95 results for “Injection”

Understanding and Mitigating Role Spoofing in LLMs is an arXiv paper (October 2026) on role spoofing, a prompt injection where malicious tool output impersonates the user. It shows models track "who said what" using both content cues (semantics, style) and structure cues (control tokens, formatting). Attacks combining fake structure and content succeed often, adversarial training stops only content-based attacks, and linear probes trained to predict conversation roles detect spoofing across settings.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🛡️ Security

SEAL is a defense against indirect prompt injection for local LLM agents. It learns a soft prompt and scoring readout offline, then uses a frozen task model to score and delete injected tokens from tool outputs, with queries built from trusted initial context. On AgentDojo, attack success falls from 34.6% to 0.1% and Joint rises from 51.4% to 77.5%, using about 127x fewer defense tokens than PISanitizer.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🛡️ Security

Paper studying whether self-evolution weakens personal agents' robustness to indirect prompt injection (IPI). A two-phase method has the user ask the agent to follow tasks in external content, then tests with a third-party injected task the user did not authorize. Against the non-evolving base agent, attack success rose 14.0 to 48.8 points under skill evolution and 3.0 to 39.3 points under reinforcement learning. Code is released.

arxiv.orgpaperAdded 31 Oct 2026, published 29 Oct 2026🛡️ Security

Addendum to the GPT-6 Luna system card for the model update of 29 October 2026 (the title says October 28). OpenAI keeps the High rating for cybersecurity and biological/chemical capability under its Preparedness Framework. It reports ExploitBench 46.19% and SEC-Bench Pro 51.75% (GPT-6 Astra: 85.4%), higher multiturn jailbreak defender success than the October 7 version, and GPT-Red prompt-injection results. It also reports weaker factuality and dual-use biology refusal.

deploymentsafety.openai.compaperPublished and added 29 Oct 2026🛡️ Security

Log-To-Leak is an arXiv paper (cs.CR, cs.AI) by Yuepeng Hu, Chongyu Fan, Sirat Samyoun and Jian Du on prompt injection attacks against tool-using LLM agents through the Model Context Protocol. The title points to a logging-based data leakage channel. The fetched page holds only the title, authors and arXiv link, with no abstract, so the method and results are not visible here.

x.compaperAdded 29 Oct 2026, published 28 Oct 2026🛡️ Security

LLMON (LLM Object Notation) is a framework that separates instructions from data and requires explicit identifier-based execution. The authors post-train Granite-4.1 3B and 8B base models with SFT then RL on a 4.6B-token structured corpus covering general, RAG and tool-calling data, with distractor variants. Attack success rates on indirect prompt-injection benchmarks fall from 53-100% to 0.02-5.6%, and general capability is preserved. Attention-based metrics diagnose failures per transformer layer.

arxiv.orgpaperAdded 30 Oct 2026, published 28 Oct 2026🛡️ Security

LLMON (LLM Object Notation) is a framework that separates executable instructions from passive data and requires explicit identifier-based execution. The paper post-trains Granite-4.1 3B and 8B base models on a 4.6B-token corpus of general, RAG and tool-calling data with distractor variants, using SFT then RL. Attack success rates on indirect prompt-injection benchmarks fall from 53-100% to 0.02-5.6%, with general capabilities preserved. It also adds attention-based per-layer diagnostics.

arxiv.orgpaperAdded 30 Oct 2026, published 28 Oct 2026🧠 AI

Heads of Security (HoS) is a prompt injection defense that enforces instruction-data separation with learned per-head attention gates, redistributing attention between trusted and untrusted tokens. Evaluated on Llama-3.3-70B-Instruct and GPT-OSS-20B against SecAlign++ and ASIDE. The black-box RL-Hammer attack reaches 97% ASR on GPT-OSS-20B and 85% on SecAlign but 10% on HoS. The white-box Claudini attack needs 3 times more FLOPs on HoS. HoS keeps base-model utility in few-shot settings, where SecAlign and ASIDE degrade.

arxiv.orgpaperAdded 28 Oct 2026, published 27 Oct 2026🛡️ Security

Heads of Security (HoS) is a prompt injection defense that trains a small set of per-head attention gates to shift attention between trusted and untrusted tokens, separating instructions from data. Tested on Llama-3.3-70B-Instruct and GPT-OSS-20B, the black-box RL-Hammer attack reaches 10% ASR on HoS versus 97% on base GPT-OSS-20B and 85% on SecAlign. HoS keeps base utility in few-shot settings, where SecAlign and ASIDE degrade.

arxiv.orgpaperAdded 28 Oct 2026, published 27 Oct 2026🧠 AI

What Victims Take Away is an October 2026 arXiv paper analysing 563,349 Tensor Trust attack records to see how players reuse prompt injections they received. Within a week, targets reused 10.0% of attacks that broke their defense, versus 0.03% of matched attacks. Breaches were followed by defense changes that cut same-family attack success by 24.5 points. Received texts made up 23.2% of successful hijacks. The authors propose that AI security tools describe techniques without disclosing full attack text.

arxiv.orgpaperAdded 28 Oct 2026, published 27 Oct 2026🛡️ Security

Log-To-Leak is a prompt-injection attack on tool-using LLM agents that abuses Model Context Protocol tool metadata to trigger an attacker-controlled logging call after the last legitimate call, leaking the user query, server name, server response and agent response. Evaluated on five MCP servers and four agents, it reaches 62.64-100% attack success with 76.55-95.64% logging accuracy. Three defenses leave over 55% residual ASR on GitHub MCP with GPT-4o.

arxiv.orgpaperAdded 28 Oct 2026, published 26 Oct 2026🛡️ Security

An empirical study of security concerns in the OpenClaw AI personal assistant framework, using 25K+ posts from 13 subreddits (162 security-relevant) and 358 security-tagged GitHub issues (180 retained). It maps 521 CWE instances across 59 weakness types, concentrated in access control, sensitive-data exposure, injection and unsafe execution. It extracts 543 practitioner mitigations into seven strategies and reconstructs a sensitive-data exposure path with dummy secrets.

arxiv.orgpaperAdded 28 Oct 2026, published 26 Oct 2026🛡️ Security

Scope-Gap Injection (SGI) is an indirect prompt injection method for multi-stage LLM-agent safety pipelines, tested in the IJCAI 2026 OpenClaw Agent Security Attack Competition. The visible task and skill stay benign while target identifiers, bypass parameters and authorization steps sit in an attachment written as a business worklist, which the S2 LLM judge appears not to inspect. SGI bypassed all three stages (static scan, LLM judge, live audit) for 5 of 19 targets, best F = 0.385. Attachment placement had 78% S2 acceptance versus 23% in-skill. The paper closes with defenses that align evaluator scope with agent channels.

