WIREDGOV

The following is an ICO discussion regarding new security risks associated with AI, whereby the personal data initially used to train the system might subsequently be revealed by the system itself.

This post is part of the ICO’s ongoing Call for Input on developing the ICO framework for auditing AI. We encourage you to share your views by leaving a comment below or by emailing us at AIAuditingFramework@ico.org.uk.

In addition to exacerbating known data security risks, as we have discussed in a previous blog, AI can also introduce new and unfamiliar ones.

For example, it is normally assumed that the personal data of the individuals whose data was used to train an AI system cannot be inferred by simply observing the predictions the system returns in response to new inputs. However, new types of privacy attacks on Machine Learning (ML) models suggest that this is sometimes possible.

In this update we will focus on two kinds of these privacy attacks – ‘model inversion’ and ‘membership inference’.  While the ICO’s overall Security guidelines already apply, as part of our AI auditing framework we are keen to hear your feedback about what you think would be reasonable approaches to threat modelling in relation to these attacks, and other best-in-class organisational and technical controls to address them.

The NCSC’s threat assessment aims to raise awareness of state-sponsored espionage targeting high-value research, as well as the risk of financial losses at the hands of cyber criminals.  While the NCSC has been working with the academic sector on an ongoing basis to improve security practices, this is the first threat assessment it has produced specifically for universities.

The assessment notes that while cyber criminals using methods such as phishing attacks and malware pose the most immediate, disruptive threat, the longer-term threat comes from nation states intent on stealing research for strategic gain. 

To mitigate the risks, universities are encouraged to adopt security-conscious policies and access controls, as well as to ensure potentially sensitive or high-value research is separated rather than stored in one area.

Why are attacks against universities successful?

In both culture and technology, universities are one of the most open and outward facing sectors. This enables and eases collaboration between academics across borders, and is likely a key component of their success. Unfortunately, this also eases the task of an attacker.

Measures to support universities have been outlined in Trusted Research, from the Centre for the Protection of National Infrastructure (CPNI) and the NCSC, which offers accessible and actionable cyber security advice for university leaders, staff & researchers.  The threat assessment for universities can be read here with a blog post also discussing the research in detail.

A 3-year collaborative programme has been launched to improve health & care across the Liverpool City Region.

The £1.9m Health Matters programme will help businesses take advantage of knowledge & expertise in the partnership to help develop solutions to challenges, receive guidance on funding opportunities and navigate the health ecosystem.  The programme will also provide real-world validation of products & services, showing evidence of impact and benefits.

The programme, which is part-funded by the European Regional Development Fund (ERDF), aims to catalyse businesses in the Liverpool City Region to become market leaders in health, wellbeing and care by encouraging the development of products, services and solutions.

Unless the Government steps up efforts to manage the transition to automation, entire regions of the UK face being left behind & British businesses could find themselves being uncompetitive, says BEIS Committee.

The Automation and the future of work report finds the UK’s slow place in moving to automation – the UK lags behind its G7 competitors in its adoption of robots – has allowed other countries to steal a march in leading the Fourth Industrial Revolution and seizing upon the opportunities for economic growth and jobs.

To ramp up the leadership and co-ordination needed to enable the UK to capitalise on these new technologies, the report urges the Government to come forward with a UK Robot and AI Strategy by the end of 2020. The report outlines a series of measures which could be introduced as part of this strategy, to help support businesses, industries, and universities and boost the adoption of automation.