Hierarchical Indexing with Controlled Expansion for Efficient Semantic Search over Encrypted Cloud Data

Area of Science:

• Computer Science
• Cryptography
• Information Retrieval

Background:

• Cloud data outsourcing necessitates efficient encrypted data search.
• Existing schemes struggle with index stability and query expansion noise, impacting accuracy and efficiency.

Purpose of the Study:

• To propose SES-HI, a Semantically Enhanced Searchable Encryption scheme.
• To enable efficient ranked semantic search over encrypted cloud data.

Main Methods:

• Constructed a hierarchical index using a two-stage clustering pipeline (Ward → k-means).
• Implemented topic-guided query expansion (LDA) with Word2Vec-based similarity filtering.
• Introduced a dual-pruning strategy (global threshold + top-k competitive pruning).

Main Results:

• Achieved tighter pruning bounds and reduced query cost.
• Enhanced semantic coverage while suppressing expansion noise.
• Demonstrated improved accuracy-latency trade-off compared to baselines on TREC dataset.

Conclusions:

• SES-HI offers efficient and accurate ranked semantic search over encrypted cloud data.
• The scheme is secure against adaptive chosen-keyword attacks.
• SES-HI supports practical privacy-sensitive cloud applications.