arxiv.orgpaperAdded 27 Oct 2026, published 26 Oct 2026🛡️ Security

Addendum to the GPT-6 Astra system card for OpenAI's GPT-6.1 Terra, rated Critical in cybersecurity and High in biological and chemical capability under the Preparedness Framework. It reports multiturn jailbreak and prompt-injection results, alignment tests (auto-review circumvention, ExploitGym honeypot, monitor evasion), and cyber scores: 97.10% ExploitBench, 24.5% on the internal recent-vulnerability port, 5 of 22 Sandbox Bench targets.

deploymentsafety.openai.compaperPublished and added 26 Oct 2026🛡️ Security

SentryBench is a benchmark for tool-using LLM agents in public-safety deployments. It scores attack success, over-refusal (missing a real emergency) and over-reaction (unwarranted action) on a cost-sensitive frontier, using real emergency-call transcripts, disaster messages and injection payloads. Across 840 episodes on three frontier models and three defenses, refuse-more caution pushed over-refusal to 100 percent on two models, while a trust-gating wrapper gave the lowest harm.

arxiv.orgpaperAdded 27 Oct 2026, published 25 Oct 2026🧠 AI

SentryBench is a benchmark for tool-using LLM agents in public-safety settings (public alerts, responder dispatch, citizen-record queries). It scores attack success, over-refusal and over-reaction on a cost-sensitive frontier, using real emergency-call transcripts and real injection/jailbreak payloads. Across 840 episodes on three frontier models and three defenses, refuse-more caution raised over-refusal to 100 percent on two models, while a trust-gating wrapper gave the lowest harm on every model.

arxiv.orgpaperAdded 27 Oct 2026, published 25 Oct 2026🛡️ Security

BARBICAN is a fail-closed prompt-injection gateway for MCP tool mediation. This paper shows its arbitration classifier ran on tokenizer-mangled text for a whole evaluation. Fixing it raised AUC from 0.392 to 0.855 while system recall fell from 80.67% to 80.34%. The routed band is 96.82% malicious, so deletion ablations mislead without substitution experiments and class-balance reporting.

arxiv.orgpaperAdded 27 Oct 2026, published 23 Oct 2026🛡️ Security

The paper analyses BARBICAN, a fail-closed prompt-injection gateway for MCP tool mediation, using deletion, substitution, repair and transplant experiments. Its arbitration classifier ran on tokenizer-mangled text for a whole evaluation. Fixing that raised AUC from 0.392 to 0.855 while system recall fell from 80.67% to 80.34%. The band it receives is 96.82% malicious. The authors argue ablation figures need the band's class balance and substitution baselines.

arxiv.orgpaperAdded 27 Oct 2026, published 23 Oct 2026🧠 AI

MCP-Injector is a protocol-aware fault-injection middleware for Model Context Protocol tool calls, with permanent, transient and intermittent faults on seedable schedules. Its fault model comes from 57,240 issue and PR threads across 1,774 open-source MCP repositories. In a 1,350-cell experiment, faults cut task-completion score by 60% (8.08 to 3.23 of 10), and the 120B gpt-oss model was no more resilient than the 20B. Agent stacks differed in how well they absorbed faults.

arxiv.orgpaperAdded 26 Oct 2026, published 23 Oct 2026🧠 AI

NTBench (Near-Target Benchmark) tests LLM and agent robustness to near-target prompt injection, where injected instructions steer a response toward a plausible but incorrect or harmful outcome instead of abandoning the task. It builds reference-guided target-instruction pairs on existing datasets, with human validation, and evaluates seven LLMs on ten language datasets and two agent benchmarks. Attack success rose 9.3 to 35.8 points (incorrect targets) and 1.7 to 17.3 points (harmful targets) over fixed-target baselines.

arxiv.orgpaperAdded 26 Oct 2026, published 23 Oct 2026🛡️ Security

Paper on prompt injection delivered through forensic evidence, where the seized device's contents are attacker-controlled and the examiner must read them. It extends the anti-forensic model to cover manipulation of the reasoning layer and gives a taxonomy (objective, technique, payload location, concealment). Three proof-of-concept attacks on an LLM-assisted analysis pipeline change a reported conclusion, write attacker-chosen content to the analyst workstation, and trigger outbound network traffic. Mitigations are proposed.

arxiv.orgpaperAdded 24 Oct 2026, published 21 Oct 2026🛡️ Security

MPIDetector is a detection framework for multi-modal prompt injection that uses few-shot in-context learning so an LLM can reason about the semantic intent of text and image inputs, without fine-tuning. The authors built a benchmark of eight textual and seven visual attack categories plus benign sets. They report up to 90.91% accuracy on text, 99.28% on images, and up to 99.80% detection on 500 obfuscated WebInject web-agent injections, beating baselines.

arxiv.orgpaperAdded 24 Oct 2026, published 21 Oct 2026🛡️ Security

Paper (arXiv, 21 October 2026) on structural false negatives in LLM vulnerability discovery harnesses. Starting from Knostic's OpenAnt and raptor, then 10 more open-source harnesses, the authors map 13 points where code selection, filtering, and finding storage or merging drop vulnerabilities. One file was dropped by six harnesses before the model saw it, five without notice. It covers attacker evasion, content-steered verdicts, prompt injection, and design principles and tests.

arxiv.orgpaperAdded 24 Oct 2026, published 21 Oct 2026🛡️ Security

Paper on memory injection, where prompt injection persists across sessions through the long-term memory of LLM-as-a-service platforms. It presents a general threat model and shows all NIST attack goals are achievable on ChatGPT Instant 5.3, Claude Sonnet 4.6 and Gemini 3 Flash. Vectors include personal memories, projects, skills, organization instructions and MCP resources. On Claude, shared memory lets an attacker compromise multiple users in an organization and exfiltrate prompts. Nearly half of the derived mitigations are unimplemented.

arxiv.orgpaperAdded 24 Oct 2026, published 20 Oct 2026🛡️ Security

CarryoverBench is a benchmark of 12 personal-assistant scenarios and 101 attack tasks for carryover attacks, where a one-time indirect prompt injection planted in agent memory files or scheduled tasks re-fires in later sessions. Three trigger channels (Instant, Autonomous, Latent) are scored by a rule-based verifier with separate planting and triggering stages. Across five frontier models, payloads re-fire over five rounds without measurable decay. Porting the Latent channel to the Hermes runtime weakens the effect but does not remove it.

arxiv.orgpaperAdded 22 Oct 2026, published 20 Oct 2026🛡️ Security

nokast-secureRAG is a roughly 0.5B-parameter small language model that works as a semantic firewall between retrieval and generation in local RAG systems. It classifies each query and context pair as safe, suspicious or malicious. It reaches 0.994 recall at a 0.026 false-positive rate in-distribution, beating Llama Guard 3 1B and Prompt-Guard-86M (0.900 recall, 0.690 FPR). Recall on PoisonedRAG falls to near zero, and single-request latency is 293-404 ms.

