On the weekend, the media reported on a real time facial recognition and facial tracking system to be deployed for 24 hours surveillance in Georgetown and the East Bank of Demerara. It was described by a government official as, “…one of the best developments for our country…”.
This facial recognition technology has been banned in San Francisco. Silicon Valley is where companies have quickly capitalized on the science behind facial recognition, i.e. deep learning, to develop voice recognition and eye disease detection software. When the government in the region, which survives on taxing the revenues of its companies, is banning facial recognition systems, that sends a strong signal that there are serious concerns of misuse.
Microsoft, the largest technology company in the world, has refused to sell its facial recognition technology to police in California. A Microsoft executive has stated, “…if we move too fast with facial recognition, we may find that people’s fundamental rights are being broken.” When a leading technology company in deep learning technology forgoes profits because of concerns on human rights, then one should question why the Government of Guyana knows better?
There are 2 key ingredients to build a facial recognition system:
(1) A large set of photos, usually in the millions, and;
(2) Massive amounts of computing resources using specialized computer chips
For (1), it would require a significant amount of time and effort to procure enough pictures of Guyanese to build a large ‘training’ set for the system. An alternative is to use image databases that already exist. This tactic however has flaws, it has been shown in the US that facial recognition software incorrectly recognizes dark-skinned females. This is because these databases are usually built from images of people online which generally skews their bias to certain features. Will the wrong folks be arrested by the police?
It is unlikely that the Guyanese facial recognition system was developed in Guyana. It was most likely developed offshore by Huawei. This is because to train the system for facial recognition you need massive amounts of computing power using specialized chips. This raises the question of whether the government sent data of its citizens to an offshore state and if parts of the system exist offshore. The likely answer might be yes, here are the reasons:
There are ways to fine tune the facial recognition system to better recognize Guyanese given inherent biases in these systems. One of the most popular techniques is known as ‘transfer’ learning. However, given it is unlikely Guyana has the computing capability, this would mean that images of Guyanese would have to be shipped offshore to tune the system. And since the system will likely need regular tuning, this would mean a continuous stream of images of Guyanese being sent offshore without their consent.
One popular technique used by US law enforcement is to use driver’s licence pictures to identify images in videos. If the Guyanese system has this capability, then was the driver’s licence information of Guyanese sent offshore to test and tune the system?
If personal information, images and videos are being sent offshore that raises serious concerns about the privacy of Guyanese. Are we ceding control to a foreign power? The media lists a few features of the facial recognition system, but more sophisticated analysis could be performed by a foreign government. Example, they could blackmail our politicians. The images of Guyanese politicians are in the media every day, these images could be used to instrument (program) a system to automatically track politicians and their interactions.
The images and information from one of these facial recognition systems in China was recently exposed to the world because of a lack of passwords. The exposed information allowed anyone to track where people went and for how long. How are these facial recognition systems in Guyana being secured against threats from the Internet?
Over the past year, the Canadian government has been involved in a dispute with China. This was triggered by Canada following the rule of law and arresting a Huawei executive on behalf of America. But this has resulted in repercussions against Canada. Canadians have been detained in China, for unknown reasons, and kept under 24-hour artificial lighting. Canadian exports to China have been banned by the Chinese government over questionable reasons which has affected key farming regions in Canada. If we have a dispute with China, would the facial recognition system be used against us with far more dire consequences?
There are a number of Chinese laws which mandate that organizations must comply with intelligence work if asked. One has to recall that a Chinese company has a stake in the Stabroek block. Could Huawei be requested to counter any calls for oil contract renegotiation? Thus, if we have folks in the streets wearing t-shirts calling for contract renegotiation could China easily recognize and target these folks.
We have to ask if the facial recognition system is a Trojan horse? This would render it one of the worst developments for the citizens of Guyana.
Darshanand Khusial,
Master of Computer Engineering
References:
https://www.theverge.com/2019/4/17/18411757/microsoft-facial-recognition-sales-refused-police-access
https://www.cnet.com/news/why-facial-recognitions-racial-bias-problem-is-so-hard-to-crack/