Cybersecurity researchers from Trend Micro uncovered two critical authentication bypass vulnerabilities (CVE-2025-10643 and CVE-2025-10644, CVSS scores 9.1 and 9.4) in Wondershare RepairIt, an AI-powered data repair tool. The flaws stemmed from overly permissive cloud storage tokens embedded in the application’s code, granting read/write access to sensitive data—including user-uploaded images/videos, AI models, software binaries, source code, and container images—all stored unencrypted. Exploitation could enable attackers to bypass authentication, tamper with AI models or executables, and launch supply chain attacks via vendor-signed updates or malicious AI model downloads. The exposed data included customer private information, intellectual property (e.g., patents, source code), and proprietary AI models, risking intellectual property theft, regulatory penalties, and severe erosion of consumer trust. Trend Micro’s responsible disclosure via the Zero Day Initiative (ZDI) in April 2025 went unanswered, leaving users vulnerable. The incident highlights critical gaps in DevSecOps practices, where rushed innovation overlooked security implications, enabling potential large-scale abuse of user data and system integrity.
Source: https://thehackernews.com/2025/09/two-critical-flaws-uncovered-in.html
TPRM report: https://www.rankiteo.com/company/wondershare-recoverit
"id": "won5303953092525",
"linkid": "wondershare-recoverit",
"type": "Vulnerability",
"date": "4/2025",
"severity": "100",
"impact": "5",
"explanation": "Attack threatening the organization's existence"{'affected_entities': [{'customers_affected': 'Global user base (exact number '
'undisclosed)',
'industry': 'Technology (Multimedia & AI Tools)',
'location': 'Shenzhen, China (HQ)',
'name': 'Wondershare',
'type': 'Software Vendor'}],
'attack_vector': ['Network',
'Misconfigured Cloud Storage',
'Embedded Overly Permissive Tokens'],
'customer_advisories': ['Public warning issued via media and security blogs'],
'data_breach': {'data_encryption': 'None (data stored in plaintext)',
'data_exfiltration': 'Potential (no confirmed exfiltration '
'reported, but exposure enables abuse)',
'file_types_exposed': ['Images',
'Videos',
'Executables',
'Container images (Docker)',
'Scripts',
'Source code files'],
'personally_identifiable_information': 'Yes (user-uploaded '
'content may include '
'PII)',
'sensitivity_of_data': 'High (includes proprietary AI models, '
'source code, and unencrypted PII)',
'type_of_data_compromised': ['User-generated content '
'(images/videos)',
'AI models',
'Software binaries',
'Container images',
'Scripts',
'Source code',
'PII']},
'date_publicly_disclosed': '2025-05',
'description': 'Cybersecurity researchers at Trend Micro disclosed two '
'critical-rated authentication bypass vulnerabilities '
'(CVE-2025-10643 and CVE-2025-10644) in Wondershare RepairIt, '
'an AI-powered data repair and photo editing application. The '
'flaws stem from overly permissive cloud access tokens '
'embedded in the application’s code, granting read/write '
'access to sensitive cloud storage without encryption. '
'Successful exploitation could bypass authentication, enable '
'supply chain attacks via tampered AI models or executables, '
'and expose user data (including images, videos, source code, '
'and container images). The vulnerabilities also contradict '
'Wondershare’s privacy policy, as the company was found to '
'collect and leak private user data due to weak DevSecOps '
'practices. Trend Micro responsibly disclosed the issues in '
'April 2025 via its Zero Day Initiative (ZDI), but Wondershare '
'has not responded or issued a fix. Users are advised to '
'restrict interaction with the product until patches are '
'available. The incident highlights broader risks in AI/ML '
'supply chains, including model tampering, intellectual '
'property theft, regulatory penalties, and erosion of consumer '
'trust.',
'impact': {'brand_reputation_impact': 'High (erosion of consumer trust due to '
'privacy policy violations and data '
'leakage)',
'data_compromised': ['User-uploaded images/videos',
'AI models',
'Software binaries',
'Container images',