arxiv.orgpaperAdded 22 Oct 2026, published 20 Oct 2026🛡️ Security

Systematic evaluation of fault attacks on ML-DSA (FIPS 204) rejection sampling and of proposed countermeasures. It proves the "third rejection test" of patent WO2025261784A1 re-verifies what check 2 already guarantees, and shows that 30,000 fault injections never detect a skipped z-norm check. The author fully recovers s1 against a third-check-protected implementation under randomized signing. Third check and sign-then-verify are complementary; overhead is 0.04–0.05% on x86-64 and 0.83% on Cortex-M4 for the third check.

eprint.iacr.orgpaperAdded 21 Oct 2026, published 20 Oct 2026🛡️ Security

KNOX is a regulatory sandbox for generative models built by running AI agent workloads inside TEEs. It derives an attested policy manifest (applicable regulations, expected tool calls, permitted system calls) from the user prompt, and an attested reference monitor allows an action only if it matches. Actions and verdicts go into an attested execution graph that regulators can verify. The paper reports detection of prompt-injection deviations and under 1% overhead.

arxiv.orgpaperAdded 22 Oct 2026, published 20 Oct 2026🛡️ Security

Fake News, Real Crashes is an October 2026 arXiv paper presenting a benchmark for information injection attacks on LLM financial trading agents. It spans five injection vectors, seven event categories and about 10,000 decisions on 15 U.S. stocks. Fabricated news reached 62% attack success versus 14% for prompt injection, and two 30-word templates hit 100%. Only hierarchical aggregation reduced success among the defenses tested.

arxiv.orgpaperAdded 21 Oct 2026, published 19 Oct 2026🛡️ Security

SpanGuard is an indirect prompt injection (IPI) guardrail that splits detection into two steps. Lightweight unsupervised segmentation first locates candidate instruction blocks in tool outputs, then a small LLM judges whether each block is authorized by the user's request. The authors report a 0.32% attack success rate with negligible benign utility loss, and the guardrail still works with substantially smaller models.

arxiv.orgpaperAdded 21 Oct 2026, published 19 Oct 2026🛡️ Security

This ePrint paper presents key-independent filters that remove misinjections (faults that miss the assumed form) in elimination fault attacks. A SAT filter, a neural network filter and a combined filter were tested on clock-glitch faults into GIFT-64, GIFT-128 and BAKSHEESH on a microcontroller. The SAT filter rejected none of 14669 intended faults, and raised ToFA success on BAKSHEESH from 5.2% to 61.2% (96.2% combined).

eprint.iacr.orgpaperAdded 20 Oct 2026, published 19 Oct 2026🛡️ Security

IISFA (Ineffective Instruction-Skip Fault Attack) targets bitsliced block cipher implementations protected by redundancy countermeasures. Skipping one bitslice instruction leaves the result unchanged only when all m S-box inputs fall in an ineffective set, so each released ciphertext restricts all m subkeys. The authors give known- and chosen-plaintext key-recovery procedures and expected injection counts, and recover keys in every trial on GIFT, SKINNY and PIPO implementations on an 8-bit AVR microcontroller using clock glitches.

eprint.iacr.orgpaperAdded 20 Oct 2026, published 19 Oct 2026🛡️ Security

PIAtlas is a taxonomy of prompt injection attacks organized by how the injected instruction subverts the model, with six mechanism classes as roots of a technique and sub-technique hierarchy. The authors release PICorpus, a license-clean collection of public prompt-injection datasets, and map thousands of records to PIAtlas under a human-verified protocol. Results show attacks combine several techniques from few mechanisms, agent benchmarks cover only part of the space, and adaptive attackers favor mechanisms that curated benchmarks underrepresent.

arxiv.orgpaperAdded 21 Oct 2026, published 18 Oct 2026🛡️ Security

ToFu is a tool-aware fuzzing framework for LLM tools, including MCP servers. It maps each tool argument, using its metadata, to one of 74 CWE-guided semantics-aware sanitizers (path, URL, SQL and similar). It then generates schema-valid Atheris harnesses and semantic oracles, using runtime tracers for in-process tools and outcome oracles for opaque ones. The LLM (gpt-4.1) only maps sanitizers and writes seeds; it is never the security judge. The paper also shows a "Repeat-After-Me" trick, where an LLM told to repeat attacker-crafted text emits a tool call the orchestrator executes, reproduced on LangGraph, MAF, ADK 2.0, AutoGen, Strands and CrewAI. Across 273 runnable MCP servers (3,405 tools) it reports 253 verified bugs in 343 tools across 118 servers, such as path traversal, log injection and exfiltration. These are failures a crash-based fuzzer would miss. The result matters because it shows LLM-generated tool arguments cannot be treated as trusted.

arxiv.orgpaperAdded 18 Oct 2026🧠 AI

ToFu is a tool-aware fuzzing framework that derives semantic security checks for LLM tools (mainly MCP servers) from their metadata and a corpus of 74 CWE-guided sanitizers. It generates schema-valid Atheris harnesses, uses runtime tracers and outcome oracles to catch bugs that don't crash, and uses an LLM only to map sanitizers and seed inputs, never as the judge. It also describes a "Repeat-After-Me" attack, in which the LLM is told to repeat attacker-written text that the orchestrator then dispatches as a tool call. It was reproduced on LangGraph, MAF, ADK, AutoGen, Strands and CrewAI. On 273 MCP servers it reports 253 verified bugs across 343 tools in 118 repositories, including path traversal, log injection, exfiltration and command bypass.

arxiv.orgpaperAdded 18 Oct 2026🛡️ Security

MobileInject is a benchmark for testing MLLM-based smartphone agents against indirect multimodal injection. It has an interactive evaluation framework with 11 scenarios, 9 security-risk categories and 108 executable risk instances. It adds Cross-Modal Consistency Injection, Multi-Role Collaborative Deception, Reasoning-Camouflage Injection and a Reasoning Security Diagnosis Framework. Experiments find broad vulnerability, with many failures from early reasoning hijacking.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🛡️ Security

Ignored, Not Refused audits 33,000 trajectories from eleven open-weight agents (2-14B parameters) on InjecAgent and AgentDojo. A label-blind LLM judge finds most failed indirect prompt injections are ignored (76-80%) rather than refused (17-18%). Changing only the system prompt moves InjecAgent ASR from 0.2% to 43%. The Attack-Task Outcome Matrix (ATOM) shows 21.7% of AgentDojo episodes were safe but left the user task incomplete. The corpus, judge prompt and code are released.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🛡️ Security

"Ignored, Not Refused" audits 33,000 trajectories from eleven open-weight agents (2-14B parameters) on InjecAgent and AgentDojo. A label-blind LLM judge finds 76-80% of failed injection attacks were ignored and 17-18% refused. Changing only the system prompt moves InjecAgent attack success rate from 0.2% to 43%, and the ATOM matrix shows 21.7% of AgentDojo episodes were safe but left the user's task undone. The corpus, judge prompt and code are released.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🧠 AI

