CAN THE MOSSAD AND ORDINARY CITIZENS BEAT FACIAL RECOGNITION?

03 - October 2018

www.jpost.com

Mossad director Yossi Cohen made a rare public speech on Monday and admitted that facial recognition technology is challenging the Mossad as it trickles out and around the world even to “non-hi-tech” countries.

Cohen explained that spying is getting harder because the same technologies that catch terrorists can sometimes uncover foreign intelligence operations.

“Everywhere we go, we have to take into account the fact that security services are getting stronger,” he said.

How does the Mossad overcome these new challenges? While current and former Mossad agents refrained from commenting, a former IDF intelligence agent interviewed with the Jerusalem Post on Tuesday about new techniques. Also, in the past former Mossad officials have spoken to the Post about the issue.

There are two main ways to beat facial recognition: one is through cutting-edge technology to defend against these programs and the other is to alter one’s appearance.

Gil Perry, a former IDF intelligence official, is the founder and CEO of D-ID, a company that has developed anti-facial recognition technology. The technology was not engineered for the Mossad, but rather for business people who want to share their appearance in marketing material without it being scooped up later by identity thieves and used against them in other unintended contexts.

Perry said that his technology is already being used by Cloudinary, a cloud technology service with 350,000 customers.

He added that one of the largest Japanese conglomerates will soon be using it, and that he is in advanced talks with top players in finance, health, government and camera manufacturing as well.

Perry and two of his partners, Eliran Kuta and Sella Blondheim – who are also former IDF intelligence or special forces officials – became interested and started working on the issue six years ago when their work in intelligence prevented them from sharing photos online.

D-ID was incorporated two years ago as deep-learning and facial-recognition programs brought a massive expansion of very accurately-deployed systems. He added that while the general public has started to understand that they are under surveillance or likely to be photographed by other people’s smartphones in many public areas, they do not realize how broad the issue is.

The former intelligence official said that a large volume of companies, banks and online platforms are also tapping people’s photos in all sorts of ways and using facial recognition on customers, unbeknownst to them.

Perry said that D-ID’s technology “protects organizations, enterprises, governments and databases of photos and videos” so that photos being shared or stored for marketing purposes “look the same to the human eye, but not to artificial intelligence and facial-recognition algorithms.”

How does the technology work? It uses advanced image processing, deep-machine learning, morphological transformation and generative adversarial network attacks to slightly alter the photo in a way that throws off facial recognition programs, without being noticed by the human eye. It creates a kind of “noise” that “confuses” such programs. The technology is designed to be adaptable in a way that facial recognition programs will repeatedly be fooled and cannot make a small adjustment to catch up, Perry said.

The bottom line is facial recognition programs “cannot recognize the subject” in a photo, but humans for whom the photos are intended in a marketing campaign cannot tell that the photos have been slightly altered. D-ID’s technology can be used to block “identity theft, fraud, reuse of biometric data” and “to comply with the GDPR” – the EU’s new privacy protection law.

“Today, everything is hacked. With Cambridge Analytica, Google Plus and others, everything is breached – people are becoming aware that being a private citizen is also important for business,” Perry said.

While Perry’s technology is utilized to protect the misuses of marketing materials, it is not a far stretch to think of a large number of ways that similar technologies could be used by the Mossad to beat facial recognition in a variety of contexts.

Other tools which are much less sophisticated and impractical for businesses – but possibly highly useful in other contexts for Mossad agents – are eyeglasses which use infrared light or flashes to fool facial recognition, but are undetectable to the human eye.

A similar adaptation of infrared technology can be hidden under a baseball hat and possibly under a wig, an umbrella or hair. One application of the technology projects dots of light onto the wearer’s face in a way that not only obscures their identity, but could even facilitate impersonating someone else.

Hacking a particular facial recognition guard station and combinations of pixelation of the face are other hi-tech tactics the Mossad might use. Perry dislikes the latter, since in business it distorts the person’s face in a notable way that can ruin the marketing purpose.

Then there are the old-school tactics. Some privacy activists are able to beat facial recognition with low-tech makeup or paint placed in pinpoint areas of contrast on a human face – where a nose is located or where the chin becomes the neck – that have been used by the Mossad and other intelligence agencies for a while.

In the past, former Mossad officials have mentioned using hats along with fake mustaches, fake beards and other lowtech non-permanent facial adjustments to avoid being identified, although this can still work against facial recognition technologies in certain contexts.

With plenty of new developments and advances in modern technology, the facial recognition tech-race is off at full speed.