What prevents the adoption of differential privacy validation methods?

Differential privacy promises a simple solution to the thorny problem of privacy. It offers a relative definition of disclosure with many attractive properties but has proven to be deceptively complex to implement correctly. The difficulty in designing differentially private algorithms was identified early after the introduction of differential privacy, and methods to verify and validate algorithms have been developed which span a wide array of techniques including frameworks, type systems and novel programming languages, and statistical methods. Despite over a decade of research into the topic, these methods have not made their way into the mainstream and ad-hoc validation still appears to be the go-to method when implementing differentially private algorithms.

To understand the gap between research into validation methods and implementation. I plan to take a dual-sided approach which will be focused on:

- evaluating the solutions proposed in the research literature for potential issues that would affect their adoption into practice

- surveying the attitudes to validation methods in the settings where differential privacy is used in practice.


To evaluate differential privacy validation methods, I will develop a consistent methodology which can provide consistent criteria and results in a heterogeneous field. This methodology will be based on the experience gained from looking at one validation method, StatDP developed by Ding et al. It will need to incorporate which variants of differential privacy are supported, at what stage of the development process it applies--for example, is it a framework or programming language used during development, or a testing tool--the time required for it to produce results, and which programming languages are supported. Other criteria for evaluation will very likely be introduced as the methodology is developed and applied.

The other side of evaluation is surveying the knowledge and attitudes towards validation of those that implement differentially private systems. To this end, I intend to engage with the programming community by reaching out to the maintainers of open-source libraries and systems, and to companies and other organisations which use differentially private systems. The methodology for gathering information from the programmers and analysts that use differential privacy will likely require mixed-methods and will make use of questionnaires, surveys, and semi-structured interviews depending on the type of questions I will need to ask. The exact format of these methods is yet to be designed. Initially, more exploratory research will benefit from a free-form interview format, but once themes emerge, later research may require more structured methods. Similarly, I plan to engage with researchers who have authored papers in this field to understand what efforts they have undertaken to have their work be used in practical applications. Such research will require ethics approval, but I do not foresee any specific issues arising with the ethics process, as it will use standard methods.

The long-term plan is heavily dependent on the insights gleaned using the previously described work. It may be the case that more time is needed to examine the cause for the gap between the research and practice. Once a set of motivating factors have been identified, I will start looking into practical ways to ameliorate the situation and bridge the gap.

The EPSRC Centre for Doctoral Training in Cybersecurity will train over 55 experts in multi-disciplinary aspects of cybersecurity, from engineering to crime science and public policy.

Short term impacts are associated with the research outputs of the 55+ research projects that will be undertaken as part of the doctoral studies of CDT students. Each project will tackle an important cybersecurity problem, propose and evaluate solutions, interventions and policy options. Students will publish those in international peer-reviewed journals, but also disseminate those through blog posts and material geared towards decision makers and experts in adjacent fields. Through industry placements relating to their projects, all students will have the opportunity to implement and evaluate their ideas within real-world organizations, to achieve short term impact in solving cybersecurity problems.

In the longer term graduates of the CDT will assume leading positions within industry, goverment, law enforcement, the third sector and academia to increase the capacity of the UK in being a leader in cybersecurity. From those leadership positions they will assess options and formulate effective interventions to tackle cybercrime, secure the UK's infrastructure, establish norms of cooperation between industries and government to secure IT systems, and become leading researcher and scholars further increasing the UK's capacity in cybersecurity in the years to come. The last impact is likely to be significant give that currently many higher education training programs do not have capacity to provide cybersecurity training at undergraduate or graduate levels, particularly in non-technical fields.

The full details of our plan to achieve impact can be found in the "Pathways to Impact" document.