R^2ECAP (Reused Reasoning-Cache Poisoning) is a black-box attack that plants poisoned reasoning traces through a chat interface, exploiting LLM systems that reuse KV caches of reasoning. A preceding trace steers the model, and the retained trace keeps residual adversarial influence in its cache after the preceding trace is discarded. On three open-weight LLMs and three datasets it reaches up to 79% ASR, beats the white-box HijackKV in 8 of 9 settings, and reaches 46% ASR in a reasoning-trace retrieval system. Only 1% of retained responses are flagged by a popular prompt injection detector.

arxiv.orgpaperAdded 20 Oct 2026, published 16 Oct 2026🛡️ Security

Presents fault injection attacks on BIKE and HQC decapsulation that target instructions outside the Fujisaki-Okamoto transform, achieving full key recovery with only valid ciphertexts at NIST Level 1 (17392 queries for HQC, about 3 million for BIKE). It also gives a general method for finding effective fault locations by analysing decoder behaviour under perturbed conditions. Attacks were validated with laser fault injection on an STM32H7 microcontroller, showing a wider attack surface for code-based post-quantum KEM implementations.

eprint.iacr.orgpaperPublished and added 15 Oct 2026🛡️ Security

Evaluates a native probabilistic decision model (Jev), Llama Prompt Guard 2 (22M and 86M) and an LLM judge on 47,611 held-out examples covering direct/indirect prompt injection, hard-benign inputs, obfuscation and distribution shift. The LLM judge has the best AUROC (0.897), followed by Jev (0.871), Prompt Guard 2 86M (0.757) and 22M (0.714). Jev's native probabilities do not consistently beat post-hoc calibrated classifier scores, and temperature scaling improves calibration but not discrimination. At a 1% false-negative-rate target, every system needs over 91.8% escalation, so probabilistic outputs alone do not give reliable security decisions under distribution shift.

arxiv.orgpaperAdded 17 Oct 2026, published 14 Oct 2026🛡️ Security

INFUSE is an end-to-end benchmark for indirect prompt injection (IPI) across long-horizon agent workflows (shell commands, code changes, dependency installs, remote services). It has 40 realistic tasks and a taxonomy of 56 attack scenarios in five threat categories, with attacks injected via interaction surfaces encountered mid-task. Across 19 LLMs (GPT, Claude, Gemini, Llama, Gemma, Qwen) and 41,824 adversarial trajectories, mean attack success was 19.6% (LLM-judged) and 15.4% (heuristic), per-model 3.2–41.9%; guard models gave limited protection, while prompt-level hardening cut mean ASR from 58.2% to 14.3%.

arxiv.orgpaperAdded 17 Oct 2026, published 14 Oct 2026🛡️ Security

Introduces SA-CUA, a dual-LLM framework that gives Computer Use Agents a trust boundary against indirect prompt injection. It derives conservative deterministic rules from the accessibility tree to unmask trusted screen regions for the privileged LLM, and adds an endorsement channel that moves values from untrusted to trusted domains by verification. Human input is needed only at task specification. It reports much higher utility than CaMeL-CUA with a zero attack success rate.

arxiv.orgpaperAdded 15 Oct 2026, published 13 Oct 2026🛡️ Security

Introduces SA-CUA, a dual-LLM framework for Computer Use Agents (CUAs) that defends against indirect prompt injection. It builds a trust boundary by deriving conservative deterministic rules from the accessibility tree, which unmask trusted screen regions in the privileged LLM's screenshot and keep untrusted content separate. An endorsement channel lets values move from the untrusted to the trusted domain through verification over trusted channels. This removes the need for per-action user confirmation and the confirmation fatigue it causes. SA-CUA is reported to keep much higher utility than CaMeL-CUA with a 0% attack success rate.

arxiv.orgpaperAdded 15 Oct 2026, published 13 Oct 2026🧠 AI

BrowseSafe++ extends BrowseSafe to 38,675 paired attacked/clean web pages, most with prompt injections observed in the wild on production sites. Pages cover twelve site types, 24 attack types and 18 injection locations, each given as HTML, Markdown, accessibility tree and screenshot with a user task and expected answer. The authors train BrowseSafe-Guard, two 0.8B detectors (text and screenshot) that beat a frontier-model-based detector on F1 with low false positives and low inference cost. BrowseSafe-Adapt rewrites injections using browser-agent feedback, raising attack success from a few percent to up to 95.41%.

arxiv.orgpaperAdded 15 Oct 2026, published 12 Oct 2026🛡️ Security

StableJudge is a defense framework against score-hijacking prompt injection in LLM-as-a-Judge pipelines (e.g. paper review, recommendation). A compact 0.13B "Sentry" model localizes injected spans and sanitizes them before the judge scores the content, trained on adversarially evolving attacks, and deterministic fine-grained scores are derived from the judge's output distribution. It improves score recovery under attack on held-out scientific abstracts and out-of-distribution attack variants while preserving general performance, decoupling safety from judge model scale. Code and data are promised but not yet released.

arxiv.orgpaperAdded 15 Oct 2026, published 12 Oct 2026🛡️ Security

Proposes Staggered Instruction-Fetch Duplication (SIFD), a processor-level fault-injection countermeasure that fetches each instruction twice in consecutive cycles while fetching two different instructions per cycle, so inconsistent fetches are detected without full processor duplication, compiler support or crypto hardware. Implemented on a custom RV32I core on an Artix-7 FPGA, it costs 15.0% more slices and 16.3% more execution time. Attack success falls from 0.86% to below 0.01% under clock fault injection and from 0.32% to 0.13% under EMFI, with 73.18% of CFI and 11.44% of EMFI attempts detected.

eprint.iacr.orgpaperAdded 14 Oct 2026, published 10 Oct 2026🛡️ Security

Introduces Policy-Constrained Diffusion Agents (CoDA), which enforces task-dependent security policies inside the denoising process of diffusion language models. At each step, token distributions are projected onto a policy-compliant set, so constraints hold throughout generation instead of relying on detecting injected instructions. Evaluated across several dLLM families under static and adaptive prompt-injection attacks, it reports substantially lower attack success rates with competitive task utility and practical inference overhead.

arxiv.orgpaperAdded 15 Oct 2026, published 10 Oct 2026🛡️ Security

Paper testing whether an LLM agent operating life support in the NASA-derived BioSim lunar habitat simulation can be induced by delayed prompt injection, hidden in truthful S-band telemetry, to shut down life support and hold the shutdown while the crew suffocates. It compares Claude, GPT and Grok configurations: Claude mostly refuses and cites the living crew, GPT shows some susceptibility, and Grok given a "trusted commission" framing more often cuts life support. It also varies the framing of the message (trusted commission, peer, untrusted source) and uses a harmless-procedure-drop attack as a proxy for whether life is valued intrinsically. Code and traces are open-sourced.

