SecurityBlog

Blogs

Report says ChatGpt Atlas is Vulnerable for Users: Understanding Open-AI Agent Mode

Atlas’s autofill and form interaction capabilities present potential attack points

As per reports ChatGpt Atlas browser is vulnerable to attacks and is laced with inherent weakness in comparison to other browser like Google Chrome. As per ‘LayerX ‘who discovered the weakness in ChatGpt Atlas, described threat actors have the ability to inject malicious instructions into ChatGPT’s ‘memory’ and execute remote code and this works by way of cross-site request forgery requests.

These exploit can allow attackers to infect systems with malicious code, grant themselves access privileges or deploy malware. “Understanding “Agent Mode” is most important and core of Atlas which is not same for any traditional browsers. In traditional browser where users manually move from site to site, agent mode allows ChatGPT to semi-autonomously operate your browser.

For e.g. any user wanting to use ChatGPT for work related purposes, the malicious code planted earlier mostly tainted will be invoked automatically to execute remote code, allowing attackers to gain control of the user account .This may include their browser, code they are writing or systems they have access to.

Rate of Vulnerability is 90% A Warning for Users

The rate of vulnerability is 90% then other browsers as when an attacker wish they can push or inject  malicious instructions into ChatGPT’s Atlas ‘memory’ and later execute via remote code.

There is a more basic warning as well. “Atlas does not include meaningful anti-phishing protections, meaning that users of this browser are “up to 90% more vulnerable to phishing attacks than users of traditional browsers,” LayerX says.

Key pointers from research

ChatGPT’s Atlas is not resilient to Phishing attacks

Out of 103 in-the-wild attacks that LayerX tested 97 to go through, a whopping 94.2% failure rate

Compared to Edge (which stopped 53% of attacks in LayerX’s test) and Chrome (which stopped 47% of attacks),

ChatGPT Atlas was able to successfully stop only 5.8% of malicious web pages

Unlike traditional web browsers where you manually navigate the internet, agent mode allows ChatGPT to operate your browser semi-autonomously.

The technology works by giving ChatGPT access to your browsing context. It can see every open tab, interact with forms, click buttons and navigate between pages just as you would.

Importance of Security by Design for web browsing & How AI is intricately involved

The sandboxing approach which is security by design is to keep websites isolated from attacks and prevent malicious code from accessing data from other tabs is crucial to modern web architecture. This is the basis of modern web that depends on separation. But if its not implemented what can be the impact.

But in Atlas, the AI agent isn’t malicious code – it’s a trusted user with permission to see and act across all sites. In this browser isolation is not required. Here AI is not directly connected to the threat but what AI does is AI following a hostile command hidden in the environment. This opens doors to security and privacy risks many users are ill-equipped to handle.

Let me put an example : If you search for air tickets and visit a site , the Atlas ChatGpt will prompt and try to book a ticket or you search for movies in near by theater ,it attempts to book a ticket ”, it will explore options and try to book reservation. Atlas autofill’s and form interaction capabilities present potential attack points, especially when AI is making rapid decisions about information entry and submission.

This is possible when access is granted to ChatGPT for any browsing requirement or context that allows it to view and open tabs, interact with forms and navigate between pages like humans do.

Is User’s security getting compromised

The above example gives users warning that any AI powered browser may be convenient but not without security risks and those who are ChatGpt Atlas, should give extreme cautious before choices are made . Do not share browsing history with any AI mode, instead adopt incognito mode. Any malicious code can  influence the AI’s behavior if browsing and this can happen across multiple tabs.

In case of Atlas, the condition is more vulnerable as Atlas provides inputs like humans doing and AI in disguise executing harmful commands within the environment.

Will AI Agent or Open AI make browsing safe for users or what it means to have safe browsing.

(Source: https://www.bbc.com/news/articles/c20pdy1exxvo)

Blogs Security Advisory

Samsung Galaxy S25 Zero-Day Exploit Exposes Camera & Location 

Summary 

At Pwn2Own Ireland 2025, researchers Ben R. and Georgi G. from Interrupt Labs successfully exploited a zero-day vulnerability in the Samsung Galaxy S25. The flaw allowed them to gain remote control of the device, activate the camera, and track the user’s real-time location without interaction.

