Integrity Coded Database (ICDB)

What is an ICDB?

Relational databases are great for structured persistent storage, and there is no shortage of cloud providers that can store one's data without needing their own hardware or servers. However, in doing so, the data owner relinquishes control of their data and must trust that any data returning back from the cloud has not been tampered in any way. A malicious insider may be able to add, delete, or modify existing data without detection.

An ICDB is a relational database containing a middleware component that ensures that no data was tampered with or corrupted by the cloud provider. The ICDB middleware computes Integrity Codes (ICs), i.e., cryptographic hashes, and stores them alongside the original data. By recomputing the ICs for each query, they can be used to detect unauthorized data modifications.

This Java ICDB implementation can detect:

Data Addition - Adding new columns/rows to the DB
Data Modification - Modifying columns/rows in the DB
Replay Attacks / Stale Data Attacks - Old data that has been modified/removed by the data owner

Currently, this implementation cannot detect:

Data Omission - Not all data was returned

Detecting if data was omitted is much more difficult than detecting addition/modification/replay, and more research is necessary to propose a time and memory efficient architecture.

This project was a part of Boise State University's Summer 2015 Research Experience for Undergraduates (REU) in Software Security.

Initial Setup

At this time, this project only supports Linux.

To be able to build and run against a database, the following must be installed:

Maven
MariaDB (MySQL)

Then run the following commands to build the project:

$ git clone https://github.com/Hyperparticle/IntegrityCodedDatabase.git
$ cd IntegrityCodedDatabase/ICDB
$ make

Make will simply run the command mvn package -DskipTests to compile icdb.jar under the target folder.

Running the ICDB tool

The arguments for interacting with the ICDB tool is as follows:

$ icdb [-c config-file] [command] [options]

icdb is a bash script that simply runs the compiled jar. This can be run directly:

$ java -jar target/icdb-capsule.jar [-c config-file] [command] [options]

All interactions with the tool will require a config file containing a JSON object with several parameters:

ip          - the target MySQL database IP address
port        - the port the database is running on
user        - database user
password    - database password (if any)
schema      - database schema to use (for conversion)
icdbSchema  - ICDB database schema name (for execution and verification)
algorithm   - the encryption algorithm to use (RSA, AES, or SHA)
granularity - use code per field or code per tuple (FIELD or TUPLE)
macKey      - 128-bit MAC key encoded as a base64 string
rsaKeyFile  - PEM file containing public and private RSA keys

For convenience, a config file is given at ./ICDB/config.json, which will be loaded by default if the -c option is not specified.

The default config provides the following JSON object:

{
  "ip": "localhost",
  "port": 3306,
  "user": "root",
  "password": "",
  "schema": "employees",
  "icdbSchema": "employees_icdb",
  "algorithm": "SHA",
  "granularity": "TUPLE",
  "macKey": "qyPTqFrPGUpxcIo9sz2MdQ==",
  "rsaKeyFile": "key.pem"
}

Commands available

1. convert-db      - Converts an existing DB to an ICDB (both schema and data)
2. convert-query   - Converts a DB query to an ICDB query
3. execute-query   - Executes an ICDB query and verifies all returned data

There are also additional options for each command.

Convert DB Command

convert-db [--skip-duplicate] [--skip-schema] [--skip-data] [--skip-load]

The convert-db command has 4 phases, any of which can be skipped: --skip-duplicate - If set, the duplicate DB step will be skipped --skip-schema - If set, the schema conversion step will be skipped --skip-data - If set, the data conversion step will be skipped --skip-load - If set, the data load step will be skipped

Example:

$ cd <project-root>/ICDB
$ ICDBTool -c ./src/main/resources/config-sample.json convert-db --skip-duplicate

Convert Query Command

convert-query [-q query] [-g granularity]

The convert-query command takes the SQL query as an input and converts it to an ICDB query. The conversion requires: -q - The SQL query, passed in as a string -g - The granularity of the ICDB query (TUPLE or FIELD), by default TUPLE

Example:

$ cd <project-root>/ICDB
$ ICDBTool -c ./src/main/resources/config-sample.json convert-query -q "SELECT * FROM employees;" -g "FIELD"

Execute Query Command

execute-query [-q query] [--convert]

The convert-data command takes a SQL query as input, executes, then verifies any returned data. -q - The SQL query, passed in as a string --convert - Convert the query to ICDB before executing, false by default

Sources

This project uses example databases generated for MySQL use.

Employees DB: https://github.com/datacharmer/test_db

Troubleshooting

If you get the error: The server time zone value 'MDT' is unrecognized or represents more than one time zone. run the query in MySQL: SET time_zone = 'America/Denver'; SET global time_zone = 'America/Denver'; or whatever timezone city you live nearby. This is a workaround for a VERY ANNOYING bug in MySQL.

Hyperparticle / integrity-coded-database