Privacy Preserving Ranked Multi Keyword Context Sensitive Fuzzy Search Over Encrypted Cloud Data

The privacy preserving search feature is very useful for a cloud user to retrieve the desired encrypted documents easily, securely and cost effectively in the cloud. However, a search query issued by the user may sometimes have mis-typos i.e. wrongly typed words. The mis-typos could occur because of the addition or drop of letter(s) from the word or by swapping of characters in a word. At times such mis-typos may have many spelling suggestions against them within a given threshold, of which only a few makes sense as per the context of the query. Also the mistyped word may sometimes result in another valid word from the dictionary or the word list and hence the mis-spelt word may go unnoticed. Recent works in cloud address the issue of fuzzy search. However, these approaches do not suggest a word suitable as per the context and co-occurrence with other words of the query for such mis-typos. This paper presents a privacy preserving scheme ‘Context Sensitive Fuzzy Search’ (CSFS) in a cloud computing environment that address these issues. CSFS uses Levenshtein distance and neighbor co-occurrence statistics computed from encrypted query click logs and suggest(s) word(s) from the generated suggestions and set distance as per their co-occurring frequency with other words of the query. The results achieved show that the spelling suggestions listed are as per the context of the query and have high recall value.