Digital Economy

Arming oneself against fake news: tools for prebunking and debunking

A guide to tools for understanding sources and possibly verifying articles, photos and videos.

16 June 2024

6' min read

More and more people are falling victim to fake news, i.e. false or misleading information content that can spread rapidly, thanks also to technological advances that facilitate the creation of realistic and increasingly sophisticated deepfakes. A dangerous spiral that leads to reinforcing prejudice and misinformation, identified by the World Economic Forum's Global Risks Report 2024 as the 'greatest short-term global danger to democracies'. A phenomenon that raises concerns about the reliability of online information and the impact that deliberately altered news can have in shaping public opinion. A scenario made even more complex by distrust of traditional media, which leads many to turn to alternative sources such as social networks and messaging apps, fertile ground for the proliferation of fake news. The picture is exacerbated by the interference of certain states, as in the case of the falsehoods made in China on the events in Taiwan, up to the Russian interference of disinformation propagated in the past on various geopolitical scenarios, including the USA. The stakes are soaring in the run-up to the EU and US elections.

The importance of prebunking

In this context, it becomes crucial to promote the practice of 'prebunking', i.e. providing tools to recognise false news in advance, rather than relying only on traditional a posteriori 'debunking', the prerogative of journalists and specialised sites. A need highlighted by Walter Quattrociocchi, a researcher at the Data Science Centre of Sapienza University and author of a study in Nature that investigated the toxic nature of interactions in digital communities.

"Studies show that we tend to give credence to sources that confirm our prejudices," explains Quattrociocchi. "Social media have ridden a business model based more on entertainment than information, fragmenting opinions into self-referential 'echo chambers'. A perverse dynamic whereby 'disproving sources does not work, because people remain anchored to their mental schemes even when faced with the evidence of facts'. The only effective solution, according to the researcher, is 'to provide the tools to understand the mechanisms of disinformation and to become aware of the cognitive biases that influence the reception of news'. This is the only way to really stem the spread of fake news.

The big-tech counteroffensive: Google sharpens its weapons

The search engine Google has created a portal, also in Italian, bringing together all important information on disinformation, misinformation and manipulation, in cooperation with Moonshot and other local partners. This initiative is part of a pre-bunking campaign in the European Union, aimed at countering some of the most widespread tactics to manipulate people online. The page features videos explaining scapegoating and decontextualisation techniques, i.e. the use of unrelated media to support a claim. There are also links to institutional sources and also to Elections24Check, which makes it easy to find and have at your fingertips verified voting information and data from over 40 European fact-checking organisations in 37 countries across the continent. Added to these are the features introduced by Google in April, prebunking tools that help users assess the reliability of online news sources, encouraging them to develop a personal database of credible sources to rely on over time.

In the first case, when doing a search, one can click on the three vertical dots that appear next to each result. A side pop-up opens, giving additional information on the source, the date of indexation, links to social profiles, and a series of useful reviews from the web and sites such as Trustpilot. In addition, results 'from searches related to the names of the source' are shown, providing the information needed to learn more about the site from which the information is being obtained.

For images, the noteworthy new feature is the 'About this image' option. Again by clicking on the three dots, the option provides access to useful details such as the history of the image or how news and fact-checking sites describe it. Although launched last year, the function is now available in 40 more languages worldwide, including Italian.

On the fact-checking front, Google has introduced the Fact Check Explorer tool, a list of checks that have already been made on the subject or character, listed with the most recent at the top, which is continuously updated and shows the title of the hoax, the related keywords, and the outcome of the check. Or you can have an image verified as having been used within a verified article by fact-checking organisations by simply uploading the URL of the image itself. A tool you should always have at hand when surfing the web, to quickly find out if what you are reading is true.

Other Debunking Tools for Images

There is no shortage of other tools to tell whether an image or video is created by artificial intelligence. A popular tool used by professional debunkers is InVID. This is a series of software solutions developed as part of an EU-funded research project and available from www.invid-project.eu. The main one of these solutions is an extension for Firefox and Chrome that allows you to quickly obtain contextual information on Facebook, Twitter and YouTube videos, search for images in Google, Yandex, Bing, Tineye, Baidu or Karmadecay (for Reddit), extract and process frames from Facebook, Instagram, YouTube, Twitter and Dailymotion videos, read the metadata of videos and images, check copyrights and apply forensic analysis filters to images to reveal manipulations. A comprehensive tool available free of charge.

SCREENSHOT3

We tested InVID by having it analyse the photograph created by photojournalist Barbara Zanon, whose work has been aimed at raising awareness of Ai issues generating truthful photos. In this case, the photo of the elderly man with flowers in his hand was created with version 5 of Midjourney, which is already known for generating fascinating and disturbing images such as the photo of Pope Francis wearing a white duvet or the shots of Donald Trump's fake arrest. The InVID tool gave me several elements to understand that it was not a real image. Firstly, by showing the Google search linked to the photo, the artificial nature of the image emerged thanks to in-depth links. But not only that: the image in question lacked metadata, i.e. all the information that is recorded once the photo has been taken, such as the type of camera, date of capture or location. Information that is altered if the image has been retouched.

TinEye

Another popular tool is TinEye, a free reverse image search platform that allows you to quickly locate all pages where an image or a similar version of it has been published. Unlike other reverse search tools, TinEye allows you to set filters for searches, thus offering more diverse and relevant results. Using the 'Most digitally manipulated' filter, the software shows all possible versions of the photo found online, allowing the user to understand how many times and in what ways the image has been manipulated. With the 'Older' filter, the software sorts the results in chronological order, making it easier to quickly find the original image.

The apps: Fake Image Detector and Veracity

Fake Image Detector is a website and app for Android devices that allows you to check whether an image in your device's gallery has been digitally manipulated. Once the image is uploaded, the app provides an indication of possible authenticity using two analysis methods: error level analysis (ELA) and metadata analysis. ELA is based on the premise that a compromised or altered image has greater digital compression in the altered areas than in the rest of the image. By analysing the composition and compression of the photo, the app can indicate whether the image has been manipulated. However, the programme can generate false positives for images from Telegram, WhatsApp or other messaging apps that digitally compress files. In these cases, the app still judges images received by message as 'digitally manipulated'.

Veracity, on the other hand, is an application only available for iOS devices and allows the user to perform a reverse search of an image, finding out on which other websites it has been used. It also has a function to check the authenticity of profile photos used on social networks to detect fake profiles.

Google Images

The simplest and most widely used method is undoubtedly the Google image search. The way it works is simple: one can type a keyword on the home page of the service or use the reverse search by uploading an image to find similar images or trace the source through the 'find image source' button in the sidebar. If the search produces no results or if the image appears in a news item that is not confirmed by any other source, this could potentially be misinformation.