The New Issue of Analytics in Government Quarterly Addresses Data Security Issues
In this Analytics in Government Quarterly (AGQ), we explore issues related to data security. Arnold Toporowski discusses the use of data that include personal information and anonymization approaches that might help relieve privacy issues. Hubert Laferrière examines cybersecurity and the potential of using AI in the battle against intrusions into an organization’s network. He points out that research on AI attacks and other anomalies is in its infancy. Kevin Kells explores the importance of human oversight of machine learning algorithms and Betty Ann Turpin discusses a data governance framework that can adjust to continually changing contexts. I contribute an article that examines issues of data poisoning of AI algorithms and discusses the emerging concept of certified external data sets and a “blockchain for AI” idea that secures both the data and the algorithms.
There is no doubt that, as more government organizations adopt AI and Machine Learning approaches, there will be more attempts at intrusions. In the past, these intrusions might be seen as little more than a nuisance that cost time and effort to recover our data or to protect our systems. With AI-enabled decision making however, intrusions could lead to errors in classification of individuals (consider for example, applying for licenses of various types, loans, etc.). Moreover, as we move towards an autonomous vehicle and personalized medicine future, AI algorithms that mistake a stop sign for a speed limit sign or a malignant tumour as benign could lead to dire consequences.
As always, we welcome your comments and suggestions on the articles in this issue. Please do contact us here: firstname.lastname@example.org