Deepfakes, shallowfakes & cheapfakes – Seeing is believing


Nasir Memon (computer science and engineering, Vice Dean for Academics and Student Affairs) is one of the co-founders of NYU Tandon's Center for Cybersecurity (CCS). He is also the founder of the OSIRIS Lab, CSAW, The Bridge to Tandon Program, and the Cyber Fellows program.

In this episode of MySecurity Media's podcast, Memon summarizes the early advancements of AI technology, and how a combination of “deep” neural networks an advancements in processing power led to the ever-growing area of modern "deepfakes". Putting deepfakes in historical context, he differentiates between what have come to be known as "cheapfakes" and "shallowfakes." 

He also explains how the online environment has changed, as deepfakes become more and more popular, including the newly-surfaced examples of various money-making and malicious schemes involving deepfakes, as well as their addition to the cyberattack arsenals of nation-state actors aiming to spread misinformation maligning the leaders of adversarial nations.

Memon explains that, ironically, the accurate detection of a deepfake can be as difficult, if not more difficult, than the creation of a convincing deepfake, and stresses the need to move past simply identifying them to preventing their generation through such means as embedding digital "watermarks" that prove a image's veracity. He also urges society as a whole to cultivate vigilance, healthy skepticism and discrimination, and not to "jump into conclusions ... look for corroborative evidence.”