This project includes a set of programs that can be used to encrypt tables (consisting of rows and columns) so that the owner of the table can create tokens to allow third party users to perform queries on the encrypted tables.

The current release allows very simple queries in which it is possible to read specific columns of all rows that satisfy a certain predicate. Only equality predicates are currently supported. These roughly correspond to SQL queries of the following form:

select col3,col5 from table where col1='AA' and col2='BB' and col4='DD'

that shows the third and fifth column of all rows in which the first column is 'AA', the second is 'BB' and the fourth is 'DD'.

The following is a typical workflow:

1. Generate keys:

   run command GenKey <num_col> <key_file>

   • <num_col>: number of columns in each row of the table we intend to encrypt;
   • <key_file>: the name of the file in which the key will be saved;

2. Encrypt rows:

   run command EncRow <key_file> <rows> <encrows> <noise> <num_threads>

   • <key_file>: name of the file that contains the master key to be used for the encryption;
   • <rows>: name of the file that contains the rows to be encrypted;
   • <encrows>: the encrypted rows will be added to files <encrows>.msg and <encrows>.ct
   • <noise>: the noise parameter;
   • <num_threads>: number of threads that will execute the encryption;

3. Token generation:

   run command GenToken <key_file> <query> <noise>

   • <key_file>: name of the file that contains the master key to be used for the token generation;
   • <query>: name of the file that contains the query;
   • <noise>: the noise parameter;

4. Predicate token execution:

   run command ApplyPToken <token> <encrows> <results> <num_threads>

   • <token>: name of the file that contains the token;
   • <encrows>: the encrypted rows are found in files <encrows>.msg and <encrows>.ct;
   • <results>: name of the file in which the indices of the selected rows will be stored;
   • <num_threads>: number of threads that will execute the decryption;

5. Message token execution:

   run command ApplyMToken <token> <encrows> <results> <num_threads>

   • <token>: name of the file that contains the token;
   • <encrows>: name of the file that contains the encrypted rows;
   • <results>: name of the file that contains the indices of the selected rows that are to be decrypted;
   • <num_threads>: number of threads that will execute the decryption.

• Table format:

  The EncRow command expects rows to be structured as follows:

  row1cell1#row1cell2#row1cell3#....#row1celln
  rowmcell1#rowncell2#rowncell3#....#rowmcelln

  Cells of the same row are separated by '#' character and rows are separeted by a newline.
  Examples are in files 'row_120'(1 row, 120 cells) and 'rows_8_40'(40 rows, 8 cells).

• Query format:

  Queries for an n-column table are encoded for the GenToken command in the following way:

  Query that shows r columns c1,c2,...,cr for rows such that columni=wi for i=1,...,n (where wi is either a string or a don't care symbol) are encoded by file whose first line contains c1#c2#...#cr
  and subsequent lines contain wi or is empty, if wi is a don't care symbol. Examples are in files 'query_8' and 'query_120'.

• Random noise is added to ciphertexts during encryption (one per each row) and to tokens during token generation (one per each select parameter of the query). If a token and an encrypted row have the same noise then the query has success even if the rows would not have been selected by the query. The noise parameter denotes the size of the interval from which the random noise is picked.