Tag
#vulnerability
### Description A critical vulnerability exists in the `llamafy_baichuan2.py` script of the [LLaMA-Factory](https://github.com/hiyouga/LLaMA-Factory) project. The script performs insecure deserialization using `torch.load()` on user-supplied `.bin` files from an input directory. An attacker can exploit this behavior by crafting a malicious `.bin` file that executes arbitrary commands during deserialization. ### Attack Vector This vulnerability is **exploitable without authentication or privileges** when a user is tricked into: 1. Downloading or cloning a malicious project folder containing a crafted `.bin` file (e.g. via zip file, GitHub repo). 2. Running the provided conversion script `llamafy_baichuan2.py`, either manually or as part of an example workflow. No elevated privileges are required. The user only needs to run the script with an attacker-supplied `--input_dir`. ### Impact - Arbitrary command execution (RCE) - System compromise - Persistence or lateral movement in sh...
PostHog ClickHouse Table Functions SQL Injection Remote Code Execution Vulnerability. This vulnerability allows network-adjacent attackers to execute arbitrary code on affected installations of PostHog. Authentication is required to exploit this vulnerability. The specific flaw exists within the implementation of the SQL parser. The issue results from the lack of proper validation of a user-supplied string before using it to construct SQL queries. An attacker can leverage this vulnerability to execute code in the context of the database account. Was ZDI-CAN-25350.
### Description Before version 1.7.0, utls did not implement the TLS 1.3 downgrade protection mechanism specified in RFC 8446 Section 4.1.3 when using a utls ClientHello spec. This allowed an active network adversary to downgrade TLS 1.3 connections initiated by a utls client to a lower TLS version (e.g., TLS 1.2) by modifying the ClientHello message to exclude the SupportedVersions extension, causing the server to respond with a TLS 1.2 ServerHello (along with a downgrade canary in the ServerHello random field). Because utls did not check the downgrade canary in the ServerHello random field, clients would accept the downgraded connection without detecting the attack. This attack could also be used by an active network attacker to fingerprint utls connections. ### Fix Commit or Pull Request refraction-networking/utls#337, specifically refraction-networking/utls@f8892761e2a4d29054264651d3a86fda83bc83f9 ### References - https://github.com/refraction-networking/utls/issues/181
### Impact It is possible for a remote unauthenticated user to escape from the HQL execution context and perform a blind SQL injection to execute arbitrary SQL statements on the database backend, including when "Prevent unregistered users from viewing pages, regardless of the page rights" and "Prevent unregistered users from editing pages, regardless of the page rights" options are enabled. Depending on the used database backend, the attacker may be able to not only obtain confidential information such as password hashes from the database, but also execute UPDATE/INSERT/DELETE queries. The vulnerability may be tested in a default installation of XWIki Standard Flavor, including using the official Docker containers. An example query, which leads to SQL injection with MySQL/MariaDB backend is shown below: ``` time curl "http://127.0.0.1:8080/rest/wikis/xwiki/query?q=where%20doc.name=length('a')*org.apache.logging.log4j.util.Chars.SPACE%20or%201%3C%3E%271%5C%27%27%20union%20select%20...
### Impact It is possible for a user with SCRIPT right to escape from the HQL execution context and perform a blind SQL injection to execute arbitrary SQL statements on the database backend. Depending on the used database backend, the attacker may be able to not only obtain confidential information such as password hashes from the database, but also execute UPDATE/INSERT/DELETE queries. The vulnerability may be tested in a default installation of XWIki Standard Flavor, including using the official Docker containers. For example, with a MySQL or MariaDB database, you can use the following script (which a user having SCRIPT right but not PROGRAMMING right) to get the content of the xwikistrings table (which contain all the short string fields stored in objects, including passwords): ``` {{velocity}} $services.query.hql("where 1<>'1\'' union select concat(XWS_NAME, XWS_VALUE) from xwikistrings #'").execute() {{/velocity}} ``` ### Patches This has been patched in 16.10.1, 16.4.6 and...
The path shortening function is used in pnpm： ``` export function depPathToFilename (depPath: string, maxLengthWithoutHash: number): string { let filename = depPathToFilenameUnescaped(depPath).replace(/[\\/:*?"<>|]/g, '+') if (filename.includes('(')) { filename = filename .replace(/\)$/, '') .replace(/(\)\()|\(|\)/g, '_') } if (filename.length > maxLengthWithoutHash || filename !== filename.toLowerCase() && !filename.startsWith('file+')) { return `${filename.substring(0, maxLengthWithoutHash - 27)}_${createBase32Hash(filename)}` } return filename } ``` However, it 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 /node_modoules/, there are no version numbers for the libraries they refer to.  In t...
AOA, DaVita, and Bell Ambulance hit by ransomware in 2025. Over 245K affected as hackers steal patient data,…
Cisco Talos discovered a sophisticated attack on critical infrastructure by ToyMaker and Cactus, using the LAGTOY backdoor to orchestrate a relentless double extortion scheme.
## Description https://github.com/vllm-project/vllm/security/advisories/GHSA-rh4j-5rhw-hr54 reported a vulnerability where loading a malicious model could result in code execution on the vllm host. The fix applied to specify `weights_only=True` to calls to `torch.load()` did not solve the problem prior to PyTorch 2.6.0. PyTorch has issued a new CVE about this problem: https://github.com/advisories/GHSA-53q9-r3pm-6pq6 This means that versions of vLLM using PyTorch before 2.6.0 are vulnerable to this problem. ## Background Knowledge When users install VLLM according to the official manual  But the version of PyTorch is specified in the requirements. txt file  So by default when the user install VLLM, it will install the PyTorch with version 2.5.1  in the tags feature. Any user with the ability of create or modify tags can inject malicious JavaScript code in the name field.