Research Overview

Related Work

I strongly believe that high quality research is necessary to achieve high quality education and visa versa.  As a faculty at the Department of Computer Science and Engineering, my long-term goal is to establish a nationally recognized research and education program in information assurance (IA).  I am the founder and director of the Information Security Laboratory (ISL).  ISL activities include IA research, educational curriculum development, seminar organizations, and outreach activities increase IA awareness.  Since the year 2000, we have developed and offered 9 IA courses and a variety of IA topics as directed reading courses, and hired new faculty members with primary research interest in IA.  The enrollment in the IA courses during the last 5 years has increased from 1 course/year (27 students) to 4-5 courses/year (80-120 students).  We have also engaged more than 25 graduate students and several faculty members in IA research.

Since my employment at USC, I’ve led the effort of developing the IA curriculum and mapping our courses to the NSTISSI 4011, 4013, and 4014 standards.  I have developed and taught 4 different graduate level IA courses and numerous directed reading courses.  I’ve graduated 2 Ph.D. and 6 M.S. students.  Currently I advise 5, and co-advise 2 Ph.D. students.  Some of my research is in collaboration with other faculty members and graduate students.  Following is a brief overview of my main research areas and links to the Web sites containing detailed information.

My main research interests are information assurance and privacy. More specifically, I'm interested in the following areas: (1) the inference problem in database management systems, (2) authorization model for Semantic Web data, (3) privacy protection in self-organizing electronic communities, (4) legal and financial analysis of cyber incidents, and (5) information confidentiality in emerging Web applications.

 

Database Inference Problem

( http://www.cse.sc.edu/research/isl/dbInferPbm.shtml )

One of the main research areas in database security research is the development of authorization models to protect data confidentiality, integrity, and availability. While these models do protect sensitive data from direct data accesses, indirect secrecy violations via inference channels may still occur. The detection and removal of unauthorized inference channels are necessary to provide secure database systems.

In my Ph.D. dissertation, under the advisement of Drs. A. Brodsky and S. Jajodia, I developed an integrated security architecture that guarantees data confidentiality by extending a standard access control model with a Disclosure Inference Engine (DIE).   After each query request, DIE generates all the information that can be disclosed by the query requestor, using the requestor’s past and current queries (results) and the database constraints.  I introduced fundamental notions of data-dependent and data-independent disclosures, and showed that the problem of data-dependent data disclosure is decidable.  Our results in relational databases were extended to the inference problem in semi-structured databases and inferences via linear constraints in numeric databases.   

Currently I am co-advising the PhD student T. Toland with Dr C. Eastman.  His research addresses the effects of database updates on the inference problem.  We study database updates from two perspectives:

  1. Update of an attribute value that has been previously released to a user causes the released value incorrect, i.e., it is not in the database any longer.  Inference engines may use this incorrect value of the user’s past query results to infer disclosed data.  Such “incorrect” inferences may unnecessarily cause the denial of user queries.  Our work use a combination of logging database updates and stamping the users’ history files with the updated values if applicable.
  2. The second aspect of the updates is that observation of the changes in the database may release sensitive information.  This problem is similar to the statistical inference problem. 

Both aspects of inference detection require the usage of update logs and the maintenance of a history file for each user.  Some of

Future extensions of the current results include the development of models addressing collaborative attacker.  Currently each user is monitored independently and inferences are generated only on his/her history files.  Also, I’m planning to evaluate the inference problem from the perspective of privacy and fairness.  Disclosing seemingly unimportant or non-sensitive information may give advantage to an adversary.  In particular, it may play an important role in interactive negotiation and trust management. 

 

Secure Semantic Web

( http://www.cse.sc.edu/research/isl/SSW/index.shtml )

The focus of this project is to develop models and technologies for XML and RDF access control, and for prevention of security threats via illegal inferences in semantically enhanced semi-structured information.  XML and RDF data has been increasingly used for storing and exchanging information, and representing metadata.  Further, semantic annotation of Web data and the development of tools to interpret such annotations support the intelligent integration of large amounts of data.  This large-scale data integration may pose significant security and privacy threats by data aggregation, inference disclosure, and data mining.   Our research targets some of the above problems:

·        Access control models for XML.  For this, we address the problem of generating XML Views that satisfy the security policies, develop access control language for XML, develop method and a practical implementation to handle XML updates without violating document integrity or the security policy,  and to prevent indirect disclosure of XML data via ontology supported inferences.  Two Ph.D. students A. Stoica and V. Gowadia, and one M.S. students, D. Roy, are involved in this project.

