Lucene search
+L

10 matches found

Veracode
Veracode
•added 2026/10/29 4:06 a.m.•22 views

VLLM: Phi-4 Multimodal Audio Encoder Materializes And Retains Full Attention Matrices: A Single Max-length Audio OOM-cra

Summary The Phi-4 multimodal audio encoder "phi4mmaudio.py" / "phi4mmutils.py" runs its conformer attention in eager mode by default: the official HF config of "microsoft/Phi-4-multimodal-instruct" does not set "useptscaleddotproductattention", so the default "False" applies "phi4mmaudio.py:166",...

6.5CVSS6.2AI score0.00329EPSS
SaveExploits1References4Affected Software1
Packet Storm News
Packet Storm News
•added 2026/10/29 12:00 a.m.•20 views

Systematic Evaluation of Adversarial Jailbreaking Attacks and Defenses across Seven Large Language Models

Large language models LLMs deployed in safety-critical settings must resist adversarial inputs designed to elicit harmful outputs. We present a systematic empirical evaluation of five adversarial jailbreaking attack families---direct requests, roleplay-based jailbreaking, hypothetical framing,...

5.9AI score
SaveExploits0
osv_ghsa
osv_ghsa
•added 2026/10/28 9:15 p.m.•17 views

GHSA-8397-8QF6-F2GM vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine

Summary The Phi-4 multimodal audio encoder phi4mmaudio.py / phi4mmutils.py runs its conformer attention in eager mode by default: the official HF config of microsoft/Phi-4-multimodal-instruct does not set useptscaleddotproductattention, so the default False applies phi4mmaudio.py:166, :876;...

6.5CVSS6.2AI score0.00329EPSS
SaveExploits1References6
Github Security Blog
Github Security Blog
•added 2026/10/28 9:15 p.m.•31 views

vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine

Summary The Phi-4 multimodal audio encoder phi4mmaudio.py / phi4mmutils.py runs its conformer attention in eager mode by default: the official HF config of microsoft/Phi-4-multimodal-instruct does not set useptscaleddotproductattention, so the default False applies phi4mmaudio.py:166, :876;...

6.5CVSS6.2AI score0.00329EPSS
SaveExploits1References6Affected Software1
osv_cve
osv_cve
•added 2026/10/26 4:56 p.m.•25 views

CVE-2026-115913 vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine

vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mmaudio.py and phi4mmutils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forwardattenti...

6.5CVSS6.2AI score0.00329EPSS
SaveExploits1References6
CVE
CVE
•added 2026/10/26 4:56 p.m.•60 views

CVE-2026-115913

Prior to version 0.32.0 , the vLLM inference and serving engine is vulnerable to a Denial of Service (DoS) condition via its Phi-4 multimodal audio encoder. Specifically, the components phi4mm_audio.py and phi4mm_utils.py use eager attention by default, causing MultiHeadedAttention.forward_attent...

6.5CVSS6.2AI score0.00329EPSS
SaveExploits1References4Affected Software1
Vulnrichment
Vulnrichment
•added 2026/10/26 4:56 p.m.•11 views

CVE-2026-115913 vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine

vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mmaudio.py and phi4mmutils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forwardattenti...

6.5CVSS6.2AI score0.00329EPSS
SaveExploits1References4
Cvelist
Cvelist
•added 2026/10/26 4:56 p.m.•37 views

CVE-2026-115913 vLLM: Phi-4 multimodal audio encoder materializes and retains full attention matrices: a single max-length audio OOM-crashes the vLLM engine

vLLM is an inference and serving engine for large language models. Prior to 0.32.0, the Phi-4 multimodal audio encoder in phi4mmaudio.py and phi4mmutils.py uses eager attention by default, allowing a remote client to submit one legal long audio clip that causes MultiHeadedAttention.forwardattenti...

6.5CVSS0.00329EPSS
SaveExploits1References4
Positive Technologies
Positive Technologies
•added 2026/10/26 12:00 a.m.•14 views

PT-2026-123436

Name of the Vulnerable Software and Affected Versions vLLM versions prior to 0.32.0 Description The Phi-4 multimodal audio encoder in phi4mm audio.py and phi4mm utils.py uses eager attention by default. This allows a remote client to submit a long audio clip that triggers the...

6.5CVSS6.1AI score0.00329EPSS
SaveExploits1References9
Schneier on Security
Schneier on Security
•added 2026/09/23 11:03 a.m.•50 views

Research on Models Engaging in Genie-Like Behavior

New paper: "Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training." Abstract: We discover a novel and surprising phenomenon of unintentional misalignment in reasoning language models RLMs, which we call self-jailbreaking. Specifically,...

5.9AI score
SaveExploits0
Rows per page
Query Builder