Design Patterns for Securing LLM Agents against Prompt Injections

As AI agents powered by Large Language Models LLMs become increasingly versatile and capable of addressing a broad spectrum of tasks, ensuring their security has become a critical challenge. Among the most pressing threats are prompt injection attacks, which exploit the agent's resilience on...

CVE-2024-37405

Livechat messages can be leaked by combining two NoSQL injections affecting livechat:loginByToken pre-authentication and livechat:loadHistory...

CVE-2023-1097

Baicells EG7035-M11 devices with firmware through BCE-ODU-1.0.8 are vulnerable to improper code exploitation via HTTP GET command injections. Commands are executed using pre-login execution and executed with root permissions. The following methods have been tested and validated by a 3rd party...

CVE-2023-34420

A valid, authenticated LXCA user with elevated privileges may be able to execute command injections through crafted calls to a specific web API...

CVE-2023-24508

Baicells Nova 227, Nova 233, and Nova 243 LTE TDD eNodeB and Nova 246 devices with firmware through RTS/RTD 3.6.6 are vulnerable to remote shell code exploitation via HTTP command injections. Commands are executed using pre-login execution and executed with root permissions. The following methods...

CVE-2021-24889

The Ninja Forms Contact Form WordPress plugin before 3.6.4 does not escape keys of the fields POST parameter, which could allow high privilege users to perform SQL injections attacks...

CVE-2020-9021

Post Oak AWAM Bluetooth Field Device 7400v2.08.21.2018, 7800SD.2015.1.16, 2011.3, 7400v2.02.01.2019, and 7800SD.2012.12.5 is vulnerable to injections of operating system commands through timeconfig.py via shell metacharacters in the htmlNtpServer parameter...

CVE-2020-14075

TRENDnet TEW-827DRU devices through 2.06B04 contain multiple command injections in apply.cgi via the action pppoeconnect, rupppoeconnect, or dhcpconnect with the key wanifname or wan0dns, allowing an authenticated user to run arbitrary commands on the device...

CVE-2019-7256

Linear eMerge E3-Series devices allow Command Injections...

CVE-2014-10386

The wp-live-chat-support plugin before 4.1.0 for WordPress has JavaScript injections...

CVE-2015-5355

Multiple cross-site scripting XSS vulnerabilities in GetSimple CMS before 3.3.6 allow remote attackers to inject arbitrary web script or HTML via the 1 post-content or 2 post-title parameter to admin/edit.php...

Researchers Expose PWA JavaScript Attack That Redirects Users to Adult Scam Apps

Cybersecurity researchers have discovered a new campaign that employs malicious JavaScript injections to redirect site visitors on mobile devices to a Chinese adult-content Progressive Web App PWA scam. "While the payload itself is nothing new yet another adult gambling scam, the delivery method...

Can Large Language Models Really Recognize Your Name?

Large language models LLMs are increasingly being used to protect sensitive user data. However, current LLM-based privacy solutions assume that these models can reliably detect personally identifiable information PII, particularly named entities. In this paper, we challenge that assumption by...

Lessons from Defending Gemini against Indirect Prompt Injections

Gemini is increasingly used to perform tasks on behalf of users, where function-calling and tool-use capabilities enable the model to access user data. Some tools, however, require access to untrusted data introducing risk. Adversaries can embed malicious instructions in untrusted data which caus...

Improving LLM Outputs against Jailbreak Attacks with Expert Model Integration

Using LLMs in a production environment presents security challenges that include vulnerabilities to jailbreaks and prompt injections, which can result in harmful outputs for humans or the enterprise. The challenge is amplified when working within a specific domain, as topics generally accepted fo...

CVE-2025-0668 BOINC Server Multiple SQL Injections

Improper Neutralization of Input During Web Page Generation XSS or 'Cross-site Scripting' vulnerability in BOINC Server allows Stored XSS.This issue affects BOINC Server: before 1.4.5...

The Obvious Invisible Threat: LLM-Powered GUI Agents' Vulnerability to Fine-Print Injections

A Large Language Model LLM powered GUI agent is a specialized autonomous system that performs tasks on the user's behalf according to high-level instructions. It does so by perceiving and interpreting the graphical user interfaces GUIs of relevant apps, often visually, inferring necessary sequenc...

CVE-2025-0723

The ProfileGrid – User Profiles, Groups and Communities plugin for WordPress is vulnerable to blind and time-based SQL Injections via the rid and search parameters in all versions up to, and including, 5.9.4.7 due to insufficient escaping on the user supplied parameter and lack of sufficient...

SUSE SLED15 / SLES15 / openSUSE 15 Security Update : emacs (SUSE-SU-2025:0599-1)

The remote SUSE Linux SLED15 / SLEDSAP15 / SLES15 / SLESSAP15 / openSUSE 15 host has packages installed that are affected by a vulnerability as referenced in the SUSE-SU-2025:0599-1 advisory. - CVE-2025-1244: improper handling of custom 'man' URI schemes allow for shell command injections...

CVE-2024-13900

The Head, Footer and Post Injections plugin for WordPress is vulnerable to PHP Code Injection in all versions up to, and including, 3.3.0. This makes it possible for authenticated attackers, with Administrator-level access and above, to inject PHP Code in multisite environments...