New technologies increasingly ease access to on-line services where one can post public information which reaches thousands of people. When doing so, rights and laws which are designed to protect us are oftentimes breached, as for example, when posting friends’ photographs without their consent or when making offensive comments.
Each time more often, youths venture into social networks, even when they are younger than the necessary minimum legal age of 14, and are not aware of the consequences of their acts, which can lead up to cases of cyberbullying. In order to correct this scenario, minors using the Internet must be educated in the responsible use of social networks as well as be made aware of the responsibilities and the consequences of their actions.
Tecnalia has joined efforts with Entelgy Ibai, an information technologies enterprise, with Seinale, a law firm specialised in information technology law, and Irungo LaSalle, a technical college, to develop a software tool -Gazela- that monitors the content shared by minors in Facebook.
Each user must sign into Gazela in order to be authorised to access content in their Facebook account and in order to obtain their biometric profile via a webcam. From then on, Gazela collects and analyses all the user’s published information in their Facebook account and searches for offensive texts and private information, such as telephone numbers or addresses. However, Gazela is not limited to monitoring the contents of texts, but also processes the photographs that are uploaded to Facebook, recognising other registered users by means of facial recognition algorithms.
This tool notifies the user of offensive content that has been published, of harassment it may be subject to and of images that have been published without authorization. It also provides a tool to request such authorisations and also notifies the user of the posting of images by other registered users of him/her.
Gazela has a “virtual attorney” that can counsel the harassed user/victim of steps/actions to take and also warns the harassing user of the consequences of one’s acts.
One of the most innovative modules is its biometric analysis. A facial recognition system has been chosen as it is one of the most widely used biometric tools and has greater user acceptance.
Our efforts have been focused on providing robustness and improving the analysed photographs by employing luminosity correction algorithms, which provides system reliability versus shine or glare. Furthermore, the problems associated with “the pose” have been dealt with the multiple compositions and the dynamic generation of models by grouping different poses, whereby in a discrete manner for the user, multiple facial models are obtained from him/her thus achieving improved results in recognition and facial location. This allows the Gazela system, which has images of groups of people in many different situations and in non-controlled locations, to adapt in a discrete way to the user, his/her facial models making them more robust to the passage of time or to new locations.
A pilot test has been carried out during the 2012-2013 academic year at LaSalle School in Irun to validate its functioning in a real environment and to test the user’s degree of acceptance.
Gazela is still being developed and is entering its third phase in 2013. Tecnalia’s Semantics and Artificial Intelligence Group continues to work on new models for the processing of natural language in order to detect harassment and shall extend its integration with Tuenti, Twitter and Youtube, while Tecnalia’s Computer Vision Group shall be responsible for the video analysis.
We are developing technologies of object and action recognition within the research line of multimedia content comprehension and analysis that will help identify bullying scenarios in images and videos and will detect non-authorised content (pornography, violence, etc.).