New technologies challenge fake news and misinformation
Read news on the go. — © Digital Journal
Scientists seek to automatically differentiate original media content from fake, to improve online trust and find a way to detect fake content. This is an important consideration for those looking to improve what is shared on social media.
Thanks to advances in technology, those wishing to modify content can take audiovisual files and develop “deepfakes”, forming montages that look like real footage. These doctored images range from fun to the use of digital media that aim to cause harm.
Today, most fake content requires a human to detect and report it. On platforms like Facebook and Twitter, this is a huge task. Although some forms of machine learning can help, the technology has not been advanced enough to be completely reliable.
To help address these concerns, researchers are looking to combine forensic digital content analysis, watermarking and artificial intelligence techniques.
Although a work in progress, this development is part of an ongoing battle as photo and video editing and artificial intelligence tools become more sophisticated. The proposed solution uses artificial intelligence and data hiding techniques to help users automatically differentiate between original and adulterated media content.
The development is led by the Universitat Oberta de Catalunya (UOC). According to lead researcher Professor David Megías: “The project has two goals: first, to provide content creators with tools to watermark their creations, making any changes easily detectable; and second, to provide social media users with tools based on next-generation signal processing and machine learning methods to detect fake digital content.
As development progresses, there is unlikely to ever be a one-size-fits-all solution and detection will need to be performed with a combination of different tools. It is on this choice to explore the concealment of information (watermarks), forensic analysis techniques of digital content (based on signal processing) and machine learning.
Taking one of them as an example, watermarking uses several techniques in the field of data hiding that embed imperceptible information in the original file to be able to automatically verify a media file.
The new technology will take advantage of signal processing technology to detect the intrinsic distortions produced by the devices and programs used when creating or modifying any audiovisual file.
The technology to date is described in the ARES-Workshopswith a research paper titled “Architecture of a Fake News Detection System Combining Watermarking, Signal Processing, and Machine Learning”.