Read on Twitter
Deepfakes present very specific opportunities for malignant activities on social media. This latest exposé of deepfake-enabled dirty tricks, astutely written by @Razhael, highlights several points of concern. https://www.reuters.com/article/us-cyber-deepfake-activist/deepfake-used-to-attack-activist-couple-shows-new-disinformation-frontier-idUSKCN24G15E
As some of you may well remember, @KeirGiles, Kim Hartmann and I authored a report on malign influence campaign, featuring the now infamous but still very non-existing #KatieJones, who had targeted key people working in security via @LinkedIn. https://twitter.com/STRATCOMCOE/status/1197088378203856897
Like #KatieJones, #OliverTaylor was created as part of an elaborate deception campaign. Oliver appears designed to help establish, shape and take hold of narratives. @Razhael goes into deeper detail about this in his thread. https://twitter.com/razhael/status/1283357080146018307
I am mentioned briefly in this accompanying piece which further describes how to identify deepfakes. There is no denying that GANs are getting more and more sophisticated. But my talking point was not about identifying them. https://graphics.reuters.com/CYBER-DEEPFAKE/ACTIVIST/nmovajgnxpa/index.html
Which brings me to the focus of this thread: not about how to identify deepfakes, but to recognise how deepfakes are being used. Think of them as a blank canvas. I mentioned in the article that they carry “no history or baggage".
This is important.
This is important.
By "blank canvas", I meant that you have a GAN photo with unlimited possibilities - a photo of a non-existent person whose legend can be constructed from scratch with no constraint whatsoever.
You as the creator of the picture have the power and freedom to conjure or mould their backstory for whatever nefarious purpose or context you want.
For instance, in this case, the creator designed him to be a political commentator, so they padded the profile with think pieces and a Quora profile to add some depth to his personality.
As well as unique images being innately suspicious, OSINT investigations can easily identify that this is not a real person if their background is not properly constructed. But @razhael’s story shows once again that nobody even looks... until they were alerted.
And this is the real danger, and a key point from our @stratcomcoe report. No matter how good the technical tools, there is no substitute for informed awareness and alertness by users. https://techhq.com/2019/11/deepfakes-are-coming-for-your-linkedin-feed/
However, excitement about deepfakes shouldn't blind us to the fact that simple forgeries (“cheapfakes”) are still easy, and can be even more effective. https://nakedsecurity.sophos.com/2020/01/08/facebook-bans-deepfakes-but-not-cheapfakes-or-shallowfakes/
In concluding this thread, deepfakes facilitate very specific forms of deception or fraud. Malign influence of other forms can be more pernicious and even harder to detect.
Furthermore, deepfakes are a single-shot weapon. Malign influence, by contrast, is a process not an event.
Furthermore, deepfakes are a single-shot weapon. Malign influence, by contrast, is a process not an event.