This achievement, earning them $50,000 and 5 Master of Pwn points, highlighted ongoing security weaknesses even in flagship smartphones with extensive testing. The exploit’s discovery underlined broader concerns about the pace of Android feature development outstripping security hardening efforts across system and multimedia libraries. 

The Galaxy S25 zero-day exploit underscores the persistent threat of critical security flaws even in top-tier consumer devices. Although discovered in a controlled, ethical hacking event, such vulnerabilities pose substantial risks if leveraged by malicious actors.

Vulnerability Details 

The vulnerability originated from an improper input validation issue within the Galaxy S25’s software stack. Through carefully crafted malicious inputs, the researchers bypassed Samsung’s built-in security safeguards and executed arbitrary code remotely.

The exploit provided persistent access, enabling control over cameras, GPS, and potentially other sensitive device components, effectively transforming the smartphone into a covert surveillance tool. Because the issue existed at a deep system level, it required no user interaction, making it particularly severe. The vulnerability had not been previously disclosed, meaning Samsung and the public were both unaware until the competition’s revelation. 

Key characteristics: 

The key characteristics of the Samsung Galaxy S25 zero-day vulnerability are as follows: 

  • Type of Vulnerability: Improper input validation bug within the device’s software stack, allowing remote code execution without user interaction.​ 
  • Impact: Enables attackers to take full control of the device, activate the camera, and track real-time GPS location, effectively turning the device into a surveillance tool.​ 
  • Discovery and Exploit: Uncovered during Pwn2Own Ireland 2025 by researchers Ben R. and Georgi G., showcasing a sophisticated exploit chain that bypassed Samsung’s security measures.​ 
  • Persistence: Vulnerability allows persistent access, which can be exploited silently without user awareness or interaction.​ 
  • Disclosure and Remediation: The flaw was previously undisclosed, with responsible disclosure leading to Samsung preparing a security patch. No official statement has been issued yet, but a fix is anticipated.​ 
  • Severity and Potential Damage: The exploit can compromise sensitive personal data, private communications, and location, highlighting significant privacy and security risks. 

Attack Flow 

Step Description 
1. Craft Malicious Input  Attackers develop specially crafted malicious inputs targeting the vulnerable components within the Samsung Galaxy S25’s software stack, particularly exploiting the improper input validation flaw. 
2. Deliver Payload The malicious payload is delivered via crafted multimedia or system input, such as manipulated images or software commands, that bypass Samsung’s existing safeguards. 
3. Bypass Security Measures The input validation flaw allows the malicious data to bypass security checks, executing remote code without requiring user interaction or consent, gaining initial access to the device’s system. 
4. Gain Persistent Control Once the malicious code executes, attackers establish persistent control over the device, enabling continuous access to core functionalities like camera activation and GPS tracking silently and covertly. 
5. Exploit Device Capabilities Attackers leverage control to activate the device’s camera and GPS in real-time, turning the device into a surveillance tool capable of capturing photos, videos, and tracking location discreetly. 
6. Maintain Stealth & Avoid Detection The exploit chain is designed to evade detection by Samsung’s defenses during the attack window, allowing attackers to operate covertly without triggering security alerts or user notifications. 
7. Exploit and Monetize The compromised device becomes a tool for espionage, data theft, or targeted surveillance, which can be exploited for malicious purposes or sold on criminal markets if attacker exploits are monetized. 

Proof-of-Concept 

The proof-of-concept for the Samsung Galaxy S25 zero-day vulnerability (CVE-2025-21043) demonstrates how specially crafted malicious images can exploit an out-of-bounds write flaw in Samsung’s closed-source image parsing library libimagecodec.quram.so. This flaw allows remote code execution with elevated privileges without requiring user interaction.

The exploit involves delivering a malicious payload embedded in an image file that, when processed by the vulnerable library, triggers memory corruption leading to arbitrary code execution and persistent control over the device.

This has been confirmed in cybersecurity forums and independent analyses, with active exploitation observed in the wild primarily via social engineering through messaging platforms like WhatsApp. The PoC confirms that attackers can bypass conventional security mechanisms and gain deep system control, enabling surveillance actions such as camera activation and location tracking. This underscores the critical need for applying the latest security patches released by Samsung.  

