Reporters Without Borders (RSF) introduces this series on Open-Source Intelligence (OSINT), an ensemble of techniques intended to help gather information from publicly available resources. The first article explains how journalists can use OSINT to analyse and extract data from online images.
Open-Source Intelligence (OSINT) techniques are used to collect, analyse, and utilise publicly available information to support fact-finding and investigation, without having to rely on internal or classified data. This article will focus on extracting information from images available online using metadata, which are information embedded into digital image files that record their history.
Metadata can reveal a lot about an image
- Time and date the image was originally taken;
- Time and date the image was last edited;
- GPS location of where the image was taken;
- Name of the owner of devices which stored or edited the image;
- Original filename of the image;
- Model of camera used;
- Camera settings such as focal length, lens type, aperture information, and flash setting.
The quantity of metadata available varies based on the image
JPG, one of the most common file formats for digital photography, usually contains a wide range of metadata, while other formats such as BMP or PNG may include less. Some software like Photoshop will alter the original image’s metadata when exported, while some service providers such as Facebook will strip most of them when the image is uploaded to the site.
Extracting metadata is simple with the right tools
Metadata attached to digital photographs is called Exif (Exchangeable image file format) data. The simplest way to view Exif data is with an online Exif reader such as Exifinfo where journalists can directly upload, or link the URL of, an image.
Online Exif readers are a good tool for analysing online, publicly available images, but they are likely to store uploaded images on a third-party server. An alternative journalists can use to check the information of sensitive photos is Exiftool, an open-source application which can analyse metadata offline on a private device.
Metadata can be used to detect generative AI and photo manipulation
Exif data analysis can be used to validate the authenticity of images. Metadata such as camera settings will likely be absent in generative AI images. Some AI image generators, such as DALL-E, Midjourney and Generative Fill will even embed their digital signature in the metadata. Online AI image detectors such as Is It AI can also be used to assist detection of AI-generated images.
Journalists can check online photographs for possible manipulation through techniques like error-layer analysis or noise analysis. These strategies require some technical knowledge, but open-source tools like FotoForensics or Forensically have made it easy for less experienced users to spot signs of alteration and filters, and read metadata.
Left: Sample image. Right: Image analysed by Forensically using error-layer filter. Showing inconsistent lighting elements.
Safety recommendations for journalists regarding image metadata
Some OSINT tools online, such as Exif readers, are likely to store uploaded images on a third-party server that may not necessarily be secure. Therefore, if journalists want to extract information from sensitive photos provided by sources or from their own private folders, they should favour offline desktop apps such as Exiftool, which also supports a wide range of file and metadata formats.
Furthermore, adversaries may apply OSINT techniques on online images to inflict harm against journalists or human rights defenders, for instance by finding their personal information within Exif data in order to dox them. Journalists can protect their sources and themselves by altering metadata of images before making them public. ExifCleaner is a safe desktop app which removes metadata from image, video and PDF files.
