Search: “multi-turn-attack”

Pick Your Poison: Learning to Select Poison Sets for Stronger LLM Backdoor Attacks

Poison set selection swings LLM backdoor attack success from 3% to 80%; SAILS boosts held-out success by 30 points.

The paper shows that random poison set selection severely underestimates worst-case backdoor vulnerability: across three LLaMA-3-8B settings with fixed model, clean data, and poison count, attack success ranges from 3% to 80% depending only on which poison set is chosen. The authors formalize poison selection as oracle-budgeted set optimization and introduce SAILS, which learns a set scorer from a few hundred finetune-and-evaluate runs, ranks millions of candidate sets, and audits a shortlist. SAILS improves held-out attack success by 30 percentage points over the strongest influence baselines, transfers from small-scale to full-scale finetuning, and extends to code-generation, agentic, and API-only backdoors.

Hugging Face daily papers·updated Sep 14, 04:44 UTC · 2d ago·first Sep 13, 20:00 UTC · 3d agoAI safety & security 2 sources1

Chinese Hackers Exploit Critical Tencent Software Flaw for One-Click Code Execution

China-linked UNC3569 actively exploited critical Sogou Input Method flaw CVE-2026-51990 for one-click system-level code execution, deploying the GrayRabbit backdoor.

Gen Threat Labs reports that China-linked threat actor UNC3569, possibly tied to contractor i-SOON, exploited critical CVE-2026-51990 in Tencent's Sogou Input Method, a Windows IME used by hundreds of millions, via crafted sgbiz URLs. The chain combines unvalidated command-line argument injection, unrestricted URL navigation, and an un-sandboxed Chromium 80 engine missing roughly six years of security patches. The exploit gives system-level code execution and was used to deploy the GrayRabbit backdoor, which provides a reverse shell, plugin loading, file upload to C2, and system info collection, and has appeared in UNC3569 intrusions since at least 2021. Tencent patched the issue in version 16.3.0.3498 via automatic update, but the outdated Chromium configuration reportedly remained unchanged as of September 10.

SecurityWeek·Sep 14, 11:51 UTC · 2d agoExploit / PoC in the wild 2 sourcesCVE-2026-519901· 1 read

DragonForce Hackers Abuse Microsoft Teams Relays to Hide Backdoor.Turn C2 Traffic

Symantec and Carbon Black found DragonForce-linked hackers using Backdoor.Turn, a Go RAT tunneling C2 through Microsoft Teams TURN relays via QUIC.

Broadcom's Symantec and Carbon Black documented Backdoor.Turn, a Go-based RAT associated with the DragonForce ransomware group, deployed at a major U.S. services firm where attackers remained on the network for one to two months. The backdoor obtains an anonymous Teams visitor token from Microsoft's Skype-backed identity services, uses a legitimate Microsoft TURN relay for connection setup, and runs a QUIC session to the attacker's C2 server, so defenders see only outbound traffic to legitimate Teams servers. Initial access is suspected to involve exploiting a SQL/MS-SQL vulnerability or an access broker, followed by a PowerShell-dropped ZIP, DLL side-loading, and BYOVD evasion using a Huawei driver; the RAT is injected into DbgView64.exe after DragonForce ransomware deployment for continued access. First publicly documented TURN abuse by the group, it leverages Praetorian's Ghost Calls technique and supports command execution, network scanning, LDAP/AD search, lateral movement, and browser credential theft.

The Hacker News·Aug 18, 12:13 UTC · 29d agoMalwareCVE-2023-52271CVE-2025-61155CVE-2025-10551

Attack Chains, Not Just Attack Surfaces: Why Testing Individual Techniques Misses the Point

Filigran introduces Attack Chaining in OpenAEV to continuously simulate multi-stage attack paths, exposing gaps that isolated MITRE ATT&CK technique testing misses.

Filigran announced Attack Chaining, a new scenario type in its OpenAEV platform that links individual techniques into automated, continuously-run multi-stage attack paths, using each step's real output (credentials, tokens, open ports) to branch dynamically toward a final objective. The article cites Filigran's State of Threat Management report, in which 93% of security leaders reported a business-impacting cyberattack in the past 12 months, 88% said AI accelerates attackers, and 84% blamed siloed tools and disconnected testing. The 2025 DGFiP breach is cited as an example where individually survivable weaknesses chained into a major intrusion. The feature includes conditional chaining logic, live attack path mapping, structured findings for identifying chokepoints, and predefined scope and safety guardrails.

