Performance analysis of filtering software using Signal Detection Theory

Software filters are increasingly being touted as a solution to restrict access to inappropriate information in a variety of settings. Families want to protect young children from pornographic sites, corporations are searching for ways to minimize trivial use of the Internet by their employees, and non-profit organizations look to control information access to reflect the value system of their communities. Despite the exponential increase in software filter usage, its effectiveness is not clear. In addition, critics of the approach argue that mandated use of filtering software on public computers like libraries may result in denial of vital information to poorer sections of the society who do not have independent access to the Internet, thereby curbing intellectual freedom and creating inequity in access to information. The purpose of this study is to analytically evaluate the performance of software filters using the Signal Detection Theory (SDT) framework. Two types of software filters are modeled and analyzed—simple software filter (single method) and a sequence of software filters (multiple methods). Analysis shows the limited capability of both types of filters, with the multiple-method filter outperforming the simple filter. Results of this study caution proponent of filter-based solution to be realistic with expectations of the benefits of filtering based solutions. Implications of the findings for proper use and design of software filters are discussed.