The exponential growth of textual data, particularly in Vision-and-Language Navigation (VLN) applications, poses significant challenges for efficient storage and management in cloud-based environments. While data deduplication is a vital technique for minimizing storage requirements, it often introduces critical security concerns. This paper proposes a novel deduplication framework aimed at enhancing storage efficiency without compromising data security. By integrating deduplication processes on both the client and cloud sides, the proposed system effectively reduces data redundancy while safeguarding confidentiality. Its lightweight preprocessing design makes it well-suited for deployment on resource-limited devices, such as those in IoT ecosystems. Furthermore, the system incorporates advanced security measures to defend against side-channel attacks and unauthorized access. Experimental evaluations using the Touchdown dataset reveal that the proposed framework achieves a notable compression rate of approximately 66%, significantly reducing storage overhead while preserving data integrity. These results underscore the system’s potential for enabling secure and scalable textual data management in modern cloud infrastructures.