Wondershare RepairIt Exposed User Photos and AI Models via Hardcoded Cloud Credentials
Trend Micro researchers found that Wondershare RepairIt hardcoded write-capable cloud credentials in its binary, exposing user photos, AI models, and signed executables to supply-chain attacks.
Trend Micro Research has disclosed that Wondershare RepairIt, an AI-powered application for repairing damaged images and videos, exposed sensitive user data and internal assets through insecure cloud storage practices. The analysis, published on September 23, 2025, reveals that the application hardcoded overly permissive cloud access tokens directly into its compiled binary, granting both read and write access to a cloud storage bucket that contained not only user-uploaded photos but also AI models, software binaries, container images, and company source code.
The vulnerabilities, assigned CVE-2025-10643 and CVE-2025-10644, were disclosed to the vendor through Trend Micro's Zero Day Initiative in April 2025. Despite the company's privacy policy explicitly stating that user data would not be stored, researchers observed that sensitive user photos were retained and subsequently exposed due to the security oversights. The cloud storage bucket was configured without encryption, making the data accessible to anyone with basic technical knowledge who could download and exploit it.
The hardcoded credentials included cloud object storage identifiers, a secret access ID and key, and defined bucket names, all embedded in the application's binary. This configuration allowed attackers to not only access user data but also manipulate AI models or executable files, potentially conducting sophisticated supply chain attacks. Such an attack could distribute malicious payloads to legitimate users through vendor-signed software updates or AI model downloads, compromising the integrity of the software supply chain.
Trend Micro noted that it is unclear whether the vulnerable code was written by a developer or an AI coding agent that provided a vulnerable code snippet. Regardless, the incident highlights the critical importance of secure DevSecOps practices, particularly when integrating cloud services. The researchers emphasized that a single access token leakage can have disastrous consequences, including the insertion of malicious code into distributed software binaries.
The exposed cloud storage contained multiple signed application executables developed by Wondershare, in addition to AI models and user data. This means an attacker with access to the bucket could replace legitimate signed binaries with trojanized versions, which would then be distributed to users through official update channels. The write permissions on the token made this scenario particularly dangerous, as it allowed both reading and writing to the storage.
Trend Micro made proactive efforts to contact Wondershare through the Trend Zero Day Initiative, but had not received a response as of the publication date. The final draft of the blog entry was shared with the vendor prior to publication. The vulnerabilities were disclosed on September 17, 2025, and the research was made public on September 23.
This case underscores the growing risks associated with AI-powered applications that handle sensitive user data. As AI tools become more prevalent, organizations must ensure that their data handling practices align with their privacy policies and that cloud storage configurations are properly secured. The incident also serves as a reminder that hardcoded credentials, even when intended for convenience, can expose organizations to significant vulnerabilities if not implemented with strict access controls.