarks, the vulnerability of their safety alignment has not been extensively studied. This is particularly troubling given the potential harm that LLMs can inflict. Existing attack methods on LLMs often rely on poisoned training data or the injection of malicious prompts. These approaches compromise the stealthiness and generalizability of the attacks, making them susceptible to detection. Additionally, these models often demand substantial computational resources for implementation, making them less practical for real-world applications. Inspired by recent success in modifying model behavior through steering vectors without the need for optimization, and drawing on its effectiveness in red-teaming LLMs, we conducted experiments employing activation steering to target four key aspects of LLMs: truthfulness, toxicity, bias, and harmfulness - across a varied set of attack settings. To establish a universal attack strategy applicable to diverse target alignments without depending on manual analysis, we automatically select the intervention layer based on contrastive layer search. Our experiment results show that activation attacks are highly effective and add little or no overhead to attack efficiency. Additionally, we discuss potential countermeasures against such activation attacks. Our code and data are available at https://github.com/wang2226/Backdoor-Activation-Attack Warning: this paper contains content that can be offensive or upsetting.

The widespread distribution of Large Language Models (LLMs) through public platforms like Hugging Face introduces significant security challenges. While these platforms perform basic security scans, they often fail to detect subtle manipulations within the embedding layer. This work identifies a novel class of deployment phase attacks that exploit this vulnerability by injecting imperceptible perturbations directly into the embedding layer outputs without modifying model weights or input text. These perturbations, though statistically benign, systematically bypass safety alignment mechanisms and induce harmful behaviors during inference. We propose Search based Embedding Poisoning(SEP), a practical, model agnostic framework that introduces carefully optimized perturbations into embeddings associated with high risk tokens. SEP leverages a predictable linear transition in model responses, from refusal to harmful output to semantic deviation to identify a narrow perturbation window that evades alignment safeguards. Evaluated across six aligned LLMs, SEP achieves an average attack success rate of 96.43% while preserving benign task performance and evading conventional detection mechanisms. Our findings reveal a critical oversight in deployment security and emphasize the urgent need for embedding level integrity checks in future LLM defense strategies.

Dense retrieval systems have been widely used in various NLP applications. However, their vulnerabilities to potential attacks have been underexplored. This paper investigates a novel attack scenario where the attackers aim to mislead the retrieval system into retrieving the attacker-specified contents. Those contents, injected into the retrieval corpus by attackers, can include harmful text like hate speech or spam. Unlike prior methods that rely on model weights and generate conspicuous, unnatural outputs, we propose a covert backdoor attack triggered by grammar errors. Our approach ensures that the attacked models can function normally for standard queries while covertly triggering the retrieval of the attacker's contents in response to minor linguistic mistakes. Specifically, dense retrievers are trained with contrastive loss and hard negative sampling. Surprisingly, our findings demonstrate that contrastive loss is notably sensitive to grammatical errors, and hard negative sampling can exacerbate susceptibility to backdoor attacks. Our proposed method achieves a high attack success rate with a minimal corpus poisoning rate of only 0.048\%, while preserving normal retrieval performance. This indicates that the method has negligible impact on user experience for error-free queries. Furthermore, evaluations across three real-world defense strategies reveal that the malicious passages embedded within the corpus remain highly resistant to detection and filtering, underscoring the robustness and subtlety of the proposed attack Codes of this work are available at https://github.com/ruyue0001/Backdoor_DPR..

Multimodal large language models with Retrieval Augmented Generation (RAG) have significantly advanced tasks such as multimodal question answering by grounding responses in external text and images. This grounding improves factuality, reduces hallucination, and extends reasoning beyond parametric knowledge. However, this reliance on external knowledge poses a critical yet underexplored safety risk: knowledge poisoning attacks, where adversaries deliberately inject adversarial multimodal content into external knowledge bases to steer model toward generating incorrect or even harmful responses. To expose such vulnerabilities, we propose MM-PoisonRAG, the first framework to systematically design knowledge poisoning in multimodal RAG. We introduce two complementary attack strategies: Localized Poisoning Attack (LPA), which implants targeted multimodal misinformation to manipulate specific queries, and Globalized Poisoning Attack (GPA), which inserts a single adversarial knowledge to broadly disrupt reasoning and induce nonsensical responses across all queries. Comprehensive experiments across tasks, models, and access settings show that LPA achieves targeted manipulation with attack success rates of up to 56%, while GPA completely disrupts model generation to 0% accuracy with just a single adversarial knowledge injection. Our results reveal the fragility of multimodal RAG and highlight the urgent need for defenses against knowledge poisoning.