·        We are currently developing access control model suitable for RDF.   Our aim is to support flexible data granularity, satisfy context-based and semantic requirements, and provide protection against undesired inferences.  Ph.D. student A. Jain is involved in this project.

·        Developing security framework for SMIL (Synchronized Multimedia Integration Language) formatted streaming data.  SMIL, an XML-like language, supports operational semantics. We provide language based security that respects continuity and synchronization constructs of SMIL.  This work is with Ph.D. student N. Kodali and Professor D. Wijesekera from George Mason University, Va.

Future extensions of the current results include the formal development of security models and completion of our prototype systems.  In addition, we are planning to study the security requirements of Web applications, like Web Services, and how these requirements impact the authorization models for XML, RDF, and OWL.  In particular, we are planning to address the need to formally define the intended meaning of XML documents and use this meaning (instead of the currently used syntactic constructs) to develop authorization model for XML.   

Finally, our future work also involves the study of offensive and defensive application of our methods, focusing on the security issues created by large-scale and focused data integration.  A defensive approach aims to develop models and tools to protect against (or warn about) unauthorized data disclosure; an offensive approach addresses information gathering techniques and techniques to release erroneous data in a stealthy manner that leading an adversary to a desired, wrong conclusion.

 

Legal and Financial Analysis

( http://www.cse.sc.edu/research/isl/SSW/themis.shtml )

Cyber attacks represent serious financial and legal burden.  Current inadequacies in national law and ambiguous interpretations of international treaties make it difficult to prosecute cyber attackers and/or provide an acceptable self-defense for legitimate counter attackers. With collaborative support from USC Ph.D. students R. McCraw and S. Saxena, GMU Ph.D. student Liesheng Peng, and Drs. J. B. Michael, Naval Postgraduate School, D. Wijesekera, George Mason University, and T. Wingfield, the Potomac Institute for Policy Studies, we develop legal and economic models that evaluate the effects of cyber attacks.  Currently we are working on a system that evaluates the direct and cascading effects of cyber attacks, and use this evaluation to perform reasoning about the severity of the attack and “lawful” response strategies.  We also studying the different economic models used by cyber insurance companies to estimate the insurance premium and the level of compensation.

 

Secure, Self-Organizing Communities

(http://www.cse.sc.edu/research/isl/anonimSys.shtml )

In this project we propose a new approach to provide accountability in self-organizing Web communities, while guaranteeing high level of privacy.  We present a framework for electronic communities that support dynamic grouping and collaborations.  The system is controlled by competition among communities.  The security protocols we developed for the system build upon community-based trust and limits exposure of personal information on a trusted third party.  We propose a two-layered privacy protection architecture, that allows enforcement of internal- (web community) and external (real world) accountability.  Enforcement of external accountability requires the release of mappings between real users and their virtual identities, enforcement of internal accountability requires the release of mappings among the virtual users.  This work is in collaboration with Ph.D. students G. Ziegler and Dr. A. Lorincz at Eotvos Lorand University, Budapest. 

Our research to improve the efficiency of Web crawlers is closely related to our anonymity project.   A fleet of crawlers is observed as a self-organizing community with needs for sharing and security.  Our current work focuses on the improvement of learning algorithms used by Web crawlers.  Our future work addresses the security protocols needed for these communities.  This work is in collaboration with Ph.D. students Zs. Palotai and Dr. A. Lorincz at Eotvos Lorand University, Budapest. 

 

My Other Research Areas Include

 

In my future research I’m planning to study the following research areas.  A brief overview of our current activities, results, and publications for each area can be reached by following the links.  

 

·         Network Security

·        Secure telephone conferencing

·        Honeynet data analysis

·        Agent-based Intrusion Detection

·         Access Control Models

·        Automated support for security policy integration

·        Access control for GIS applications