One of the issues with the Conversation Injection technique is that the output from SearchGPT appears clearly to the user, which will raise a lot of suspicion. We discovered a bug with how the ChatGPT website renders markdown that can allow us to hide the malicious content. When rendering code blocks, any data that appears on the same line as the code block opening (past the first word) does not get rendered. This means that unless copied, the response will look completely innocent to the user, despite containing the malicious context, which will be read by ChatGPT.
Now in 2025, zero-days remain the leading method of initial compromise, accounting for roughly one-third of all intrusion attempts. It might mean setting clear policies about which tools can be used for what kinds of content. Every device that accesses sensitive communications must be kept up to date with the latest operating system patches. Antivirus software should be standard, and full-disk encryption should be mandatory. Passwords must be strong and unique, and biometric authentication should be enabled wherever possible.
In organizations like the Balearic Islands Health Service , instant messaging has become an essential tool for rapid communication between professionals, but that doesn’t mean you can use just any app you have on your personal phone. The choice of tool directly impacts patient data confidentiality and legal compliance. Messaging apps have become the backbone of modern communication — from birthday planning to boardroom discussions, and even customer support. Their convenience makes them indispensable, but it also introduces serious security risks. Although TM SGNL’s use by U.S. agencies came under scrutiny recently, contracts with TeleMessage date back to at least 2023, well before the current administration.
Bad actors such as cybercriminals might have different objectives, Hong says, “but if you just do a few relatively simple things, you can actually protect yourself from the vast majority of those kinds of threats.” The FBI and CISA also advise users to set their phones to update operating systems automatically. In full end-to-end encryption, tech companies make a message decipherable only by its sender and receiver — not by anyone else, including the company.
Finally, it’s important not to forget the role of messaging apps as a distribution channel for malicious files . The breach exploited a critical vulnerability in TM SGNL’s architecture—a lack of proper end-to-end encryption, which allowed attackers to intercept and access stored chat logs, message metadata, and other application heap dumps. As IM apps integrate more features (payments, mini-programs), their attack surfaces expand, requiring stricter access controls and faster patch cycles.
Security
The U.S. has been working since late spring to determine the extent of their activities. This month, the Biden administration said at least eight telecommunications infrastructure companies in the U.S., and possibly more, had been broken into by Chinese hackers. “These are for legitimate wiretaps that have been authorized by the courts,” Hong says. But in hackers’ hands, he says, the tools could potentially be used “to surveil communications and metadata for lots of people. And it seems like the hackers’ focus is primarily Washington, D.C.” She recommends getting 2FA messages through an app like Google Authenticator or Authy or by using a physical security key to verify access. The encrypted Signal app is what Defense Secretary Pete Hegseth and other leading national security officials within the administration used to discuss bombing Houthi sites this month.
- It is important to note that memories could contain private information about the user.
- During our research of the url_safe endpoint, we noticed that bing.com was a whitelisted domain, and always passed the url_safe check.
- As IM apps integrate more features (payments, mini-programs), their attack surfaces expand, requiring stricter access controls and faster patch cycles.
Core Dump Exposure Flaw (cve-2025-
That means that any website that is indexed on Bing has a bing.com URL that will redirect to it. Attempts to steal credentials often involve phishing attacks, malicious malware, and social engineering. Attackers are using AI to accelerate reconnaissance, compromise identities and escalate access. In a statement shared with The Hacker News, WhatsApp said it sent in-app threat notifications to less than 200 users who may have been targeted as part of the campaign. Ó Cearbhaill described the pair of vulnerabilities as a “zero-click” attack, meaning it does not require any user interaction, such as clicking a link, to compromise their device. The Meta-owned company said the issue “could have allowed an unrelated user to trigger processing of content from an arbitrary URL on a target’s device.”
Each platform served a unique audience, but all became prime targets for cybercriminals due to their scale and sensitivity of stored data. In addition to its long-term memory feature, ChatGPT considers the current conversation and context when responding. If the user asks it to remember something, or if there is some information that the engine deems important even without an explicit request, it can be remembered using memories. As is seen in the System Prompt, the memories are invoked internally using the bio tool and sent as a static context along with it. It is important to note that memories could contain private information about the user.