Prompt injection attacks pose serious security risks across a wide range of real-world applications. While receiving increasing attention, the community faces a critical gap: the lack of a unified platform for prompt injection evaluation. This makes it challenging to reliably compare defenses, understand their true robustness under diverse attacks, or assess how well they generalize across tasks and benchmarks. For instance, many defenses initially reported as effective were later found to exhibit limited robustness on diverse datasets and attacks. To bridge this gap, we introduce PIArena, a unified and extensible platform for prompt injection evaluation that enables users to easily integrate state-of-the-art attacks and defenses and evaluate them across a variety of existing and new benchmarks. We also design a dynamic strategy-based attack that adaptively optimizes injected prompts based on defense feedback. Through comprehensive evaluation using PIArena, we uncover critical limitations of state-of-the-art defenses: limited generalizability across tasks, vulnerability to adaptive attacks, and fundamental challenges when an injected task aligns with the target task. The code and datasets are available at https://github.com/sleeepeer/PIArena.

This paper introduces a new method for adversarial attacks on large language models (LLMs) called the Single-Turn Crescendo Attack (STCA). Building on the multi-turn crescendo attack method introduced by Russinovich, Salem, and Eldan (2024), which gradually escalates the context to provoke harmful responses, the STCA achieves similar outcomes in a single interaction. By condensing the escalation into a single, well-crafted prompt, the STCA bypasses typical moderation filters that LLMs use to prevent inappropriate outputs. This technique reveals vulnerabilities in current LLMs and emphasizes the importance of stronger safeguards in responsible AI (RAI). The STCA offers a novel method that has not been previously explored.

Prompt leakage poses a compelling security and privacy threat in LLM applications. Leakage of system prompts may compromise intellectual property, and act as adversarial reconnaissance for an attacker. A systematic evaluation of prompt leakage threats and mitigation strategies is lacking, especially for multi-turn LLM interactions. In this paper, we systematically investigate LLM vulnerabilities against prompt leakage for 10 closed- and open-source LLMs, across four domains. We design a unique threat model which leverages the LLM sycophancy effect and elevates the average attack success rate (ASR) from 17.7% to 86.2% in a multi-turn setting. Our standardized setup further allows dissecting leakage of specific prompt contents such as task instructions and knowledge documents. We measure the mitigation effect of 7 black-box defense strategies, along with finetuning an open-source model to defend against leakage attempts. We present different combination of defenses against our threat model, including a cost analysis. Our study highlights key takeaways for building secure LLM applications and provides directions for research in multi-turn LLM interactions

This paper identifies and analyzes a novel vulnerability class in Model Context Protocol (MCP) based agent systems. The attack chain describes and demonstrates how benign, individually authorized tasks can be orchestrated to produce harmful emergent behaviors. Through systematic analysis using the MITRE ATLAS framework, we demonstrate how 95 agents tested with access to multiple services-including browser automation, financial analysis, location tracking, and code deployment-can chain legitimate operations into sophisticated attack sequences that extend beyond the security boundaries of any individual service. These red team exercises survey whether current MCP architectures lack cross-domain security measures necessary to detect or prevent a large category of compositional attacks. We present empirical evidence of specific attack chains that achieve targeted harm through service orchestration, including data exfiltration, financial manipulation, and infrastructure compromise. These findings reveal that the fundamental security assumption of service isolation fails when agents can coordinate actions across multiple domains, creating an exponential attack surface that grows with each additional capability. This research provides a barebones experimental framework that evaluate not whether agents can complete MCP benchmark tasks, but what happens when they complete them too well and optimize across multiple services in ways that violate human expectations and safety constraints. We propose three concrete experimental directions using the existing MCP benchmark suite.