Privacy-Preserving Text Indexing for Search of Documents

Protection of content of sensitive text documents is important in enterprise intranets. An index structure is needed to support efficient search and retrieval, but it can lead to information leakage; by statistical attacks an adversary can draw probabilistic inference about the contents of document collection. Zerr and others present a confidential index structure and the ranking of retrieved documents for the query, but only for singleterm queries. The solution proposed in the paper generalizes Zerr’s method by using an anonymization parameter and query-dependent anonymized inverse document frequency factors; thereby it provides better ranking and gives possibility of multi-term queries. Key-Words: information retrieval, confidential text indexing, inverted index, posting list, inverse document frequency, r-confidentiality, anonymization, ranking