Baffle, a leader in data protection, has announced a significant expansion of its capabilities to include pgvector on PostgreSQL, making it the first to offer Real Queryable Encryption (RQE) for vector databases. This extension provides enhanced security for sensitive data stored in vector databases, such as text and embeddings, used by Generative AI (GenAI) applications. As organizations increasingly rely on vector databases for memory and retrieval tasks in their GenAI applications, this new feature ensures sensitive data remains protected while maintaining functionality like similarity searches and other operations.
1. Baffle’s Real Queryable Encryption for Vector Databases
Baffle’s Real Queryable Encryption now secures data within vector databases like pgvector on PostgreSQL, ensuring that embeddings of sensitive information can be processed by GenAI applications without exposure:
- Encryption of Sensitive Data: Embeddings, which often contain sensitive information, are securely encrypted while still allowing for operations such as similarity searches.
- No Code Changes Required: Organizations can apply encryption without modifying existing application code, maintaining ease of integration and security.
2. The Need for Enhanced Security in GenAI Applications
According to Gartner, vector databases and large language models (LLMs) used in GenAI applications are not inherently secure and could expose sensitive data. Baffle’s approach addresses this critical security gap:
- Security in GenAI: Baffle ensures that even while GenAI applications process sensitive data, it remains encrypted, meeting current data security and privacy compliance requirements.
- Embedding Protection: The embeddings in vector databases are kept secure, even while GenAI models perform necessary operations on them.
3. Baffle’s Comprehensive Data Protection Platform
Baffle’s data protection platform offers a no-code solution for securing sensitive data, and its Real Queryable Encryption enhances security for vector databases:
- Enterprise-Class Security: Baffle supports encryption, tokenization, and masking with role-based access control at multiple levels, including logical database, column, row, and field levels.
- Cloud Data Security: Baffle protects regulated data at rest, in use, and in transit across cloud environments, supporting PostgreSQL databases on services like Amazon RDS and Amazon Aurora.
4. Looking Forward: Expanding to More Vector Databases
Currently, Baffle’s encryption capabilities are available for pgvector on PostgreSQL, but the company plans to extend this feature to additional vector databases in the future, further enhancing GenAI application security.
Baffle’s new integration with pgvector on PostgreSQL offers organizations enhanced security for sensitive data used in GenAI applications. By implementing Real Queryable Encryption, Baffle ensures that sensitive embeddings remain protected while still enabling critical vector operations like similarity searches. As data security becomes increasingly important in the era of Generative AI, Baffle’s platform provides the necessary tools to ensure compliance with privacy and security regulations without the need for code changes.
