The rise of generative AI is expected to revolutionize the ways we use our security systems along with the information that these systems will be able to provide. These promises of new levels of efficiency, accuracy and depth of data from gen-AI are capturing most of the attention and discussions around this technology and its influences on the security market.

However, there are other impacts from generative AI that can pose threats to the integrity of the industry’s core set of technologies. Perhaps the most pressing concern is the rise of deepfake or altered videos, made possible by new gen-AI video creation and editing tools that are becoming widely available. 

These tools can pose significant challenges to the integrity of video surveillance footage and organizational trust in video evidence, thanks to deepfakes and other manipulations of video that could be seamless to the viewer. Some examples of specific changes to video that could be commonplace include the fabrication of a different timestamp to provide incorrect information about the date and time of an event, cutting out specific frames of video from an event to remove the event of interest in the video, or altering a face in a scene with that of another person or placing a firearm in an individual’s hand using generative AI. 

All of this can create significant doubt in the integrity of video as one of the most crucial pieces of evidence used in criminal investigations, court proceedings, and internal investigations within an organization. In many countries, there is a very robust chain of custody process required during investigations and admitting the video as evidence in court. The use of video as a credible source could be rendered useless if we’re unable to trust that the video itself is not an accurate view of the scene.

In the future, traditional forensic techniques will not be able to adequately safeguard video from the widely available gen-AI tools that are quickly growing more sophisticated. This highlights the need for industry collaboration and standardization as a way to preserve the integrity of video and the institutional trust in the footage as an accurate view of a scene.

ONVIF is working on a video method authentication that provides proof that the video has not been edited since it has left the camera sensor. Securing the video at the earliest point, at the point of video capture from the sensor in the camera, is key to ensuring the authenticity and trustworthiness of the video footage from camera to court.

On a technical level, a signature (or unique digital fingerprint) is created for each video frame, and the frames are then packaged and “signed” using a digital certificate that is unique to that individual camera. This digital key, or signature, allows a supporting viewer (video management client, video player, etc.) to verify that the video data originated directly from the specific camera and has not been tampered with. Specifically, ONVIF defines how this certificate is included in the video stream to be later verified by a compatible software client. There are other existing ways of protecting video but most of these are initiated at the point of the VMS and not at the actual source of the video.

Standardizing video authentication with ONVIF enables a single mechanism for video management systems to verify the authenticity of the video it has received. This can help to streamline processes for video users such as law enforcement and other criminal justice personnel, who deal with video footage generated by systems from many different manufacturers that might use different methods for protecting video. Maintaining this chain of custody for video evidence is also just as important in a corporate setting, particularly when it comes to internal threats, such as those with access to an organization’s network or other systems that may be used to store video and other evidence.

Security is only one industry that relies on the authenticity of video footage and the ramifications of generative AI on digital video. ONVIF is planning to release the implementation of video authentication as an open source project. This will provide a wide community of developers access to these video authenticity specifications to improve the implementation based on changing trends and conditions. Making these standards available via open source will also create transparency in the implementation, preserving trust in the authentication process and the integrity of the video itself.