15 March 2016 14:08

Computer scientists at Lancaster University are working on smarter ‘privacy setting’ algorithms to prevent embarrassing photos being shared on social media.

Currently, when updates and photographs are posted on sites such as Facebook only the person that posted is able to set the privacy settings – even when there are other people in the image, mentioned in a comment, or invited to an event.

“This is a massive and serious problem as users’ individual privacy preferences for co-owned items, such as photos, usually conflict, so applying the preferences of only one party risks these items being shared with undesired recipients,” says Dr Jose Such, Lecturer at Lancaster University’s School of Computing and Communications. “This can lead to privacy violations, which in some cases can have severe consequences such as people losing their jobs or being cyberstalked.”

In a paper, ‘Resolving multi-party privacy conflicts in social media’, published by the journal IEEE Transactions on Knowledge and Data Engineering, researchers have shown it is possible to automatically detect when conflicts over privacy settings will occur when social media posts are made.

This research also demonstrates that algorithms can be used to drastically reduce the number of conflicts that have to be resolved manually. When a clash of privacy settings happens it is currently down to the individuals concerned to communicate with each other – via messaging, email or phone calls - to resolve the issue.

Dr Such, lead researcher on the paper, said: “Computational mechanisms that can automate the negotiation process have been identified as one of the biggest gaps in privacy management in social media. We want to minimise the burden on the user to resolve multi-party privacy conflicts.”

The paper proposes an automated mediator, which would examine all of the individual’s privacy settings and identify any conflicts. It then estimates, using privacy preferences, how sensitive an item is to individuals and proposes solutions. If all users accept the solution it is applied.

The algorithm was tested on a group of 50 people, and the solutions it proposed were shown to clearly outperform the current method followed by mainstream social media such as Facebook - which is that the person uploading the item decides who can see it.

Dr Such said: “This is the first mechanism for detecting and resolving privacy conflicts in social media that is able to adapt the conflict resolution strategy based on the particular situation. It has the potential to reduce the amount of manual work needed to achieve a satisfactory solution for all parties involved in social media privacy conflicts.”

The other researcher who worked on the paper is Dr Natalia Criado from King’s College London.

The paper represents a stepping stone for future research, and it has already influenced research now being conducted as part of another follow-up project called RePriCo, which is lead by Dr Such and funded by EPSRC. RePriCo is looking at understanding better multi-party privacy conflicts and the factors that contribute to acceptable solutions to develop truly “multi-party” privacy tools.  

More information about Lancaster University’s School of Computing and Communications is available by visiting http://www.lancaster.ac.uk/scc/