Abstract: Recently, generative adversarial networks (GANs) have achieved stunning realism, fooling even human observers. Indeed, the popular tongue-in-cheek website http://thispersondoesnotexist.com, taunts users with GAN generated images that seem too real to believe. On the other hand, GANs do leak information about their training data, as evidenced by membership attacks recently demonstrated in the literature. In this work, we challenge the assumption that GAN faces really are novel creations, by constructing a successful membership attack of a new kind. Unlike previous works, our attack can accurately discern samples sharing the same identity as training samples without being the same samples. We demonstrate the interest of our attack across several popular face datasets and GAN training procedures. Notably, we show that even in the presence of significant dataset diversity, an over represented person can pose a privacy concern.
Entries Tagged "face recognition"
Page 1 of 4
Fascinating research: “Generating Master Faces for Dictionary Attacks with a Network-Assisted Latent Space Evolution.”
Abstract: A master face is a face image that passes face-based identity-authentication for a large portion of the population. These faces can be used to impersonate, with a high probability of success, any user, without having access to any user-information. We optimize these faces, by using an evolutionary algorithm in the latent embedding space of the StyleGAN face generator. Multiple evolutionary strategies are compared, and we propose a novel approach that employs a neural network in order to direct the search in the direction of promising samples, without adding fitness evaluations. The results we present demonstrate that it is possible to obtain a high coverage of the population (over 40%) with less than 10 master faces, for three leading deep face recognition systems.
Fawkes is a system for manipulating digital images so that they aren’t recognized by facial recognition systems.
At a high level, Fawkes takes your personal images, and makes tiny, pixel-level changes to them that are invisible to the human eye, in a process we call image cloaking. You can then use these “cloaked” photos as you normally would, sharing them on social media, sending them to friends, printing them or displaying them on digital devices, the same way you would any other photo. The difference, however, is that if and when someone tries to use these photos to build a facial recognition model, “cloaked” images will teach the model an highly distorted version of what makes you look like you. The cloak effect is not easily detectable, and will not cause errors in model training. However, when someone tries to identify you using an unaltered image of you (e.g. a photo taken in public), and tries to identify you, they will fail.
EDITED TO ADD (8/3): Kashmir Hill checks it out, and it’s got problems.
OneZero is tracking thirty countries around the world who are implementing surveillance programs in the wake of COVID-19:
The most common form of surveillance implemented to battle the pandemic is the use of smartphone location data, which can track population-level movement down to enforcing individual quarantines. Some governments are making apps that offer coronavirus health information, while also sharing location information with authorities for a period of time. For instance, in early March, the Iranian government released an app that it pitched as a self-diagnostic tool. While the tool’s efficacy was likely low, given reports of asymptomatic carriers of the virus, the app saved location data of millions of Iranians, according to a Vice report.
One of the most alarming measures being implemented is in Argentina, where those who are caught breaking quarantine are being forced to download an app that tracks their location. In Hong Kong, those arriving in the airport are given electronic tracking bracelets that must be synced to their home location through their smartphone’s GPS signal.
The Chinese facial recognition company Hanwang claims it can recognize people wearing masks:
The company now says its masked facial recognition program has reached 95 percent accuracy in lab tests, and even claims that it is more accurate in real life, where its cameras take multiple photos of a person if the first attempt to identify them fails.
Counter-intuitively, training facial recognition algorithms to recognize masked faces involves throwing data away. A team at the University of Bradford published a study last year showing they could train a facial recognition program to accurately recognize half-faces by deleting parts of the photos they used to train the software.
When a facial recognition program tries to recognize a person, it takes a photo of the person to be identified, and reduces it down to a bundle, or vector, of numbers that describes the relative positions of features on the face.
Hanwang’s system works for masked faces by trying to guess what all the faces in its existing database of photographs would look like if they were masked.
