Deserialization of Untrusted Data

Overview chattts is an A generative speech model for daily dialogue Affected versions of this package are vulnerable to Deserialization of Untrusted Data when loading the tokenizer with torch.load. An attacker can exploit this by supplying a malicious tokenizer.pt file, leading to code execution...

PT-2024-40623 · Git +1 · Llama.Cpp

Name of the Vulnerable Software and Affected Versions: No specific software or versions are mentioned in the provided description. Description: The issue is related to a crash type of UNKNOWN READ. The crash state includes functions such as llama output reserve and llama decode, and is located in...

PT-2024-40620 · Git +1 · Llama.Cpp

Name of the Vulnerable Software and Affected Versions: No specific software or versions are mentioned in the provided description. Description: A crash occurred due to a segmentation fault on an unknown address. The crash state includes functions such as llama output reserve and llama decode, whi...

Malicious code in css-tokenizer (npm)

--- -= Per source details. Do not edit below this line.=-...

MAL-2024-9577 Malicious code in css-tokenizer (npm)

--- -= Per source details. Do not edit below this line.=-...

RHEL 8 : grafana (Unpatched Vulnerability)

The remote Redhat Enterprise Linux 8 host has one or more packages installed that are affected by multiple vulnerabilities that have been acknowledged by the vendor but will not be patched. - nodejs-underscore: Arbitrary code execution via the template function CVE-2021-23358 - node-fetch is...

UBUNTU-CVE-2024-39705

NLTK through 3.8.1 allows remote code execution if untrusted packages have pickled Python code, and the integrated data package download functionality is used. This affects, for example, averagedperceptrontagger and punkt...

Server-side Template Injection (SSTI)

litellm is vulnerable to Server-Side Template Injection SSTI. The vulnerability is due to insufficient sanitization within the hfchattemplate method in factory.py, which processes the chattemplate parameter from the tokenizerconfig.json file using the Jinja template engine, allowing attackers to...

GHSA-46CM-PFWV-CGF8 LiteLLM has Server-Side Template Injection vulnerability in /completions endpoint

BerriAI/litellm is vulnerable to Server-Side Template Injection SSTI via the /completions endpoint. The vulnerability arises from the hfchattemplate method processing the chattemplate parameter from the tokenizerconfig.json file through the Jinja template engine without proper sanitization...

PT-2024-22934

Name of the Vulnerable Software and Affected Versions BerriAI/litellm affected versions not specified Description The issue arises from the hf chat template method processing the chat template parameter from the tokenizer config.json file through the Jinja template engine without proper...

PT-2023-32836 · Hugging Face · Transformers

Name of the Vulnerable Software and Affected Versions: huggingface/transformers versions prior to 4.36 Description: The issue concerns the deserialization of untrusted data, which can compromise model integrity and potentially lead to remote code execution RCE. This is particularly relevant when...

GO-2023-1914 Brute-force of token secrets in github.com/superfly/tokenizer

Brute-force of token secrets in github.com/superfly/tokenizer...

Brute Force Token Secrets

superfly/tokenizer is vulnerable to brute-force of token secrets vulnerability. The vulnerability is due to not restricting formatting in fmt parameter to simple formatting and allowing fmt/dst parameters to be specified at request time leading to attacker brute forcing secret values using...

Tokenizer vulnerable to client brute-force of token secrets

Impact Authorized clients, having an injectprocessor secret, could brute-force the secret token value by abusing the fmt parameter to the Proxy-Tokenizer header. Patches This was fixed in https://github.com/superfly/tokenizer/pull/8 and further mitigated in...

GHSA-F28G-86HC-823Q Tokenizer vulnerable to client brute-force of token secrets

Impact Authorized clients, having an injectprocessor secret, could brute-force the secret token value by abusing the fmt parameter to the Proxy-Tokenizer header. Patches This was fixed in https://github.com/superfly/tokenizer/pull/8 and further mitigated in...

Security Bulletin: Open Source Dependency Vulnerability

Summary IBM Edge Application Manager 4.5 has resolved the vulnerability. Vulnerability Details CVEID:CVE-2022-25758 DESCRIPTION: Node.js scss-tokenizer module is vulnerable to a denial of service, caused by a regular expression denial of service ReDoS vulnerability in the loadAnnotation function....

SUSE CVE-2023-30402

YASM v1.3.0 was discovered to contain a heap overflow via the function handledotlabel at /nasm/nasm-token.re. Note: This has been disputed by third parties who argue this is a bug and not a security issue because yasm is a standalone program not designed to run untrusted code...

SUSE CVE-2009-4214

Cross-site scripting XSS vulnerability in the striptags function in Ruby on Rails before 2.2.s, and 2.3.x before 2.3.5, allows remote attackers to inject arbitrary web script or HTML via vectors involving non-printing ASCII characters, related to HTML::Tokenizer and...

SUSE CVE-2017-11626

A stack-consumption vulnerability was found in libqpdf in QPDF 6.0.0, which allows attackers to cause a denial of service via a crafted file, related to the QPDFTokenizer::resolveLiteral function in QPDFTokenizer.cc after four consecutive calls to QPDFObjectHandle::parseInternal, aka an "infinite...

SUSE CVE-2017-12595

The tokenizer in QPDF 6.0.0 and 7.0.b1 is recursive for arrays and dictionaries, which allows remote attackers to cause a denial of service stack consumption and segmentation fault or possibly have unspecified other impact via a PDF document with a deep data structure, as demonstrated by a crash ...