Efficient Single-Round Obfuscation of Search and Result Patterns in Searchable Encryption

This is the full implementation for our FROST.

WARNING: This is an academic proof-of-concept prototype and has not received careful code review. This implementation is NOT ready for production use.

Required Libraries

1. GMP 6.3.0

2. NTL v11.6.0

You can run the script file auto_setup.sh (user needs root permission) to automatically install the required libraries.

./auto_setup.sh

Dataset

1. Download Enron email dataset at: https://www.cs.cmu.edu/~enron/, then extract to obtain the folder maildir.
2. Download Wiki dataset at Wikipedia database backup dump: https://dumps.wikimedia.org/enwiki/latest/enwiki-latest-pages-articles.xml.bz2, then use WikiExtractor to extract articles in JSON format:

python3 -m wikiextractor.WikiExtractor enwiki-latest-pages-articles.xml.bz2 --json -o extracted

Finally, execute the python script in scripts/extract_random.py to randomly sample 500K documents, excluding the empty ones without any keywords.

Build & Compile

Go to the folder frost then execute:

make -j8

This is going to create an executable file named main in the folder frost.

Testing

• Go to the directory frost and launch the executable file main:

./main [-d <path_to_dataset_directory>] [-m <Bloom_filter_size>] [-n <number_of_documents>]

For example:

./main -d ./maildir 

By default, it starts the process with default parameters Bloom_filter_size = 2912 and number_of_documents = 1024.

• Wait for the initialization process to finish, then enter a keyword to perform a search (e.g., "security" without quotes). When keyword search is complete, it outputs processing latency and bandwidth overhead of keyword search for the true positive rate TPR = 0.95 w.r.t various false positive rates FPR = {0.025, 0.05, 0.1, 0.15, 0.2}.