Uniqueness and How it impacts privacy

The purpose of this research were to evaluate the characteristics of two social science datasets of high volume. Furthermore, the researchers want to understand how having unique features within the data collection process affects privacy. This was achieved by using two data sets provided by The Kinsey Institute. They performed a simple score re-identification experiment to see if they could re-identify an individual. They were successful in achieving this with great accuracy however when error was added these findings changed. This research tells me that even though someone may try to limit an attackers ability to re-identify an individual it was still possible to do so a high percentage of the time. This would indicate to me that there must be other safeguards on data such as encryption or another medium in protecting user data.

k-Anonymity: A Model for Protecting Privacy

In this paper the researchers explored whether using k-anonymity was enough to help de-identify patients in a hospital, while still keeping useful information to the researchers. Various algorithms were used to try and limit what information can be found by using the data set beyond the information contained in the data. They found that you can come close to controlling the disclosure of information by examining the relationships between attributes. While some attributes are traditional identifiers others are known as quasi-identifiers which include information that can be public or private.

Relational Learning For Sustainable Health

The researchers in this paper were attempting to use machine learning to predict whether an individual would have some health related issue. Three case studies were presented to demonstrate their ability to various issues an individual can have. In the first case study they demonstrate how machine learning can predict whether an individual will have a chronic heart problem between 0 and 20 years. Various risk factors were used as data for the machines to make their predictions such as: age, sex, cholesterol, etc. The experiment revealed that they are able to predict with great accuracy whether a person will have a chronic heart condition. The second case study discusses how machine learning can be used to predict the likelihood of someone with mild cognitive impariment moving to Alzheimer’s disease. This was done by performing MRIs. The results of the MRI are used as input to create a graph and the machine will examine both of these to come to a conclusion. The third case study tried to use machine learning to determine the success or failure of different drugs on patients. This work is very interesting because it demonstrates three different ways in which machine learning can help solve problems that are very taxing on today’s society.

Common Pitfalls in Writing about Security and Privacy…

In this paper the authors provided advice to authors to when conducting experiments on human subjects to make less common mistakes during data collection and documentation of experiments. The method in which this was done was by means of creating a guide to researchers and authors discussing how to avoid common mistakes. A few of the guidelines were to clarify and state the hypothesis, avoid misleading readers, explain what was observed in detail, and one interesting guideline is to ask for help when needed. This paper would help in any research because at some point one would need to document that research so having this guide is helpful in writing an effective paper.

Forward and Backward Citation

Forward:

• McCabe, Janice M., et al. “Methodological considerations from a Kinsey Institute mixed methods pilot project.” International Journal of Multiple Research Approaches 7.2 (2013): 178-188.
• Manfredi, Veronica. “Privacy Implications of New York City’s Stop-and-Frisk Data.” (2015).
• Hill, Raquel, et al. “An empirical analysis of the utility of a differentially private social science dataset.” Workshop for privacy enhancing tools. 2013.

Backward:

• B. Malin and L. Sweeney. Re-identification of DNA through an automated linkage process. In Proceedings of the American Medical Informatics Association Fall Symposium, pages 423–427, 2001.
• A. C. Solomon, R. Hill, and E. Janssen. Poster: Privacy and de-identification in high-dimensional social science datasets, 2011. Presented at IEEE Security and Privacy 2011.
• Y. Xiao, L. Xiong, and C. Yuan. Differentially private data release through multidimensional partitioning. In W. Jonker and M. Petkovic, editors, Secure Data Management, volume 6358 of Lecture Notes in Computer Science, pages 150–168. Springer Berlin / Heidelberg, 2010.

Wednesday was very interesting. I enjoyed the mobile app development preparation and am excited to create an app with my team. I also enjoyed the laser cutting and 3-D printing. The laser cutter was surprisingly fast which was very cool! Overall my experience was positive for today and look forward to what is to come.