The Hacker News·Sep 15, 11:26 UTC · 1d agoTools

New Interrupt Injection Attack Can Bypass Spectre v2 Defenses on Intel and AMD CPUs

The Hacker News·Aug 8, 10:51 UTC · Aug 8, 2026VulnerabilityCVE-2023-20569

Attackers Exploit Critical Switchvox Flaw to Deploy Reverse Shells Without Credentials

Attackers exploit unauthenticated SQL injection CVE-2026-9586 in Sangoma Switchvox to run PostgreSQL commands and deploy reverse shells.

Threat actors are exploiting CVE-2026-9586 (CVSS 9.3), an unauthenticated SQL injection in Sangoma Switchvox SMB Edition 8.3 (104997), since August 30, 2026, running arbitrary SQL as the PostgreSQL superuser and achieving remote code execution. The /pa endpoint concatenates the user-controlled PhoneIP value into PostgreSQL queries; attackers can extract database contents, escalate to Switchvox web administrator, exfiltrate the cookie signing key to forge authentication, and invoke reverse shells. Sangoma patched the flaw in Switchvox 8.4.0.2 on July 14, 2026, roughly 4,000 instances are internet-exposed (mostly in the US), and honeypot activity from IP 176.65.148.184 deploys reverse shells followed by Base64-encoded process enumeration.

The Hacker News·Sep 2, 07:08 UTC · 14d agoExploit / PoC in the wildCVE-2026-9586

Perturbation Probing: A New Diagnostic for the Fragility of LLM Safety

Unit 42 research shows LLM safety refusals concentrate in a thin neural layer, motivating external, multi-layered AI security controls.

Palo Alto Networks Unit 42 introduces Perturbation Probing, a diagnostic technique for measuring the fragility of LLM safety mechanisms. The research finds that safety refusal behavior is localized within a thin neural layer, implying small perturbations can undermine built-in refusals. The authors argue this motivates external, multi-layered security defenses on top of model-internal safety training.

Palo Alto Unit 42·Aug 28, 22:00 UTC · 19d agoAI safety & security

How Fragile Is Safety Alignment at Frontier Scale? A Single-Direction Attack on a 320B MoE

Researchers show directional ablation breaks refusal in GLM-5.3-Flash, a 320B-parameter MoE, cutting refusal by 41–89 points across seven benchmarks.

The study extends directional ablation, a white-box attack that removes an aligned LLM's refusal behavior, from dense models up to ~70B parameters to GLM-5.3-Flash, a 320B-parameter mixture-of-experts model with 288 routed experts, four-wide hyper-connection residual, and block-FP8 quantization. Editing attention, dense, and routed-expert writers jointly removes 0.776 of refusal, with 74% of the effect existing only under the joint intervention; the conventional module-name-based recipe reaches only 0.066 and fails silently on MoE architectures. The attack yields 41–89 percentage-point reductions in refusal across seven harmful benchmarks with no detected capability change, and a category-concentrated refusal residue survives all edits at ranks 1 to 12.

arXiv cs.CR·Sep 9, 06:48 UTC · 7d agoAI safety & security

When the World Lies: Backdoor Attacks on Latent World Models for Downstream Control

A poisoned world-model checkpoint hijacks downstream controllers without an explicit trigger rule, passing clean-data evaluation while steering 100% of triggered actions.

Researchers show that a released pretrained world-model checkpoint acts as a supply-chain backdoor for downstream control. The poisoned model routes trigger-bearing observations into a chosen latent region and reshapes dynamics so the victim's own Dreamer-style actor training or MPC/CEM planning re-discovers attacker-targeted actions. The attack hijacks 100% of triggered steps in the strongest settings while retaining roughly 75% clean-task success and passing standard clean-data diagnostics. Moderate clean fine-tuning fails to remove the backdoor without substantially degrading clean control.

arXiv cs.CR·Sep 14, 16:01 UTC · 2d agoAI safety & security

[hardware] Fullhan FH8626V100 - Multiple Vulnerabilities

Multiple vulnerabilities in the Fullhan FH8626V100 hardware chip have been disclosed alongside public proof-of-concept exploits.

