BadReward: Clean-Label Poisoning of Reward Models in Text-To-Image RLHF

Reinforcement Learning from Human Feedback RLHF is crucial for aligning text-to-image T2I models with human preferences. However, RLHF's feedback mechanism also opens new pathways for adversaries. This paper demonstrates the feasibility of hijacking T2I models by poisoning a small fraction of...

PYSEC-2025-43

vLLM is an inference and serving engine for large language models LLMs. In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image...

PYSEC-2025-43

vLLM is an inference and serving engine for large language models LLMs. In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image...

CVE-2025-46722

The CVE-2025-46722 entry concerns vLLM (versions 0.7.0–0.8.x) where MultiModalHasher in vllm/multimodal/hasher.py hashes PIL.Image.Image objects using only obj.tobytes(). This excludes image metadata (width, height, mode), enabling two images with identical pixel data but different shapes to yiel...

CVE-2025-46722 vLLM has a Weakness in MultiModalHasher Image Hashing Implementation

vLLM is an inference and serving engine for large language models LLMs. In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image...

CVE-2025-46722 vLLM has a Weakness in MultiModalHasher Image Hashing Implementation

vLLM is an inference and serving engine for large language models LLMs. In versions starting from 0.7.0 to before 0.9.0, in the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image...

vLLM has a Weakness in MultiModalHasher Image Hashing Implementation

Summary In the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image objects using only obj.tobytes, which returns only the raw pixel data, without including metadata such as the...

Incomplete Comparison with Missing Factors

Overview vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs Affected versions of this package are vulnerable to Incomplete Comparison with Missing Factors due to the implementation of image hashing in hasher.py. An attacker can achieve hash collisions and...

GHSA-C65P-X677-FGJ6 vLLM has a Weakness in MultiModalHasher Image Hashing Implementation

Summary In the file vllm/multimodal/hasher.py, the MultiModalHasher class has a security and data integrity issue in its image hashing method. Currently, it serializes PIL.Image.Image objects using only obj.tobytes, which returns only the raw pixel data, without including metadata such as the...

PT-2025-23136 · Vllm · Vllm

Name of the Vulnerable Software and Affected Versions: vLLM versions 0.7.0 through 0.8.x Description: The issue concerns a security and data integrity problem in the image hashing method of the MultiModalHasher class. Specifically, the method serializes PIL.Image.Image objects using only...

CVE-2024-48924

Impact When this library is used to deserialize messagepack data from an untrusted source, there is a risk of a denial of service attack by an attacker that sends data contrived to produce hash collisions, leading to large CPU consumption disproportionate to the size of the data being deserialize...

CVE-2023-31135

Dgraph is an open source distributed GraphQL database. Existing Dgraph audit logs are vulnerable to brute force attacks due to nonce collisions. The first 12 bytes come from a baseIv which is initialized when an audit log is created. The last 4 bytes come from the length of the log line being...

CVE-2022-28229

The hash functionality in userver before 42059b6319661583b3080cab9b595d4f8ac48128 allows attackers to cause a denial of service via crafted HTTP request, involving collisions...

CVE-2020-36563

XML Digital Signatures generated and validated using this package use SHA-1, which may allow an attacker to craft inputs which cause hash collisions depending on their control over the input...

CVE-2012-1035

AdaCore Ada Web Services AWS before 2.10.2 computes hash values for form parameters without restricting the ability to trigger hash collisions predictably, which allows remote attackers to cause a denial of service CPU consumption by sending many crafted parameters...

kernel: sched: sch_cake: fix bulk flow accounting logic for host fairness

In the Linux kernel, the following vulnerability has been resolved: sched: schcake: fix bulk flow accounting logic for host fairness In schcake, we keep track of the count of active bulk flows per host, when running in dst/src host fairness mode, which is used as the round-robin weight when...

Use Of Weak Hash

pnpm is vulnerable to Use of Weak Hash. The vulnerability is due to improper path shortening caused by the use of the md5 function for compression, which can cause different libraries to resolve to the same storage path if a hash collision occurs...

CVE-2024-47829

pnpm is a package manager. Prior to version 10.0.0, the path shortening function uses the md5 function as a path shortening compression function, and if a collision occurs, it will result in the same storage path for two different libraries. Although the real names are under the package name...

GHSA-8CC4-RFJ6-FHG4 pnpm uses the md5 path shortening function causes packet paths to coincide, which causes indirect packet overwriting

The path shortening function is used in pnpm： export function depPathToFilename depPath: string, maxLengthWithoutHash: number: string let filename = depPathToFilenameUnescapeddepPath.replace/\/:?"|/g, '+' if filename.includes'' filename = filename .replace/$/, '' .replace/\||/g, '' if...

CVE-2025-3576 Krb5: kerberos rc4-hmac-md5 checksum vulnerability enabling message spoofing via md5 collisions

A vulnerability in the MIT Kerberos implementation allows GSSAPI-protected messages using RC4-HMAC-MD5 to be spoofed due to weaknesses in the MD5 checksum design. If RC4 is preferred over stronger encryption types, an attacker could exploit MD5 collisions to forge message integrity codes. This ma...