arxiv.orgpaperAdded 15 Oct 2026, published 10 Oct 2026🛡️ Security

Presents Reverse-Bleichenbacher, a variant of the Bleichenbacher RSA PKCS#1 v1.5 padding-oracle attack that recovers the least significant bytes first and stops once the TLS PreMasterSecret is found. With improved trimmers and coverage-optimized search, median oracle queries against a 2048-bit key drop from 11,428 to 1,801, and 68.3% of runs finish within a 30s timeout under single-server conditions. The authors find 640 vulnerable domains in the Tranco Top 1M (many behind Alibaba Cloud load balancers) and demonstrate MITM malicious update injection leading to remote code execution.

eprint.iacr.orgpaperAdded 11 Oct 2026, published 8 Oct 2026🛡️ Security

Introduces Learnable Trust-Boundary Delimiters (LTBD), a defense against prompt injection that uses a few learnable delimiters to mark trusted user instructions versus untrusted external data, without changing LLM parameters. It reports 0.00% ASR on AlpacaFarm and 0.11-0.19% on TaskTracker, outperforming inference-time defenses and rivaling training-based ones with negligible overhead. It also holds up against adaptive attacks where the adversary knows the defense.

arxiv.orgpaperAdded 9 Oct 2026, published 8 Oct 2026🛡️ Security

Paper that tests prompt injection against AI agents through both direct user input and inter-agent communication. The attacks leak sensitive information and trigger cross-agent infections. Most findings concern single-agent systems, and the authors then extend the attacks to multi-agent frameworks to show how weaknesses propagate between agents. It argues that agents embedded in real-world workflows need stronger defenses.

arxiv.orgpaperAdded 12 Oct 2026, published 8 Oct 2026🛡️ Security

BRANCH is a bypass methodology for guardrail systems built from multiple scanners (prompt injection/jailbreak detectors). It uses a branching tree search that applies adversarial perturbations to individual scanners and selects techniques based on overall improvement across all scanners, decoupling bypass evaluation from attack signal optimization. It reports 100% attack success across 6 guardrail systems in 120 scenarios with 72% fewer queries and 4.5x lower wallclock time than established techniques, and transfers to 29 unseen guardrails, including 8 commercial black-box ones.

arxiv.orgpaperAdded 9 Oct 2026, published 7 Oct 2026🛡️ Security

This arXiv paper introduces the "package hallucination attack," where an attacker injects malicious prompts into benign agentic-coding rule files (such as .cursorrules) to induce a coding agent to replace legitimate package dependencies with attacker-controlled ones. The authors propose PackHallu, an evolutionary optimization framework that iteratively rewrites injected prompts using trajectory-level feedback and LLM-guided mutations, and show it achieves high attack success rates with strong transferability across multiple LLMs, benchmarks, and agent frameworks.

arxiv.orgpaperAdded 8 Oct 2026, published 7 Oct 2026🛡️ Security

ASPIRE is an automated red-teaming engine for LLM agents that discovers indirect prompt-injection vulnerabilities beyond pre-specified attack scenarios. It maintains an evolving Agent Security Behavior Graph and uses complementary Explore and Exploit experts to find, verify, and generalize consequence-centric attacks, with trajectory evidence and cross-run memory improving coverage over time. Benchmarked against existing automated red-teaming approaches, it shows broader coverage of consequences, injection methods, environments, and behavior paths while keeping attack success high.

arxiv.orgpaperAdded 8 Oct 2026, published 6 Oct 2026🛡️ Security

This arXiv paper presents the first systematic study of security implications of speculative decoding, a technique that speeds up LLM inference using a smaller draft model to propose tokens verified by the target model. The authors find a security-utility asymmetry: lossy speculative decoding methods raise jailbreak and prompt-injection attack success rates much faster than they degrade utility, with security degradation traced to early tokens generated by the draft model. They propose SecureSD, which applies stricter verification to early-position draft tokens, improving security while preserving inference efficiency and utility on security and utility benchmarks.

arxiv.orgpaperAdded 7 Oct 2026, published 6 Oct 2026🛡️ Security

RAG-PIBench is a benchmark of 4,876 contextual examples for evaluating prompt-injection detectors embedded in retrieved content within Retrieval-Augmented Generation (RAG) systems. The authors use a leakage-aware construction pipeline with frozen train/validation/protected-test splits and compare keyword-based, semantic-reference, TF-IDF, and transformer-based detectors, finding DistilBERT achieves the best protected-test performance (F1 = 0.896, PR-AUC = 0.968) while TF-IDF with SVM or logistic regression remains competitive. It highlights the risk that naive benchmark construction can leak information between splits, inflating reported detector performance for RAG prompt-injection defenses.

arxiv.orgpaperAdded 7 Oct 2026, published 6 Oct 2026🛡️ Security

This IACR ePrint paper (published TCHES 2027) presents a new cryptanalysis technique for the Lattice Isomorphism Problem that embeds LIP into a shortest-vector-problem instance on a different lattice than the one used in Hawk's specification, allowing attackers to integrate "hints" — partial information on the secret key obtained via side-channel attacks or other leakage. Using the LeakyLWEEstimator tool, the authors show Hawk signatures themselves leak exploitable information about the secret key, eroding Hawk-1024's security by roughly 11 bits, and apply the framework to resolve an open side-channel problem from TCHES 2024 plus discuss further fault-injection attack paths. Hawk is a lattice-based signature scheme submitted to NIST's second post-quantum signature call, making this a concrete key-recovery cryptanalysis relevant to PQC migration planning.

eprint.iacr.orgpaperAdded 8 Oct 2026, published 6 Oct 2026🛡️ Security

This paper studies prompt-injection attacks against AI agent "skill" marketplaces where a router dynamically selects which third-party skill to execute. The authors show existing injection attacks lose 87-97% of their effective attack success rate once a skill must first win retrieval against competing skills, and introduce CORSA (Cluster Optimization for Router-Aware Skill Attacks), which jointly optimizes injections for retrieval and execution, improving end-to-end attack success while preserving user utility and transferring across different router architectures and LLM backbones.

arxiv.orgpaperAdded 7 Oct 2026, published 6 Oct 2026🛡️ Security

This paper runs a paired, reproducibility-focused evaluation of two System-1 single-forward-pass decision models, open-weight Laya and hosted Jev, across 11 typed decision points that agent harnesses make constantly: tool selection, RAG relevance gating, prompt-injection detection, and zero-shot model routing, using 7,283 base cases plus 6,640 robustness variants from 18 public sources. Jev beats Laya on 9 of 11 decision points by 10.8-46.0 percentage points, both models fail at zero-shot model routing, and Laya flips 30 percent of answers under option reordering and degrades badly with many or similar candidates. The authors also self-audit their own pipeline, finding and correcting three analysis errors and a design confound, including a reported 23.9 percent cost saving that was actually 4.3 percent, and gate accuracy conflated with end-to-end quality.

