MCP tools are implicated in several new attack techniques. Here's a look at how they can be manipulated for good, such as logging tool usage and filtering unauthorized commands.
Over the last few months, there has been a lot of activity in the Model Context Protocol (MCP) space, both in terms of adoption as well as security. Developed by Anthropic, MCP has been rapidly gaining traction across the AI ecosystem. MCP allows Large Language Models (LLMs) to interface with tools and for those interfaces to be rapidly created. MCP tools allow for the rapid development of “agentic” systems, or AI systems that autonomously perform tasks.
Beyond adoption, new attack techniques have been shown to allow prompt injection via MCP tool descriptions and responses, MCP tool poisoning, rug pulls and more.
Prompt Injection is a weakness in LLMs that can be used to elicit unintended behavior, circumvent safeguards and produce potentially malicious responses. Prompt injection occurs when an attacker instructs the LLM to disregard other rules and do the attacker’s bidding. In this blog, I show how to use techniques similar to prompt injection to change the LLM’s interaction with MCP tools. Anyone conducting MCP research may find these techniques useful.
A new attack technique named Policy Puppetry can break the protections of major gen-AI models to produce harmful outputs.
Silent and undetectable initial access is the cornerstone of a cyberattack. MFA is there to stop unauthorized access, but attackers are constantly evolving.
Discover ClickFix, a rising social engineering threat used to deliver malware and learn how to detect and respond against it with Logpoint.
A new variation of clickjacking attacks called "DoubleClickjacking" lets attackers trick users into authorizing sensitive actions using double-clicks while bypassing existing protections against these types of attacks.
Clickjacking, also known as UI redressing, is when threat actors create malicious web pages that trick visitors into clicking on hidden or disguised webpage elements.
“Clickjacking” attacks have been around for over a decade, enabling malicious websites to trick users into clicking hidden or disguised buttons they never intended to click . This technique is becoming less practical as modern browsers set all cookies to “SameSite: Lax” by default. Even if an attacker site can frame another website, the framed site would be unauthenticated, because cross-site cookies are not sent. This significantly reduces the risk of successful clickjacking attacks, as most interesting functionality on websites typically requires authentication.
Akamai security researcher Tomer Peled explored new ways to use and abuse Microsoft's UI Automation framework and discovered an attack technique that evades endpoint detection and response (EDR).
To exploit this technique, a user must be convinced to run a program that uses UI Automation. This can lead to stealthy command execution, which can harvest sensitive data, redirect browsers to phishing websites, and more.
Detection of this technique is challenging in several ways, including for EDR. All EDR technologies we have tested against this technique were unable to find any malicious activity.
This technique can be used on every Windows endpoint with operating system XP and above.
In this blog post, we provide a full write-up on how to (ab)use the UI Automation framework (including possible attacks that could leverage it) and we present a proof of concept (PoC) for each abuse vector we discuss. We also provide detection and mitigation options.
This post is about sandbox evasion techniques and their usefulness in more targeted engagements.
There's a lot of sandbox evasion techniques, some are simple: query WMI, some are cool: parsing SMBIOS tables, most try to detect sandbox artifacts. I wanted to know if these techniques are still effective for detecting sandboxes, or if the sandboxes have since been updated to counter them.
In today’s post, We’ll explore the process of designing and developing malware for macOS, which is a Unix-based operating system. We’ll use a classic approach to understanding Apple’s internals. To follow along, you should have a basic understanding of exploitation, as well as knowledge of C and Python programming, and some familiarity with low-level assembly language. While the topics may be advanced, I’ll do my best to present them smoothly.
You can hide data in directory structures, and it will be more or less invisible without knowing how to decode it. It won't even show up as taking up space on disk. However, its real-world applications may be limited because it is the code execution itself which is often the difficulty with AV/EDR evasion.
Elastic researchers uncovered a new technique, GrimResource, which allows full code execution via specially crafted MSC files. It underscores a trend of well-resourced attackers favoring innovative initial access methods to evade defenses.
Using machine learning to target stockpiled malicious domains, the results of our detection pipeline tool highlight campaigns from phishing to scams.
Introducing a novel technique for e-mail spoofing
Our research team is committed to continuously identifying potential security vulnerabilities and techniques that threat actors may exploit to bypass existing security controls. In this blog post, our team is detailing on a comprehensive research specifically focused on process injection techniques utilized by attackers to deceive robust security products integrated into the security stack, such as EDRs and XDRs. Throughout the blog post, we will delve into various process injection techniques e
This article explores a phishing technique that emulates a file archiver software in the browser while using a .zip domain.
A newly uncovered technique to abuse Google’s ad-words powerful advertisement platform is spreading rogue promoted search results in mass. Pointing to allegedly credible advertisement sites that are fully controlled by threat actors, those are used to masquerade and redirect ad-clickers to malicious phishing pages gaining the powerful credibility and targeting capabilities of Google’s search results. Adding customized malware payloads, threat actors are raising the bar for successful malware deployments on Personal PCs with ad words like Grammarly, Malwarebytes, and Afterburner as well as with Visual Studio, Zoom, Slack, and even Dashlane to target organizations.
Partially encrypting victims' files improves ransomware speed and aids evasion. First seen in LockFile, the technique is now being widely adopted.
A new wave of phishing is currently circulating (a related story from derstandard.at newspaper can be found here). Documents are said to have been sent to you from a scanner, which you can allegedly download, as can be seen in the following image