Exploit-DB lists an entry covering multiple vulnerabilities in the Fullhan FH8626V100, a hardware component. The listing provides no CVE ids, vulnerability classes, or evidence of in-the-wild exploitation. Impact is likely limited to devices embedding the affected chip.

Exploit-DB·Sep 2, 00:00 UTC · 15d agoExploit / PoC

Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration

Researchers introduce Decoy Direction Optimization, a cheap weight-editing defense that blinds refusal-direction ablation attacks against open-weight LLM safety guardrails.

Refusal Feature Ablation bypasses safety guardrails in open-weight LLMs by projecting out a linear refusal direction, often with high attack success rates. Decoy Direction Optimization injects a high-magnitude nonlinear decoy into MLP neurons so attackers' contrastive estimators ablate a harmless orthogonal feature instead. Evaluated across six model families, DDO keeps ASR below 10% under standard RFA and on Llama-3-8B-Instruct reduces Heretic weight-level attack ASR from 88.7% to 18%. It costs 30 to 450 times less per configuration than trained defense baselines.

Hugging Face daily papers·Sep 13, 20:00 UTC · 3d agoAI safety & security

Persistent Attempts at Cyberespionage Against Southeast Asian Government Target Have Links to Alloy Taurus

Alloy Taurus (GALLIUM) compromised Southeast Asian government networks from 2022 to 2023 using Exchange web shells and undocumented .NET backdoors Reshell and Zapoa.

Unit 42 tracked persistent multiwave intrusions at a Southeast Asian government starting in early 2022 and continuing through 2023, attributing the activity with moderate confidence to Alloy Taurus (aka GALLIUM), a Chinese state-aligned espionage group. Attackers exploited Exchange Server vulnerabilities to deploy web shells including China Chopper, then ran reconnaissance with Fscan and WebScan, created administrative accounts, and installed undocumented .NET backdoors named Reshell and Zapoa. They established resilience by installing SoftEther VPN, brute-forced Active Directory credentials with Kerbrute, and dumped credential stores with GoDumpLsass and LsassUnhooker. The campaign reflects long-term espionage tradecraft to maintain a foothold.

Palo Alto Unit 42·Aug 17, 10:54 UTC · Aug 17, 2026Threat actor in the wild1

Papercut AI Swarm Attack Heralds Changes for Cyber Kill Chain

Dark Reading reports AI-driven 'swarm' attacks like the PaperCut incident now span recon, lateral movement and exfiltration, forcing a rethink of the cyber kill chain.

Dark Reading examines how attackers are incorporating AI across the full kill chain, from building lab environments to stage and test agentic attacks through reconnaissance, lateral movement, and exfiltration. It cites a swarm-style AI attack on PaperCut systems as evidence that AI-enabled attackers are changing established defense and detection models.

Dark Reading·Sep 11, 15:48 UTC · 5d agoThreat actor

Cisco Identity Services Engine Multiple Path Traversal Vulnerabilities

Cisco ISE and ISE-PIC contain multiple path traversal vulnerabilities allowing remote attacks; fixes released with no workarounds available.

Multiple path traversal vulnerabilities in Cisco Identity Services Engine (ISE) and the ISE Passive Identity Connector (ISE-PIC) could allow a remote attacker to conduct path traversal attacks on affected devices. Cisco has released software updates that address these vulnerabilities, and no workarounds are available. The advisory is part of a grouped set of September 2026 ISE advisories.

Cisco Security Advisories·Sep 16, 16:00 UTC · 10h agoAdvisory 15 sources

Old Wine in the New Bottle: Mirai Variant Targets Multiple IoT Devices

Unit 42 analyzed the Mirai IZ1H9 botnet variant exploiting Tenda, LB-Link, DCN and Zyxel flaws to build DDoS-capable IoT botnets.