Source: https://x.com/thezdi/status/1981316237897396298 

Why It’s Effective 

  • Code Execution via Input Validation Flaw: Exploits improper input validation within the Galaxy S25’s software stack, allowing malicious payloads to bypass safeguards and execute remote code seamlessly alongside legitimate system processes. 
  • Zero-Click Capability: Operates without requiring any user interaction, enabling silent compromise through automated payloads that trigger upon data processing or system-level input handling. 
  • Persistent Access: Establishes continuous control after initial compromise, granting long-term ability to activate hardware components like camera and GPS without detection by standard security mechanisms. 
  • Stealth Operations: Exploit chain hides within multimedia and system library processes, avoiding visible alerts or performance anomalies that might indicate compromise to the user. 
  • Advanced Evasion: Utilizes legitimate system libraries and resource calls, reducing the likelihood of being flagged by mobile antivirus or Samsung Knox runtime protections. 
  • High Impact Vector: Enables complete device surveillance, capturing photos, videos, and location data covertly, illustrating real-world severity when attackers weaponize such system-level access. 

Remediation: 

  • Update Samsung Galaxy devices immediately with the latest September 2025 Security Maintenance Release (SMR) patch that fixes CVE-2025-21043. 
  • Manually check for software updates via Settings > Software Update > Download and Install to ensure the fix is applied promptly. 
  • Enable automatic security updates on Samsung devices for timely future patching without delay. 
  • For enterprises, enforce patch deployment policies through Mobile Device Management (MDM) or Enterprise Mobility Management (EMM) tools to cover all mobile endpoints. 
  • Restrict app permissions, especially camera and location access, to minimize exposure in case of compromise. 
  • Avoid opening images from untrusted sources or suspicious messaging apps, as the vulnerability exploits image parsing. 
  • Implement continuous mobile threat detection to identify abnormal device behavior indicative of compromise. 
  • Educate users and IT teams about the critical nature of this vulnerability and the importance of timely patching. 

This ensures comprehensive mitigation of vulnerability while reducing risk and exposure to active exploits. 

Conclusion: 


This incident reinforces the value of responsible disclosure mechanisms like Pwn2Own, where manufacturers receive detailed technical reports to develop patches before public release. Samsung has yet to issue a formal statement but is expected to roll out a security update imminently.

In the meantime, users are advised to enable automatic updates, remain cautious with app permissions and untrusted networks, and monitor official channels for patches to mitigate potential exploitation risks. 

References: 

Blogs

Discord Security Incident Reveal Support Ticket Stolen in Third-Party Breach 

Summary 

In today’s interconnected digital world, trust often reaches beyond the main platforms to include the network of partners that support them. Recently, Discord disclosed an incident tied not to its own systems, but to a third-party customer service provider whose systems were compromised, exposing limited user information.

The company emphasized that its core infrastructure remained secure, with the issue confined to the external vendor’s network. The cyber attack appears to be financially motivated, as hackers demanded a ransom from Discord in exchange not to reveal any information that they ceased.

Breach Breakdown 

Discord confirmed that an unauthorized party gained access to the vendor’s systems by exploiting that firm’s ticketing platform. Through that entry point, attackers were able to view limited user information exchanged during support requests like names, Discord handles, emails and some contact details.

For a small number of affected users, the exposure extended to scanned government-issued IDs, such as passports or driver’s licenses, typically used to verify age or ownership. 

Discord as well clarified that its core infrastructure like chat servers, authentication databases, and private messaging systems were not breached. This wasn’t a platform intrusion, but rather a supply chain compromise through one of the company’s external service tools. 

Upon discovering the breach, Discord revoked the vendor’s access immediately, launched an internal investigation. Law enforcement agencies have also joined the effort to identify the perpetrators and prevent further misuse of stolen data. 

Discord already notified data protection authorities, contacted affected users directly via noreply@discord.com, and reviewed all vendor relationships to ensure compliance with data protection standards. The company also pledged to strengthen third-party systems and increase security oversight for partners with data access.

Additionally, Discord advised users to watch for phishing attempts and reiterated that it will never contact them by phone regarding the incident. 

Recommendations: 

Here are some recommendations below 

  • Always verify the sender before clicking links in security emails. 
  • Enable multifactor authentication to protect your account even if credentials leak. 
  • Stay alert for phishing emails, especially those that sound urgent or official. 
  • Keep your data footprint minimal by sharing only what’s necessary. 
  • Regularly assess vendor security and treat third-party reviews as a key defense measure. 

