Security Vulnerabilities
- CVEs Published In June 2025
llama.cpp is an inference of several LLM models in C/C++. Prior to version b5721, there is a signed vs. unsigned integer overflow in llama.cpp's tokenizer implementation (llama_vocab::tokenize) (src/llama-vocab.cpp:3036) resulting in unintended behavior in tokens copying size comparison. Allowing heap-overflowing llama.cpp inferencing engine with carefully manipulated text input during tokenization process. This issue has been patched in version b5721.
Gogs is an open source self-hosted Git service. Prior to version 0.13.3, it's still possible to delete files under the .git directory and achieve remote command execution due to an insufficient patch for CVE-2024-39931. Unprivileged user accounts can execute arbitrary commands on the Gogs instance with the privileges of the account specified by RUN_USER in the configuration. Allowing attackers to access and alter any users' code hosted on the same instance. This issue has been patched in version 0.13.3.
Successful exploitation of the vulnerability could allow an unauthenticated, remote attacker to send Modbus TCP packets to manipulate Digital Outputs, potentially allowing remote control of relay channel which may lead to operational or safety risks.
Successful exploitation of the vulnerability could allow an attacker to cause repeated reboots, potentially leading to remote denial-of-service and system unavailability.
Successful exploitation of the vulnerability could allow an attacker that has physical access to interface with JTAG to inject or modify firmware.
Successful exploitation of the vulnerability could allow an unauthenticated attacker to upload firmware through a public update page, potentially leading to backdoor installation or privilege escalation.
Successful exploitation of the stored cross-site scripting vulnerability could allow an attacker to inject malicious scripts into device fields and executed in other users’ browser, potentially leading to session hijacking, defacement, credential theft, or privilege escalation.
Kanboard is project management software that focuses on the Kanban methodology. Prior to version 1.2.46, Kanboard allows password reset emails to be sent with URLs derived from the unvalidated Host header when the application_url configuration is unset (default behavior). This allows an attacker to craft a malicious password reset link that leaks the token to an attacker-controlled domain. If a victim (including an administrator) clicks the poisoned link, their account can be taken over. This affects all users who initiate a password reset while application_url is not set. This issue has been patched in version 1.2.46.
Successful exploitation of the vulnerability could allow an unauthenticated attacker to conduct brute force guessing and account takeover as the session cookies are predictable, potentially allowing the attackers to gain root, admin or user access and reset passwords.
Successful exploitation of the vulnerability could allow an attacker to consume all available session slots and block other users from logging in, thereby preventing legitimate users from gaining access to the product.