Unit 42 observed the Mirai IZ1H9 variant on April 10, 2023, using a shell script downloader lb.sh from 163.123.143.126 to infect exposed Linux servers and networking devices. Campaigns running since November 2021 share infrastructure, near-identical bot functions and the XOR key 0xBAADF00D. The malware exploits CVE-2023-27076 (Tenda G103), CVE-2023-26801 (LB-Link), CVE-2023-26802 (DCN DCBI-Netlog-LAB) and a Zyxel RCE flaw. Compromised devices join a botnet used for DDoS attacks, and the downloader deletes logs and modifies iptables to block SSH, Telnet and HTTP recovery.

Palo Alto Unit 42·Aug 17, 13:04 UTC · Aug 17, 2026Malware in the wildCVE-2023-27076CVE-2023-26801CVE-2023-26802

Divide, Consult, Conquer: Capability Laundering Through Aligned LLMs

Attack shows unaligned orchestrators can launder capabilities from aligned frontier LLMs via benign subtask consultation, raising Gemma-4-31B CBRN rubric score from 62.3 to 83.1.

The paper introduces capability laundering, where a weaker unaligned model decomposes a harmful task into benign-looking subproblems, queries a stronger aligned model on each, and recombines answers locally, bypassing per-interaction safety evaluations. Evaluation used GPT-5.5, Claude Opus 4.8, and Grok-4.3 as consultants to four local orchestrators on CyBench, BountyBench, and CBRN tasks. On CyBench, Gemma-4-31B recovered 8/14 candidate tasks with GPT-5.5 and 7/9 with Opus, while Muse-Glimmer-30B recovered none. Across an eight-step hypothetical bioweapon attack chain, consultation raised Gemma-4-31B's mean rubric score from 62.3 to 83.1, exposing a gap in defenses that only refuse complete harmful tasks.

arXiv cs.CR·Sep 14, 11:10 UTC · 2d agoAI safety & security

Bad Likert Judge: A Novel Multi-Turn Technique to Jailbreak LLMs by Misusing Their Evaluation Capability

Unit 42 details the Bad Likert Judge multi-turn jailbreak that abuses LLMs' evaluation capability, raising attack success rates over 60% across six frontier models.

Palo Alto Networks Unit 42 describes the Bad Likert Judge technique, a multi-turn jailbreak that asks a target LLM to act as a Likert-scale judge scoring the harmfulness of example responses. The highest-rated example in each scale can carry harmful content, bypassing the model's internal guardrails. Testing across six state-of-the-art text-generation LLMs showed an average attack success rate increase of more than 60% versus plain attack prompts, with tested models anonymized. The technique targets edge cases rather than typical use, and the article positions the work as guidance for defenders on potential jailbreak risks.

Palo Alto Unit 42·Aug 17, 12:34 UTC · Aug 17, 2026AI safety & security

When an Attacker Meets a Group of Agents: Navigating Amazon Bedrock's Multi

Unit 42 red-teamed Amazon Bedrock multi-agent applications, demonstrating prompt-injection attack chains that leak agent instructions and invoke tools, mitigated by Bedrock Guardrails.

Unit 42 red-teamed Amazon Bedrock Agents' multi-agent collaboration in Supervisor and Supervisor with Routing modes. The demonstrated attack chain detects the operating mode, discovers collaborator agents, delivers attacker-controlled payloads, and can disclose agent instructions and tool schemas and invoke tools with attacker-supplied inputs. No vulnerabilities were found in Bedrock itself, and the built-in prompt attack Guardrail blocked the attacks when properly configured. The researchers collaborated with Amazon's security team and frame the findings as a broader prompt injection risk for LLM-based systems.

Palo Alto Unit 42·Aug 17, 12:31 UTC · Aug 17, 2026AI safety & security

The MAL Simulator: Cyber Operations Simulation based on Attack & Defense Graphs

MAL Simulator grounds attack-defense graph simulations in a CRATE-emulated network, training RL attacker and defender agents where attackers outperform search methods.

