vLLM 输入验证错误漏洞

vLLM is an open-source inference and service engine designed for LLM models, featuring high throughput and efficient memory usage. Versions of vLLM prior to 0.6.1 to 0.20.0 contained a vulnerability related to input validation errors. This vulnerability stemmed from token injection issues during...

CVE-2026-42339

New API is a large language mode LLM gateway and artificial intelligence AI asset management system. In versions 0.11.9-alpha.1 and prior, the SSRF protection introduced in v0.9.0.5 CVE-2025-59146 and hardened in v0.9.6 CVE-2025-62155 does not block the unspecified address 0.0.0.0. A regular...

CVE-2026-42339 New API: SSRF Filter Bypass via 0.0.0.0

New API is a large language mode LLM gateway and artificial intelligence AI asset management system. In versions 0.11.9-alpha.1 and prior, the SSRF protection introduced in v0.9.0.5 CVE-2025-59146 and hardened in v0.9.6 CVE-2025-62155 does not block the unspecified address 0.0.0.0. A regular...

Hard to Read, Easy to Jailbreak: How Visual Degradation Bypasses MLLM Safety Alignment

Recent advancements in visual context compression enable MLLMs to process ultra-long contexts efficiently by rendering text into images. However, we identify a critical vulnerability inherent to this paradigm: lowering image resolution inadvertently catalyzes jailbreaking. Our experiments reveal...

GHSA-V5C3-6WVC-PC2Q QuantumNous/new-api has an SSRF Filter Bypass via 0.0.0.0

SSRF Filter Bypass via 0.0.0.0 Summary The SSRF protection introduced in v0.9.0.5 CVE-2025-59146 and hardened in v0.9.6 CVE-2025-62155 does not block the unspecified address 0.0.0.0. A regular non-admin user holding any valid API token can send a multimodal request to /v1/chat/completions,...

GHSA-HPV8-X276-M59F vLLM Vulnerable to Remote DoS via Special-Token Placeholders

Summary This report explains a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder sequences supplied without matching data cause vLLM to index into empty grids during...

vLLM Vulnerable to Remote DoS via Special-Token Placeholders

Summary This report explains a Token Injection vulnerability in vLLM’s multimodal processing. Unauthenticated, text-only prompts that spell special tokens are interpreted as control. Image and video placeholder sequences supplied without matching data cause vLLM to index into empty grids during...

Redefining AI Red Teaming in the Agentic Era: From Weeks to Hours

AI systems are entering critical domains like healthcare, finance, and defense, yet remain vulnerable to adversarial attacks. While AI red teaming is a primary defense, current approaches force operators into manual, library-specific workflows. Operators spend weeks hand-crafting workflows -...

PT-2026-37318

Name of the Vulnerable Software and Affected Versions vLLM versions 0.6.1 through 0.19.x Description A Token Injection issue exists in the multimodal processing of vLLM. Unauthenticated, text-only prompts containing special tokens are interpreted as control commands. When image and video...

RoboKA: KAN Informed Multimodal Learning for RoboCall Surveillance System

Wide exploration on robocall surveillance research is hindered due to limited access to public datasets, due to privacy concerns. In this work, we first curate Robo-SAr, a synthetic robocall dataset designed for robocall surveillance research. Robo-SAr comprises of 200 unwanted and 1200 legitimat...

Jailbreaking Frontier Foundation Models through Intention Deception

Large vision-language models exhibit remarkable capability but remain highly susceptible to jailbreaking. Existing safety training approaches aim to have the model learn a refusal boundary between safe and unsafe, based on the user's intent. It has been found that this binary training regime ofte...

Text Steganography with Dynamic Codebook and Multimodal Large Language Model

With the popularity of the large language models LLMs, text steganography has achieved remarkable performance. However, existing methods still have some issues: 1 For the white-box paradigm, this steganography behavior is prone to exposure due to sharing the off-the-shelf language model between...

CVE-2026-6604

A vulnerability was identified in modelscope agentscope up to 1.0.18. Affected by this issue is the function parseurl/prepareimage/openaiaudiototext of the file src/agentscope/tool/multimodality/openaitools.py of the component Cloud Metadata Endpoint. Such manipulation of the argument...

Server-side Request Forgery (SSRF)

Overview ragas is an Evaluation framework for RAG and LLM applications Affected versions of this package are vulnerable to Server-side Request Forgery SSRF via improper validation of URLs in the retrievedcontexts parameter when processing multimodal inputs. An attacker can access arbitrary files,...

From Incomplete Architecture to Quantified Risk: Multimodal LLM-Driven Security Assessment for Cyber-Physical Systems

Cyber-physical systems often contend with incomplete architectural documentation or outdated information resulting from legacy technologies, knowledge management gaps, and the complexity of integrating diverse subsystems over extended operational lifecycles. This architectural incompleteness...

Stealthy and Adjustable Text-Guided Backdoor Attacks on Multimodal Pretrained Models

Multimodal pretrained models are vulnerable to backdoor attacks, yet most existing methods rely on visual or multimodal triggers, which are impractical since visually embedded triggers rarely occur in real-world data. To overcome this limitation, we propose a novel Text-Guided Backdoor TGB attack...

ContractShield: Bridging Semantic-Structural Gaps Via Hierarchical Cross-Modal Fusion for Multi-Label Vulnerability Detection in Obfuscated Smart Contracts

Smart contracts are increasingly targeted by adversaries employing obfuscation techniques such as bogus code injection and control flow manipulation to evade vulnerability detection. Existing multimodal methods often process semantic, temporal, and structural features in isolation and fuse them...

CVE-2026-22742

Spring AI's spring-ai-bedrock-converse contains a Server-Side Request Forgery SSRF vulnerability in BedrockProxyChatModel when processing multimodal messages that include user-supplied media URLs. Insufficient validation of those URLs allows an attacker to induce the server to issue HTTP requests...

GHSA-MHRG-94VW-45C5 Spring AI: Insufficient Validation causes SSRF when processing multimodal messages with user-supplied URLs

Spring AI's spring-ai-bedrock-converse contains a Server-Side Request Forgery SSRF vulnerability in BedrockProxyChatModel when processing multimodal messages that include user-supplied media URLs. Insufficient validation of those URLs allows an attacker to induce the server to issue HTTP requests...

EUVD-2026-16537

Spring AI's spring-ai-bedrock-converse contains a Server-Side Request Forgery SSRF vulnerability in BedrockProxyChatModel when processing multimodal messages that include user-supplied media URLs. Insufficient validation of those URLs allows an attacker to induce the server to issue HTTP requests...