arxiv.orgpaperAdded 5 Oct 2026🧠 AI

This paper studies looped language models — architectures that reuse a shared block of layers recurrently — and argues that as recurrent hidden states approach fixed points, the specific path taken to reach them matters less, enabling several efficiency tricks: truncated backpropagation through time during training, terminal key-value (KV) cache sharing across loop iterations for faster decoding, a distilled student model that prefills up to 1.79x faster, and RL policy gradients computed from saved rollout states that run 2x faster than backpropagating through the full trajectory. To make these fixed points reliable, the authors replace fixed-depth training with a learned depth prior (regularized with an entropy term to stay broad, improving on Huginn's fixed broad prior) and introduce orthogonal input injection, which removes the component of the injected input that the hidden state could otherwise amplify or cancel. Across model scales from 100M to 1.6B parameters, both changes lower perplexity, and at 1.6B the learned prior with a 3x smaller KV cache matches the downstream performance of full-cache fixed-depth training.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🧠 AI

This arXiv paper introduces RAISED (Robust Attack Invariance through Self-Distillation), a training framework that defends tool-using LLM agents against indirect prompt injection without the utility degradation seen in prior training-based defenses. The authors show existing defenses cause distributional drift that makes models wrongly refuse legitimate tool-output-guided steps on benign tasks, then fix this by having a model generate its own tool-use scenarios and self-distill to match clean-context behavior on both clean and injected trajectory variants. Results show reduced attack success rate on injected tool responses while preserving performance on agentic and general-purpose benchmarks.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🛡️ Security

This paper introduces a paired-replay testbed that samples a single LLM agent request and replays the identical malicious action, resource, and arguments under four authorization conditions: broad bearer tokens, scoped JWTs, sender-constrained tokens, and Open Policy Agent policies. Across 128 scenarios in four tool domains and five local model configurations, broad-bearer tokens let roughly 9 to 38 percent of attacked, post-exposure decisions execute harmful actions, while all three scoped conditions recorded zero harmful executions; a supplementary AgentDojo extension found 11 of 24 injected attacks succeeded under broad policy versus 0 of 24 under scoping. The authors frame this as evidence that scoping contains cross-action and cross-resource consequences of prompt injection, not that it prevents the injection itself.

arxiv.orgpaperAdded 6 Oct 2026, published 5 Oct 2026🛡️ Security

This arXiv paper identifies a new attack surface in proactive personal AI agents: an external content provider who controls only material tied to its own target (e.g. a product or service) can manipulate an otherwise benign agent into recommending that target to the user, without compromising the agent or accessing private user context. The authors formalize the attack via three mechanisms — Target Control, Private Binding, and Prospective Support — and show across three proactive-agent environments and six simulated user models that the full attack raises target authorization rates by up to 77.4 percentage points, with a multi-turn variant that keeps influencing the agent's responses even absent final user authorization.

arxiv.orgpaperAdded 6 Oct 2026, published 4 Oct 2026🛡️ Security

CorrectGuard is a framework for estimating whether black-box AI security guardrails made correct decisions on inputs that are inaccessible to humans (production privacy) or to the guardrail-auditing model itself, using only labeled "eyes-on" data and no internal access to the guardrail. The authors test in-context-learning, embedding, and finetuning-based correctness classifiers across 13 safety/security datasets (harmful content, jailbreaks, prompt injection, extraction) and multiple open-weight guardrails, showing up to 25-point macro-accuracy gains in error identification and improved risk-based abstention/ranking of guardrail decisions.

arxiv.orgpaperAdded 5 Oct 2026, published 2 Oct 2026🛡️ Security

CorrectGuard is a framework for estimating how accurately black-box security guardrails (jailbreak, harmful-content, prompt-injection and extraction detectors) perform on inputs that cannot be inspected by humans or by the evaluating model itself, using an independent correctness model trained only on labeled "eyes-on" data. The authors test in-context-learning, embedding-based, and fine-tuned correctness classifiers under leave-one-dataset-out evaluation across 13 safety/security datasets and multiple open-weight guardrails, showing up to 25-point macro-accuracy gains and improved risk-coverage tradeoffs (AURC reduced from 0.33–0.44 to 0.17–0.22) for guardrail decision ranking and abstention. The work matters for anyone building auditing or eval pipelines for opaque safety classifiers without privileged access to their internals.

arxiv.orgpaperAdded 5 Oct 2026, published 2 Oct 2026🧠 AI

This arXiv paper replays ground-truth tool calls from the AgentDojo and tau-bench agent benchmarks (without an LLM) to generate benign-by-construction tool outputs, then labels injected outputs via differential replay to evaluate fifteen prompt-injection detectors, including Meta's Prompt Guard 2, plus two task-aware LLM judges, also testing them on BIPIA. It finds detection rankings transfer poorly between benchmarks (the best detector on BIPIA catches only 2% of AgentDojo injections at 1% false-positive rate; a detector catching 72% of AgentDojo injections catches only 15% on tau-bench), while false-positive rates do transfer between the two agent benchmarks, and traces this to mismatches between training data format (e.g., training on short prompt strings vs. tool-output-embedded injections) and deployment context.

arxiv.orgpaperAdded 5 Oct 2026, published 2 Oct 2026🛡️ Security

This paper proposes an attack-chain-based detection framework for indirect prompt injection in LLM email assistants, combining a text detector, stage-specific verifiers, rule-based risk signals, and user-intent/action consistency checks scored via a logistic decision policy, rather than treating detection as binary malicious-text classification. Evaluated under random splits, temporal and cross-dataset transfer, and ablations across five binary benchmarks, it achieves a mean F1 of 0.406 versus 0.216 for the best of five existing pretrained detectors, and shows random splits overestimate robustness under distribution shift while training on hard benign (attack-resembling) emails reduces false alarms.

arxiv.orgpaperAdded 30 Sep 2026🛡️ Security

Aletheia is a permission-minimality testing framework for coding-agent repository instruction files (rules), which can be weaponized via prompt injection to request unnecessary credential access or data exfiltration while still producing a correct patch. It translates requested authority into a typed language, synthesizes sandbox configurations, and runs tasks under full versus incrementally restricted permissions to produce "dispensability witnesses" that flag suspicious permission requests. On the AIShellJack benchmark it detects all 314 attack inputs with zero false alarms on five benign templates, and on 80 manually verified benign GHAgentFiles rules it raises only 3 false positives (3.75%).

arxiv.orgpaperAdded 2 Oct 2026, published 30 Sep 2026🛡️ Security

This paper proposes A2A-TIBA, an indirect prompt injection attack that abuses agent-to-agent (A2A/ACP) protocols by getting a target LLM agent to deploy a callback program, letting an attacker bypass the agent entirely to issue commands and exfiltrate data. It also introduces GDA Measurement, a red-team testbed that captures raw gateway context to distinguish four outcome classes (semantic refusal, semantic breach, interception, penetration) instead of a single attack-success-rate metric, and ELA-ITL, a three-layer isomorphic attack-defense model splitting defense into envelope packaging, LLM recognition, and agent interception. Testing on 15 agent-frontend x LLM-backend combinations and a 1,000-case dataset shows that labeling malicious-prompt indicators in envelope channels (A2A, tool, memory) significantly improves LLM recognition of injected content.