Sometimes ChatGPT will respond with the output of SearchGPT’s browsing results as-is, and sometimes it will take the full output and modify its reply based on the question. As a method of isolation, SearchGPT has no access to the user’s memories or context. Therefore, despite being susceptible to prompt injection in the Browsing Context, the user should, theoretically, be safe, as SearchGPT is doing the browsing. Tenable Research has discovered seven vulnerabilities and attack techniques in ChatGPT, including unique indirect prompt injections, exfiltration of personal user information, persistence, evasion, and bypass of safety mechanisms. The platform has faced similar challenges before, including the notorious XcodeGhost malware in 2015 that infected WeChat version 6.2.5 alongside 38 other popular iOS applications, affecting hundreds of millions of users.
When asked to summarize the contents of the blog, SearchGPT follows the malicious instructions from the comment, compromising the user. The Pentagon advisory, according to reports, warned against using Signal even for unclassified BestDates reviews communications, citing the alleged vulnerability exploited by Russian hackers. However, Signal countered this assertion, explaining that phishing attacks, the actual threat highlighted in the advisory, are not unique to their platform and represent a persistent risk for any popular app or website. They emphasized that these attacks do not exploit flaws in Signal’s underlying encryption technology but instead rely on deceiving users into revealing their credentials or other sensitive information. Security researchers have uncovered critical vulnerabilities in instant messaging applications that could allow attackers to gain complete control over target devices through specially crafted messages. High-security messaging apps like Signal can be compromised, either by human error or cyberattacks.
The simplest way to ensure your messages are safe from snooping is to use an end-to-end encrypted app like Signal or WhatsApp, says Eva Galperin, director of cybersecurity at the Electronic Frontier Foundation (EFF). With these apps, “your communications are end-to-end encrypted every single time,” she says. Alongside international partners, the NCSC has issued actions for individuals at risk of targeted attacks against messaging apps. While Signal encrypts message content, it still transmits metadata such as who is talking to whom and when.
A poorly secured cloud backup can completely negate the benefit of end-to-end encryption. Throughout this article, we will calmly but frankly examine why these tools are not harmless, what real risks they pose, and what alternatives and best practices exist to minimize the potential for disaster. Enjoy full access to a modern, cloud-based vulnerability management platform that enables you to see and track all of your assets with unmatched accuracy. The attack, detailed in recent research by cybersecurity firm DARKNAVY, exploits WeChat’s built-in browser components and URL parsing mechanisms to execute remote code without requiring any user interaction beyond receiving the message. newlinePostMessage-tracker is a Chrome extension specifically designed to monitor and log all postMessage communications. It captures both sent and received messages, displays their origins, and allows you to inspect the message content in detail. This tool is particularly useful for understanding the complete postMessage flow in applications with multiple iframes or windows.
Additionally, we can also notice that the postMessage data includes the OAuth token, which may be exchanged for a session token later once it reaches the parent window. Your browser’s built-in developer tools can also be deployed to trace back postMessage calls. This method is particularly useful for discovering message handlers that might be hidden in third-party scripts or dynamically loaded code.
The embedded browser components in these applications often lag behind official releases, potentially exposing users to known vulnerabilities that remain unpatched. WeChat’s file processing system, designed to enhance user experience through file previews and content extraction, creates significant security exposure when handling untrusted content. Security engineers can also leverage Uncoder AI, an IDE and co-pilot for detection engineering, which is now enhanced with a new AI Chat Bot mode and the MCP tools support. Many of the phishing attacks that compromise Signal accounts rely on social engineering. Teaching employees how to spot suspicious links, question unexpected group invites, and verify QR codes can prevent many of these attacks before they start.
Vulnerabilities found in Signal, Google Duo, Facebook Messenger, and other messaging apps allowed attackers to listen in on users without their permission, security experts have warned. The security of postMessage implementations relies heavily on proper validation of these properties, particularly event.origin. When developers skip origin validation or improperly handle message data, postMessage vulnerabilities can and will emerge. Before we move to the exploitation part, let’s first understand how we can identify potential postMessage vulnerabilities.