'Scripts',
'Company source code',
'Personally Identifiable Information (PII)'],
'identity_theft_risk': 'High (unencrypted PII exposure)',
'legal_liabilities': ['Potential GDPR/CCPA violations',
'Regulatory fines'],
'operational_impact': ['Supply chain disruption',
'Loss of customer trust',
'Potential regulatory penalties'],
'systems_affected': ['Wondershare RepairIt application',
'Cloud storage (Azure/S3-like)',
'AI model repositories',
'Container registries']},
'investigation_status': 'Ongoing (vendor unresponsive; third-party research '
'disclosed)',
'lessons_learned': ['Embedding overly permissive cloud access tokens in '
'application code introduces critical risks.',
'Lack of encryption for sensitive data in cloud storage '
'exacerbates exposure.',
'Weak DevSecOps practices can lead to privacy policy '
'violations and supply chain vulnerabilities.',
'AI/ML supply chains require rigorous security controls '
'to prevent model tampering and downstream attacks.',
'Responsible disclosure does not guarantee vendor '
'responsiveness; users may need to take proactive '
'mitigation steps.'],
'motivation': ['Data Theft',
'Supply Chain Compromise',
'AI Model Manipulation',
'Intellectual Property Theft'],
'post_incident_analysis': {'corrective_actions': ['Token rotation and '
'least-privilege access '
'enforcement (pending '
'vendor action).',
'Data encryption '
'implementation for cloud '
'storage.',
'Security audit of CI/CD '
'pipelines and third-party '
'dependencies.',
'User awareness campaigns '
'on supply chain risks in '
'AI/ML tools.'],
'root_causes': ['Overly permissive cloud access '
'tokens embedded in application '
'code.',
'Lack of encryption for sensitive '
'data in cloud storage.',
'Weak DevSecOps practices (e.g., '
'hardcoded credentials, '
'insufficient access controls).',
'Inadequate vendor response to '
'responsible disclosure.']},
'recommendations': ['Immediately revoke and rotate all embedded cloud access '
'tokens in Wondershare RepairIt.',
'Implement least-privilege access controls for cloud '
'storage and enforce encryption for data at rest.',
'Conduct a thorough security audit of DevSecOps pipelines '
'to identify and remediate similar vulnerabilities.',
'Monitor cloud storage and container registries for '
'unauthorized access or tampering.',
'Users should avoid using Wondershare RepairIt until '
'patches are released and verified.',
'Organizations should assess third-party AI/ML tools for '
'supply chain risks before deployment.',
'Adopt security best practices for MCP servers, including '
'authentication and credential protection.'],
'references': [{'date_accessed': '2025-05',
'source': 'Trend Micro Research Blog',
'url': 'https://www.trendmicro.com/en_us/what-is/zero-day-initiative.html'},
{'source': 'Kaspersky PoC on MCP Server Risks'},
{'date_accessed': '2025-05',
'source': 'Palo Alto Networks Unit 42 Report on AI Code '
'Assistant Vulnerabilities'},
{'source': "Checkmarx Research on 'Lies-in-the-Loop' (LitL) "
'Attacks'}],
'regulatory_compliance': {'regulations_violated': ['Potential GDPR (EU)',
'CCPA (California)',
'Other regional data '
'protection laws']},
'response': {'communication_strategy': ['Public disclosure by Trend Micro',
'Media coverage'],
'containment_measures': ['Users advised to restrict interaction '
'with Wondershare RepairIt'],
'incident_response_plan_activated': 'No (vendor unresponsive)',
'third_party_assistance': ['Trend Micro (disclosure via Zero Day '
'Initiative)']},
'stakeholder_advisories': ['Trend Micro advisory to restrict use of '
'Wondershare RepairIt'],
'title': 'Critical Authentication Bypass Vulnerabilities in Wondershare '
'RepairIt Exposing User Data and AI Model Tampering Risks',
'type': ['Authentication Bypass',
'Supply Chain Attack',
'Data Leakage',
'AI Model Tampering'],
'vulnerability_exploited': [{'cve_id': 'CVE-2025-10643',
'cvss_score': 9.1,
'description': 'Authentication bypass '
'vulnerability in permissions '
'granted to a storage account '
'token'},
{'cve_id': 'CVE-2025-10644',
'cvss_score': 9.4,
'description': 'Authentication bypass '
'vulnerability in permissions '
'granted to an SAS token'}]}