06.06.2014 Views

Research Statement

Research Statement

Research Statement

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Research</strong> <strong>Statement</strong><br />

Jeff Pittges<br />

Motivation<br />

External attacks on computer systems are well publicized and well recognized as a<br />

serious problem. However, insiders perpetrate over 80 percent of all security breaches.<br />

Until recently corporations rarely reported data leakage (i.e., the loss of confidential<br />

information). Emerging regulations, including the Health Insurance Portability and<br />

Accountability Act (HIPAA), require corporations to protect sensitive information and<br />

publicly disclose security breaches. Analysts estimate that data leakage costs<br />

corporations over a billion dollars a year. Consequently, data security, data leakage, and<br />

privacy are becoming urgent challenges for the database community [1].<br />

Most database security architectures rely on a fortress security model in which access to<br />

the database management system (DBMS) is strictly controlled while the data within the<br />

database is stored in the clear. Not only is the data unprotected from attackers who<br />

circumvent the access controls, the data is also unprotected from authorized personnel<br />

thereby creating unnecessary threats to privacy and security.<br />

Solution<br />

One solution to this problem is to apply fine-grained role-based access privileges, but this<br />

approach still leaves the data exposed. Most popular database management systems<br />

allow database administrators to encrypt data thereby applying the principle of Defensein-Depth<br />

[10]. Data encryption adds a layer of protection but it is still possible to recover<br />

the original data by decrypting the encrypted data [4, 5, 6].<br />

A Translucent Database [7, 8] uses one-way hash functions and other cryptographic<br />

algorithms to obfuscate the original data without destroying the function of the data. For<br />

example, the Unix password file applies a one-way hash function to each password<br />

before the password is stored in the file. Each time a user logs into the system the hash<br />

function is applied to the password entered by the user and the result is compared with<br />

the user password stored in the password file. If an attacker gains access to the contents<br />

of the password file it is impossible for the attacker to obtain the original data thereby<br />

preventing access to the user passwords.<br />

Translucent data is secure and translucent data renders attacks uninteresting because the<br />

data has no value to the attacker. Translucent data also protects privacy because<br />

authorized users are unable to view the original data. The Microsoft Hailstorm project<br />

[9], an online store of personal information including an individual’s calendar, contacts,<br />

and credit card numbers, was built on the concept of a translucent database.


The techniques to create a translucent database are readily available to any application<br />

developer. However, implementing translucency in the application layer creates an<br />

undesirable dependency between the application and the database. There are also<br />

database performance considerations. Finally, application developers are not given time<br />

to implement security features and when time is allocated for security, few developers<br />

have the necessary skills to implement security features correctly. The goal of my<br />

research is to extend database management systems to provide native support for<br />

translucency thereby increasing data security and privacy and reducing data leakage.<br />

Plan<br />

The Oracle Database 10g provides a built-in package named DBMS_CRYPTO that<br />

provides functions for generating keys, managing keys, and encrypting and decrypting<br />

data [10]. However, Oracle’s DBMS_CRYPTO package uses a block cipher rather than<br />

applying encryption to specific data columns as discussed in [4]. A block cipher requires<br />

the DBMS to encrypt and decrypt data as the data moves between main memory and<br />

disk. Therefore, the data in main memory is exposed as clear text and the query<br />

execution engine does not operate on encrypted data. Hence, Oracle 10g does not<br />

provide native support for data encryption. I plan to build on the research presented<br />

above by: (1) developing native DBMS support for cryptography, (2) extending the<br />

cryptographic functions supported by the DBMS to include one-way hash algorithms, and<br />

(3) developing a framework that makes security far more accessible to practitioners. To<br />

my knowledge, no similar work exists in the published literature. Furthermore, I am<br />

optimistic the recent NSF CyberTrust and CyberInfrastructure programs will continue to<br />

support such security-related research.<br />

The first step in my research plan is to extend the Data Definition Language (DDL) and<br />

Data Manipulation Language (DML) to support policy-based security. Consider a<br />

security policy stating that all HIPAA data must be encrypted. The database<br />

administrator (DBA) should be able to define a HIPAA data type and when a database<br />

table is defined with a column of type HIPAA the DBMS should automatically encrypt<br />

the data. The DDL statements used to create the table should allow the DBA to select the<br />

cryptographic algorithm from a set of algorithms natively supported by the DBMS. The<br />

first step of my research will create a mechanism for implementing policy-based security<br />

within the DBMS that is easily accessible to practitioners.<br />

The theoretical constructs developed in the first step of my research will lack significant<br />

impact unless they can be implemented and tested in realistic scenarios. My 15 years of<br />

database experience are rife with examples of promising ideas that have been shelved due<br />

to the absence of rigorous testing academic research requires. Drawing on this<br />

experience, the second step of my research plan is to implement a system that supports<br />

the constructs developed in the first step. This system will serve as a tool to evaluate the<br />

results of the first step and guide further research. This system may be developed by<br />

extending an existing open source database such as MySQL. This work should provide<br />

interesting projects for students and challenging programming assignments for computer<br />

science classes.


While the first and second steps of my research will demonstrate the ability to store data<br />

securely, the full extend of my research goals may only be realized if the DBMS is able<br />

to operate on translucent data without recovering the original data thereby enhancing the<br />

security of the data and protecting privacy. Therefore, the third step of my research plan<br />

will develop functions that operate directly on translucent data. These functions include<br />

searching, partial matching, and data mining operations that are efficient while preserving<br />

data security and privacy.<br />

References<br />

1. Database Security (Common Sense Principles).<br />

http://www.governmentsecurity.org/articles/DatabaseSecurityCommonsensePrinciples.php<br />

2. Database Security Issues: Inference.<br />

3. http://databases.about.com/od/security/l/aainference.htm<br />

4. Mattsson, Ulf T. A Database Encryption Solution that is Protecting Against External<br />

and Internal Threats, Meeting Regulatory Requirements. February 25, 2005.<br />

http://database.ittoolbox.com/white-papers/a-database-encryption-solution-that-is-<br />

protecting-against-external-and-internal-threats-meeting-regulatory-requirements-<br />

4266<br />

5. Newman, Aaron C. Encryption of Data at Rest. May 28, 2002.<br />

http://database.ittoolbox.com/white-papers/encryption-of-data-at-rest-1922<br />

6. MacVittie, Don. Time is Right for Database Encryption. December 9, 2003.<br />

http://www.networkcomputing.com/showitem.jhtml;jsessionid=ZWGY32WNMZH0<br />

EQSNDBCCKHY?articleID=16401578<br />

7. Wayner, Peter. Translucent Databases. Baltimore: Flyzone Press. 2002.<br />

8. Garfinkel, Simon. Protecting Privacy with Translucent Databases. August 2, 2002.<br />

http://www.oreillynet.com/pub/a/network/2002/08/02/simson.html<br />

9. Microsoft Hailstorm. http://xml.coverpages.org/hailstorm.html<br />

10. Nanda, Arup. Encrypt Your Data Assets. Oracle Magazine. January/February 2005.<br />

http://www.oracle.com/technology/oramag/oracle/05-jan/o15security.html

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!