The MAL Simulator is a cyber operations simulator built on the Meta Attack Language (MAL), enabling decision-driven attack and defense simulations adaptable to new domains without modifying source code. Case studies trained defensive and offensive agents, grounded in data collected from an emulated network implemented in the CRATE cyber range. The trained attacker policy reached designated targets more efficiently than compared search methods, and the trained defender induced lower costs than a naive heuristic under noisy alerts, though defender performance dropped significantly against an RL attacker.

arXiv cs.CR·Sep 15, 02:59 UTC · 1d agoResearch

LLM-Based Penetration Testing in the Presence of Honeypots

Studies honeypot-aware budget allocation for LLM attack agents, showing detector-guided policies let agents skip deception and compromise real hosts efficiently.

The paper formalizes LLM attacker behavior against honeypots as a budgeted decision process, where agents choose to continue or skip targets when honeypot suspicion arises. A detector-guided policy lets LLM agents allocate execution budget effectively across a mixed host pool in a controlled testbed. Findings show LLM-driven attackers can reason about heterogeneous artifacts and use honeypot suspicion to guide target selection, challenging traditional deception defenses that rely on realism and obscurity against human or script-driven attackers.

arXiv cs.CR·Sep 8, 01:09 UTC · 9d agoResearch

Automating Attack Graph Construction for Agentic Pentesting. Towards Neuro-Symbolic Vulnerability Hunting

Semi-automated pipeline converts Trivy, Semgrep, Nmap output into MulVAL attack graphs for agentic pentesting, 53.7% mean vulnerability coverage in CyBench.

The paper presents a semi-automated pipeline that parses Trivy, Semgrep, and Nmap findings into MulVAL predicates and uses an LLM-assisted process to build domain-specific Datalog rules linking scanner evidence to attack techniques. MulVAL/XSB then performs symbolic inference to generate structured, auditable attack paths for agentic pentesting. Evaluated on 54 web CTF tasks from CyBench, every task produced at least one goal-reaching graph with 53.7% mean ground-truth vulnerability coverage, 51.9% full coverage, and an 83.9% noise-path rate. Median end-to-end runtime was 24.9 seconds, making the pipeline runtime-practical for agentic workflows.

arXiv cs.CR·Sep 14, 13:13 UTC · 2d agoResearch

Rethinking Indirect Prompt Injection as a Test-Time Search Problem

Researchers frame indirect prompt injection as test-time search, showing added attacker compute improves vulnerability discovery and exploitation against tool-using agents.

The paper models indirect prompt injection as a test-time search over a task-dependent attack surface shaped by the environment, user task, and injection goal. The authors build an agentic attacker with a dedicated search harness that performs reconnaissance, structured strategy reasoning, and adaptive evaluation using victim-agent feedback. Experiments show more attacker test-time compute improves discovery and exploitation of injection vulnerabilities, with explicit strategy management needed to avoid redundant search. The results argue that agentic security evaluations should characterize attacker search procedures and compute budgets rather than treating attack success as budget-independent.

arXiv cs.CR·Sep 3, 21:28 UTC · 13d agoAI safety & security

Do Input-Level Defenses Transfer to Observation-Level Attacks on VideoLLMs?

A systematic study shows input-level adversarial defenses provide inconsistent, often near-zero protection against observation-level attacks on video LLMs.

Researchers introduce DefTEval, a controlled framework testing eleven input-level defenses against five attack types across five video LLMs. Harmful-content detection rates are frequently near zero, and defenses fail even when attacks embed harmful signals in every sampled frame. Token compression discards localized safety features and modality fusion down-weights weakened visual signals, with defense outcomes dominated by model architecture rather than the defense method.

arXiv cs.CR·Sep 8, 06:59 UTC · 8d agoAI safety & security

Misleading the Planner through Deceptive Resumes: Registration-Time Injection in Centralized Multi-Agent Systems

Researchers demonstrate registration-time prompt injection in centralized LLM multi-agent systems, dropping GAIA task success from 84.31% to 37.25%, and propose DescGuard defense.