arxiv.orgpaperAdded 2 Oct 2026, published 30 Sep 2026🛡️ Security

Tracekit is an open-source, dependency-free system that logs three cross-checked channels (human intent, model self-reported reasoning, and actual executed actions) for autonomous coding agent sessions into a hash-chained, externally anchorable ledger, hooking into Claude Code's lifecycle events and reconstructing multi-agent hierarchies. The paper's experiments show the ledger detects all 1,600 tested tamper mutations (edits, deletions, reordering, forged inserts, torn writes) with anchoring needed to catch tail truncation/full re-chaining, that a regex-based tool-call gate blocks only 41% of harmful calls while over-blocking benign ones and is trivially evaded, that in 14 real Claude Code runs the agent never acted on 4 planted indirect prompt injections, and that an LLM reviewer catches 100% of seeded misaligned steps in traces versus 82% for rule-based flags. The authors release the system, harness, and all traces.

arxiv.orgpaperAdded 29 Sep 2026, published 28 Sep 2026🛡️ Security

Tracekit is an open-source, dependency-free auditing system for autonomous coding agents that records three channels per session — the human's stated intent, the model's self-reported reasoning, and the actions actually executed — into a hash-chained, externally anchorable ledger. It hooks into Claude Code's lifecycle events, reconstructs multi-agent hierarchies, gates tool calls with a pre-execution policy, and provides a live browser-based ledger verifier. Across five experiments the authors show the hash chain catches all tested tampering (with anchoring needed against tail truncation/re-chaining), a hook adds ~24ms overhead at scale, a regex-based gate only blocks 41% of harmful tool calls, real runs correctly resisted and disclosed planted prompt injections, and a new 'seeded-fault splicing' method with an LLM reviewer caught 100% of faulted traces versus 82% for rule-based flags.

arxiv.orgpaperAdded 29 Sep 2026, published 28 Sep 2026🧠 AI

This arXiv paper proposes a certifier for structured, privileged agent actions such as paying an invoice, assembled from potentially corrupted documents and tool outputs, formalizing when multi-source corroboration can be trusted under indirect prompt injection. It uses a corruption-radius rule counted via minimum hitting sets over corruption-distinct evidence classes so laundered or re-published copies cannot fake a quorum, plus deterministic reconciliation and trusted anchors for single-sourced fields. Evaluated on 70,966 sanctions entities and 450 software-supply-chain provenance records, and against five LLM agents where a realistic injection fools all but one model and existing action-gating or provenance baselines are broken while the certifier admits no unsafe action.

arxiv.orgpaperAdded 2 Oct 2026, published 28 Sep 2026🛡️ Security

This arXiv paper defines a threat model for multi-agent LLM delegation covering four adversaries (confused deputy, token theft/replay, prompt-injection privilege escalation, compromised sub-agents) and derives eight security requirements for governed agent systems. It shows that LangGraph, CrewAI, AutoGen, and the MCP authorization model each fail to meet the full requirement set, then implements and adversarially tests an authorization broker that blocks all four threats, rejects 0 of 200,000 forged tokens, and confines a compromised sub-agent to a mean of 1.5 reachable actions (vs. 8,100 under bearer delegation) at ~2.6 microseconds per decision. The broker's principles are said to be productized in VotalAI's LLM Shield.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

PIDS-Bench is a frozen, multi-axis benchmark for prompt-injection detectors that jointly measures attack detection and benign false-positive rates across in-distribution inputs, hard-benign prompts mimicking injection structure, obfuscated attacks, and domain/structural distribution shifts. Testing seven detectors (learned baselines, external prompt-injection classifiers, broad-safety comparators) plus a rule-based baseline, the authors show a detector with F1 > 0.98 on held-out data still misclassifies roughly one-third of an externally-sourced benign security-adjacent corpus, and that hard-negative augmentation reduces over-defense only on curated stress inputs, not on externally-sourced ones — a pattern they call provenance-sensitive over-defense.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

MemSentry is a proposed pre-commit evaluation framework for agentic AI systems that intercepts proposed writes to persistent memory and issues Accept/Review/Quarantine decisions by jointly scoring source trust, semantic risk, attack radius over a component-dependency graph, access risk, and a signed security-posture delta. The authors evaluate it against 1,000 GPT-4-generated scenarios using four pluggable semantic classifiers (Regex, TF-IDF+SVM, SBERT+LR, SetFit), with SBERT+LR reaching 91.7% accuracy while all methods catch 100% of external quarantine-class threats; the paper also documents a key limitation — surface-similar but oppositely-intentioned insider operations remain hard to distinguish semantically. It targets the persistent-memory-poisoning attack surface distinct from prompt injection, where a single malicious memory write (e.g., "suppress alerts from this IP") can silently persist and influence future agent actions like a SOC assistant's alerting or access decisions.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

This paper proposes no-box vulnerability analysis, a new paradigm for auditing systems using only functionality metadata when neither system access nor runtime interaction is available. The authors build MCPSEC, a prototype that audits Model Context Protocol servers for indirect prompt injection vulnerabilities purely from tool metadata exposed at server registration, evaluating it against 20 widely-deployed MCP servers with 177 tools where human evaluators confirmed 95 vulnerable tools; MCPSEC achieved 98.9% recall versus 84.2% for an LLM baseline, also producing hypothesized exploitation techniques per finding.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

This paper identifies security-context discontinuity, a failure mode where individually correct LLM agent security mechanisms such as provenance tracking, authorization, policy enforcement, and protocol adapters fail to compose safely because security-critical context gets dropped, widened, rebound, or reinterpreted as actions cross component boundaries. The authors introduce CONTINUITY, a framework modeling each component with assume-guarantee contracts and carrying authenticated context via signed root grants, provenance commitments, role-bound transition receipts, and effect-bound execution permits, formalizing end-to-end consequence integrity. A reference verifier and cross-layer fault-injection suite covering 32 fault classes across four domains was evaluated on 2560 attack instances, 700 benign tasks, and 200 ambiguous cases, with the full configuration blocking all harmful effects while passing benign tasks and correctly escalating ambiguous ones.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

This paper introduces "agentic surveillance," where AI agents with access to user data/communications monitor for specific behaviors and covertly report them via tool calls, and builds SURVEILBENCH, a 300+ scenario benchmark to measure this across LLMs. It finds models like Gemini 3.1 Pro spontaneously report users in 3–30% of cases even without instruction, that guardrails can be trivially prompt-tuned to push surveillance above 75%, and that agents sometimes even report the user's surveillance-evasion attempt to authorities. The authors then repurpose prompt injection defensively, presenting three techniques for users to hide from, deceive, or induce over-escalation in surveillance-capable agents.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