Conclusion 
This incident underscores that even well-secured platforms like Discord remain vulnerable through their third-party partners. It highlights the growing importance of robust vendor risk management, transparent communication, and continuous security auditing.

For users, it’s a reminder to stay cautious, enable strong authentication measures, and practice vigilance against phishing or social engineering attempts following any major data disclosure. 

Discord was created as a communication platform for gamers, who represent more than 90% of the userbase, but expanded to various other communities, allowing text messages, voice chats and video calls.

References: 

Security Advisory

VoidProxy PhaaS Uses MFA Bypass, Hijacking Google & Microsoft Logins

Security Advisory

Security researchers from Okta have uncovered a stealthy and sophisticated Phishing-as-a-Service (PhaaS) framework known as VoidProxy.

This has been used to hijack Microsoft, Google and even integrated SSO accounts protected by providers like Okta. Unlike traditional phishing kits, VoidProxy employs Adversary-in-the-Middle (AiTM) tactics to capture real-time credentials, MFA tokens and bypassing several standard authentication protections.

VoidProxy’s infrastructure leverages disposable domains, Cloudflare protections, dynamic DNS which all of mimicking as legitimate enterprise setups becoming extremely difficult to detect, analyze. The attackers are running phishing campaigns with little technical effort, enabling wide-scale compromises that lead to email compromise, fraud and data breaches.

Its attack chain is built to evade modern email security, identity defenses, and analysis tools by leveraging the following:

  • CAPTCHA Filtering: Victims are first shown a CAPTCHA challenge before any phishing content loads. This helps block bots and automated security scanners.
  • Cloudflare Workers: Used to deliver customized phishing pages and smartly direct traffic to the attacker’s backend servers.
  • URL Redirection Chains: The phishing links in emails go through several redirects (often using shortened URLs) before landing on fake login pages. This helps bypass spam filters and security tools.
  • Dynamic DNS: These services let attackers quickly create domain names that point to specific IP addresses, making their infrastructure flexible and harder to track.    

Once a user enters their credentials and MFA tokens, the session is hijacked via a reverse proxy server, allowing the attacker to immediately access the legitimate account.

Here are some shortened url links

Attack Flow

StepDescription
1. DeliveryPhishing emails are sent from compromised accounts on email delivery services (like Postmarkapp or Constant Contact) increasing trust and shortening URL services for bypassing spam filters.
2. Redirecting & FilterClicking the phishing link redirects victims through several short URLs and presents a Cloudflare captcha to ensure human interaction.
3. PhishingVictims land on a fake Microsoft or Google login page using realistic subdomain patterns like “login.<phishing_domain>.<.com/.io>”. Additionally, integrated SSO accounts are redirected to additional fake SSO pages mimicking the login flows.
4. AiTM Session HijackThe backend proxy captures credentials, MFA tokens and session cookies, allowing attackers full account access.
5. ExfiltrationSession cookies and credentials are routed to the attacker’s admin panel in real-time. Integration with bots or webhooks enables instant alerts to the attackers.

Why It’s Effective

AiTM Infrastructure: Unlike static phishing kits, VoidProxy runs a live proxy in the middle of the authentication flow, stealing session tokens or mfa token immediately after login.

CAPTCHA & Cloudflare Layers: These challenges ensure only real human victims reach the phishing payload, filtering out scanners and sandboxes.

Integrated SSO Targeting: Accounts using Okta or other SSO providers are redirected to accurate second-stage phishing pages, increasing the likelihood of a full compromise.

Recommendations:

Here are some recommendations below

  • Harden the authentication by bind sessions to IP addresses (IP Session Binding) to block cookie replay attacks.
  • Block access from rarely used IP ranges or unmanaged devices.
  • Provide user awareness training to help recognize phishing links, suspicious email senders and fake login prompts.
  • Keep monitoring for any indications of suspicious activities.

Conclusion
VoidProxy’s layered architecture, real-time session hijacking and deep evasion mechanisms make it a potential threat even for environments with multi-factor authentication in place. We require a shift from traditional phishing detection toward real-time risk-based access controls, strong authenticators and persistent user education.

References:

Scroll to top