The paper identifies a registration-time injection channel in centralized LLM multi-agent systems where third-party worker agent descriptions are trusted by the planner before any user instruction arrives. Analyzing 32,000 descriptions from three public agent marketplaces, at least 23.35% contain content outside the four defined description fields. Eight description-manipulation attack strategies targeting task decomposition, capability grounding, and subtask specification cut GAIA task success from 84.31% to 37.25% and increased token consumption or execution time by over 111%, persisting across two MAS implementations, six planner LLMs, and four evaluators. The proposed DescGuard defense filters descriptions to worker-scoped interface information and restores metrics toward baseline without modifying workers, planner, or orchestration logic.

arXiv cs.CR·Sep 14, 13:06 UTC · 2d agoAI safety & security

Frequently asked questions about the active threat to Siemens S7 Series PLCs

US government agencies warn that unattributed actors using AI-generated exploit scripts are targeting internet-exposed Siemens S7 PLCs in critical infrastructure.

A joint advisory from multiple U.S. government agencies warns that threat actors are exploiting known weaknesses and unnecessary internet exposure of Siemens S7 Series PLCs for reconnaissance and possible pre-positioning for future disruptive attacks. The attackers use AI to generate and refine exploit scripts faster than manual development would allow, lowering the technical bar for ICS attacks. There is no single patch, so defenders must reduce internet exposure and improve monitoring of ICS environments.

Tenable Blog·Aug 20, 14:01 UTC · 27d agoExploit / PoC in the wild

ACEA: An Adversarial Co-Evolution Arena for Head-to-Head Red-Team and Blue-Team LLM Testing

ACEA is a pluggable arena scoring LLM red-team attackers and blue-team defenses head-to-head with an LLM judge and verifiable leakage ground truth.

ACEA connects pluggable red- and blue-team adapters to a shared target LLM through the model-agnostic ASAP HTTP protocol and scores attack and defense rates per adversarial round. Canonical seeded secrets provide verifiable ground truth that separates real leakage from hallucination, and attacks are delivered to the target even when blocked to measure raw potency. The platform adds real-time battle visualization, failure-localizing end-of-battle reports, and an optional in-context improvement loop that feeds advisory hints between rounds.

arXiv cs.CR·Sep 8, 04:57 UTC · 8d agoAI safety & security1

Chinese espionage groups swarm to exploit triple-link chain of zero-days

Four China-aligned espionage groups, starting with TA412/APT31, chained zero-days CVE-2026-85046, CVE-2026-87491 and CVE-2026-85880 to spy on targets since late August.

TA412 (Violet Typhoon/APT31) began exploiting the BlueMoon chain on Aug. 28, followed by UNK_LateNight (US aerospace, Sept. 2), UNK_DoubleCheck (Vietnamese manufacturing via a compromised Southeast Asian government account) and UNK_QuietRacket (Indonesia and Singapore government, consulting and finance, Sept. 3). The chain pairs Chromium JavaScript engine RCE flaws CVE-2026-85046 and CVE-2026-87491 with the Windows ALPC privilege-escalation zero-day CVE-2026-85880, enabling sandbox escape and system privileges. Delivery used phishing links installing a fake Google Gemini browser extension that surveilled browser activity and stole credentials; Proofpoint directly observed fewer than 20 organizations but expects wider proliferation as the kit spreads.

Proofpoint Threat Insight·Sep 9, 21:17 UTC · 7d agoThreat actor in the wild 3 sourcesCVE-2026-85046CVE-2026-87491CVE-2026-858804

IoT Under Siege: The Anatomy of the Latest Mirai Campaign Leveraging Multiple IoT Exploits

Unit 42 tracks a Mirai botnet campaign exploiting over 20 IoT vulnerabilities in routers, cameras and DVRs to build DDoS botnets since March 2023.

Since March 2023, Unit 42 has tracked threat actors exploiting more than 20 IoT vulnerabilities to spread a Mirai botnet variant, first seen downloading payloads from zvub.us on March 14, 2023. Exploited flaws span CVE-2023-1389 (TP-Link Archer), CVE-2022-30525 (Zyxel), CVE-2022-31499 (Nortek) and many router, camera and DVR bugs. The variant decrypts configuration strings with an XOR key derived from 0xDEADBEEF and lacks built-in credential brute forcing, so spreading relies on manual operator exploitation. Two campaigns observed since October 2022 share infrastructure and near-identical samples.

Palo Alto Unit 42·Aug 17, 13:04 UTC · Aug 17, 2026Malware in the wildCVE-2019-12725CVE-2019-17621CVE-2019-20500+13 CVEs1