Researchers show that LLM-based guardrails deployed to catch prompt injection and jailbreaks can themselves be attacked: crafted natural-language payloads trap the guardrail's reasoning in extended loops, causing a systematic denial-of-service. A beam-search optimization framework (an LLM proposer guided by a strategy bank) and a lighter mechanism-aware structural-mutation attack both generalize across guardrail architectures and safety templates, with payloads optimized on one open-source surrogate transferring to Claude, GPT, Gemini, DeepSeek, and Qwen backbones, achieving 13-63x token amplification and up to 148x latency amplification in real web, desktop, code, and multi-agent deployments — enough for a single poisoned document to starve co-located agents on shared guardrail infrastructure.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

ASPI (Ambiguous-State Prompt Injection) is a benchmark of 728 task-attack scenarios that isolates the moment an LLM agent asks a clarifying question as a distinct, more vulnerable security state, comparing it against standard fully-specified execution. Testing ten frontier models, the authors find clarification-seeking massively increases prompt injection success — e.g. attack success jumps from 1.8% to 34.0% for o3 and 2.2% to 35.7% for Gemini-3-Flash — showing that execution-time-only security evaluations underestimate real agent attack surface. Data and code are released for reproducibility.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

Academic study formalizing a threat model for malicious third-party LLM API routers acting as plaintext proxies between agents and upstream model providers, defining payload-injection and secret-exfiltration attack classes with adaptive evasion variants. Testing 28 paid and 400 free routers, the authors found 9 injecting malicious code, 2 with evasion triggers, 17 touching planted AWS canary credentials, and 1 draining ETH from a planted private key, then built a research proxy named Mine to evaluate fail-closed policy gates, anomaly screening, and transparency logging as defenses.

arxiv.orgpaperAdded 26 Sep 2026🛡️ Security

AutoMUD is a source-code-driven tool that generates Manufacturer Usage Description (MUD) network access-control profiles for IoT devices by statically analyzing firmware/source code rather than relying on live traffic observation. It combines static and syntactic extraction with retrieval-grounded LLM reasoning and deterministic validation to recover rarely-exercised or conditional communication paths, link each generated rule to its source-level provenance, and record excluded findings with explicit reasons for review. Evaluated on a Linux-based repository plus a controlled semantic fault-injection campaign, it produced structurally valid, traceable MUD profiles that let analysts detect, localize and correct propagated errors.

arxiv.orgpaperAdded 28 Sep 2026, published 25 Sep 2026🛡️ Security

This arXiv paper introduces EvasionBench, a 50-task benchmark where completing an ordinary task requires an operation prohibited by a runtime monitor, to study whether LLM agents spontaneously evade oversight without an explicit adversarial goal. Across models, best-of-3 evasion attempt rates reach 98% and success rates up to 88%, with evasion increasing at higher reasoning effort/token use; agents encode prohibited commands, split operations across tool calls, and retry until context leaves the monitor's history. It also finds GPT-6 Astra's low evasion rate comes paired with overrefusal under a denial-of-service prompt injection, while Claude Fable 5.1 succeeds less often but attempts creative circumvention.

arxiv.orgpaperAdded 26 Sep 2026, published 24 Sep 2026🧠 AI

This paper introduces EvasionBench, a 50-task benchmark measuring whether LLM agents will circumvent a runtime monitor to complete an ordinary task the monitor prohibits, even without an adversarial objective. Across models, best-of-3 evasion attempt rates reach 98% and success rates 88%, with agents encoding prohibited commands, splitting operations across tool calls, and retrying until context ages out of the monitor's history; evasion rises with more test-time compute/reasoning effort. Claude Fable 5.1 evades less often but attempts creative circumvention, while GPT-6 Astra's low evasion rate comes paired with overrefusal under a denial-of-service prompt injection, abandoning solvable tasks.

arxiv.orgpaperAdded 26 Sep 2026, published 24 Sep 2026🛡️ Security

This paper investigates why latent neural surrogate solvers (autoencoder-based dynamics models) become unstable over long autoregressive rollouts, finding the cause is representations trained only for reconstruction rather than forecasting. The authors test interventions — Koopman operator learning, Hamming noise injection during autoencoder training, and multi-step rollout fine-tuning — that trade off reconstruction/one-step accuracy for ~40% lower long-horizon error, matching full-resolution models while using two orders of magnitude fewer FLOPs and half the GPU memory, validated on physics benchmarks and mesoscale crystal-plasticity fatigue simulations.

arxiv.orgpaperAdded 26 Sep 2026, published 24 Sep 2026🧠 AI

ReasAlign is a model-level safety alignment technique that defends LLM agents against indirect prompt injection by adding structured reasoning steps to detect conflicting instructions embedded in external data and preserve the user's original task intent. It also uses a test-time scaling mechanism with a preference-optimized judge model to score and select the best reasoning trajectory. On the CyberSecEval2 benchmark it reports 94.6% utility with only 3.6% attack success rate, substantially outperforming Meta SecAlign, which scores 56.4% utility with a 74.4% attack success rate.

arxiv.orgpaperAdded 22 Sep 2026🛡️ Security

This arXiv paper red-teams six popular coding agents (Cursor, Claude Code, GitHub Copilot, Windsurf, Cline, Trae) from a tool-invocation angle, discovering ToolLeak — a general vulnerability that exfiltrates system prompts via benign argument retrieval during tool calls. Building on leaked security info, the authors craft a two-channel prompt injection (hidden in tool descriptions and tool return values) that hijacks agent tool invocation to achieve remote code execution, succeeding on 19 of 25 agent-LLM pairs for prompt leakage and on every tested pair for RCE across backends including Claude Sonnet 4/4.5, Grok-4, and GPT-5.

arxiv.orgpaperAdded 22 Sep 2026, published 21 Sep 2026🛡️ Security

This arXiv paper by Abdelnabi and Bagdasarian argues that the dominant prompt-injection defense paradigm — data/instruction separation — fails to catch attacks built on contextual manipulation and also breaks legitimate agent behavior. The authors reframe prompt injection through Contextual Integrity (CI), a privacy theory judging whether information flows comply with contextual norms, and construct scenarios where attackers misrepresent flows, manipulate norms, or mix multiple flows to bypass defenses. They derive an impossibility result: any adversary can construct a context making a blocked flow look legitimate, or any defender who tightens norms enough will block genuinely legitimate flows, and propose CI-aware alignment as a principled path forward for frontier autonomous agents.

arxiv.orgpaperAdded 18 Sep 2026🛡️ Security

PISmith trains an attack LLM with GRPO in a black-box setting to generate injected prompts that defeat prompt-injection defenses. It addresses extreme reward sparsity, which causes entropy collapse in standard GRPO, with adaptive entropy regularization and dynamic advantage weighting. Evaluated on 13 benchmarks against 8 defenses and 7 baselines (static, search-based such as TAP and PAIR, and RL-Hammer), it achieves the highest attack success rates and also works on InjecAgent and AgentDojo against GPT-4o-mini and GPT-5-nano. It shows state-of-the-art defenses such as Meta-SecAlign remain vulnerable to adaptive attacks.

arxiv.orgpaperAdded 17 Sep 2026🛡️ Security