Cybersecurity experts have identified two significant security vulnerabilities in Wondershare RepairIt, a popular AI-driven data repair and photo editing application. These flaws not only jeopardize user privacy but also open avenues for potential supply chain attacks through AI model manipulation.
Overview of the Vulnerabilities
The identified vulnerabilities are:
– CVE-2025-10643: An authentication bypass issue stemming from overly permissive permissions granted to a storage account token.
– CVE-2025-10644: Another authentication bypass flaw related to the permissions associated with a Shared Access Signature (SAS) token.
Both vulnerabilities have been assigned high severity scores, with CVE-2025-10643 rated at 9.1 and CVE-2025-10644 at 9.4 on the Common Vulnerability Scoring System (CVSS). Exploitation of these flaws could allow attackers to circumvent authentication mechanisms, potentially leading to arbitrary code execution on users’ systems.
Implications of the Security Flaws
The ramifications of these vulnerabilities are extensive:
1. User Data Exposure: The application reportedly collects and stores user data without proper encryption. This data, stored in cloud environments, becomes susceptible to unauthorized access due to the embedded permissive cloud access tokens within the application’s code.
2. AI Model Tampering: The exposed cloud storage contains not only user data but also AI models, software binaries, container images, scripts, and company source code. This exposure provides malicious actors the opportunity to alter AI models or executables, potentially leading to supply chain attacks that could affect downstream users.
3. Supply Chain Risks: Given that the application automatically retrieves and executes AI models from unsecured cloud storage, attackers could modify these models or their configurations. Such manipulations could distribute malicious payloads to legitimate users through vendor-signed software updates or AI model downloads.
Discovery and Disclosure
Trend Micro researchers Alfredo Oliveira and David Fiser discovered these vulnerabilities and disclosed them through the Zero Day Initiative (ZDI) in April 2025. Despite multiple attempts to contact Wondershare, the vendor has not responded, and no patches have been released to address these issues.
Recommendations for Users
In light of the absence of fixes, users are advised to limit their interaction with Wondershare RepairIt. Implementing robust security practices, including regular software updates and cautious data handling, is crucial to mitigate potential risks.
Broader Implications for AI and Security
This incident underscores the critical need for integrating security measures throughout the software development lifecycle, especially in AI-driven applications. Exposing Model Context Protocol (MCP) servers without authentication or storing sensitive credentials in plaintext can lead to severe security breaches. Organizations must prioritize security to protect user data and maintain trust.
