Professors and a graduate student from the College of Engineering and Information Technology at the University of Arkansas at Little Rock (UALR) have developed a new model to manage the "vast ocean" of user-generated content being generated by the ever-growing social network sites including Facebook and Twitter.
Nitin Agarwal, assistant professor in EIT's Department of Information Science, and doctoral student M. Venkata Swamy in that department, worked with Srini Ramaswamy, former chair of the UALR computer science department and now vice president of research at ABB India, to develop a Context-Based Privacy Model. The model leverages intelligent, scalable, adaptive, and robust pattern-matching algorithms to allow Internet sites to automatically adjust privacy needs of consumers or organizations to the context in which the data is accessed.
The research was supported in part by grants from the U.S. Office of Naval Research and the U.S. National Science Foundation.
Their paper on the project was presented and awarded "Best Paper" at the Second International Symposium on Privacy and Security Applications held in conjunction with the Institute of Electrical and Electronic Engineering (IEEE) International Conference on Privacy, Security, Risk, and Trust Aug. 20-22 in Minneapolis, Minn. Only 13 percent of papers submitted at the highly competitive conference are presented.
"With the advent of social media websites such as Facebook, Myspace, and Twitter, and social health websites such as Patientslikeme that help people with health conditions connect with people with like conditions, a vast ocean of user-generated content has been created—including non-sensitive information as well as sensitive demographic, financial or health-related data," Agarwal says. "As a result, users may be unknowingly granting access to their data, leading to grave privacy concerns."
In recent years, companies' data information centers in industries are facing increasing federal regulations due to these privacy concerns, forcing them to constantly modify their privacy information-handling policies. The existing research on developing privacy models, although seem persuasive, are essentially based on user, role or service identification. Such models are incapable of automatically adjusting privacy needs of consumers or organizations to the context in which the data is accessed.
"In this work, we propose a Context Based Privacy Model (CBPM), which leverages the automatic context identification of the information consumer borrowing concepts from Object Oriented methodology," according to the researchers. A context could be defined as a secure or non-secure location, family members, or group of friends, etc.
"Considering numerous pieces of information such as name, telephone number, e-mail address, age, gender, items purchased online, social interactions each individual generates; and the number of contexts created, the CBPM matrix could quickly become huge and unmanageable."
The UALR team addresses that problem by leveraging intelligent, scalable, adaptive, and robust pattern-matching algorithms to compress the matrix, making it more manageable.
"Our work has shown the necessity of avant-garde privacy models dealing with the challenges of new types of information sources, creating a vast ocean of data with intricate access requirements and constraints, forcing us to think beyond the existing user, role, or service-based privacy models," Agarwal says. "The proposed work is unique, one of its kind emphasizing on the context more importantly than the content, with far-reaching implications in the privacy as well as the information security area."