Communities across the United States are starting to ban facial recognition technologies. In May of last year, San Francisco banned facial recognition; the neighboring city of Oakland soon followed, as did Somerville and Brookline in Massachusetts (a statewide ban may follow). In December, San Diego suspended a facial recognition program in advance of a new statewide law, which declared it illegal, coming into effect. Forty major music festivals pledged not to use the technology, and activists are calling for a nationwide ban. Many Democratic presidential candidates support at least a partial ban on the technology.
These efforts are well-intentioned, but facial recognition bans are the wrong way to fight against modern surveillance. Focusing on one particular identification method misconstrues the nature of the surveillance society we’re in the process of building. Ubiquitous mass surveillance is increasingly the norm. In countries like China, a surveillance infrastructure is being built by the government for social control. In countries like the United States, it’s being built by corporations in order to influence our buying behavior, and is incidentally used by the government.
In all cases, modern mass surveillance has three broad components: identification, correlation and discrimination. Let’s take them in turn.
Facial recognition is a technology that can be used to identify people without their knowledge or consent. It relies on the prevalence of cameras, which are becoming both more powerful and smaller, and machine learning technologies that can match the output of these cameras with images from a database of existing photos.
But that’s just one identification technology among many. People can be identified at a distance by their heartbeat or by their gait, using a laser-based system. Cameras are so good that they can read fingerprints and iris patterns from meters away. And even without any of these technologies, we can always be identified because our smartphones broadcast unique numbers called MAC addresses. Other things identify us as well: our phone numbers, our credit card numbers, the license plates on our cars. China, for example, uses multiple identification technologies to support its surveillance state.
Once we are identified, the data about who we are and what we are doing can be correlated with other data collected at other times. This might be movement data, which can be used to “follow” us as we move throughout our day. It can be purchasing data, Internet browsing data, or data about who we talk to via email or text. It might be data about our income, ethnicity, lifestyle, profession and interests. There is an entire industry of data brokers who make a living analyzing and augmenting data about who we are —using surveillance data collected by all sorts of companies and then sold without our knowledge or consent.
There is a huge —and almost entirely unregulated —data broker industry in the United States that trades on our information. This is how large Internet companies like Google and Facebook make their money. It’s not just that they know who we are, it’s that they correlate what they know about us to create profiles about who we are and what our interests are. This is why many companies buy license plate data from states. It’s also why companies like Google are buying health records, and part of the reason Google bought the company Fitbit, along with all of its data.
The whole purpose of this process is for companies— and governments —to treat individuals differently. We are shown different ads on the Internet and receive different offers for credit cards. Smart billboards display different advertisements based on who we are. In the future, we might be treated differently when we walk into a store, just as we currently are when we visit websites.
The point is that it doesn’t matter which technology is used to identify people. That there currently is no comprehensive database of heartbeats or gaits doesn’t make the technologies that gather them any less effective. And most of the time, it doesn’t matter if identification isn’t tied to a real name. What’s important is that we can be consistently identified over time. We might be completely anonymous in a system that uses unique cookies to track us as we browse the Internet, but the same process of correlation and discrimination still occurs. It’s the same with faces; we can be tracked as we move around a store or shopping mall, even if that tracking isn’t tied to a specific name. And that anonymity is fragile: If we ever order something online with a credit card, or purchase something with a credit card in a store, then suddenly our real names are attached to what was anonymous tracking information.
Regulating this system means addressing all three steps of the process. A ban on facial recognition won’t make any difference if, in response, surveillance systems switch to identifying people by smartphone MAC addresses. The problem is that we are being identified without our knowledge or consent, and society needs rules about when that is permissible.
Similarly, we need rules about how our data can be combined with other data, and then bought and sold without our knowledge or consent. The data broker industry is almost entirely unregulated; there’s only one law —passed in Vermont in 2018 —that requires data brokers to register and explain in broad terms what kind of data they collect. The large Internet surveillance companies like Facebook and Google collect dossiers on us are more detailed than those of any police state of the previous century. Reasonable laws would prevent the worst of their abuses.
Finally, we need better rules about when and how it is permissible for companies to discriminate. Discrimination based on protected characteristics like race and gender is already illegal, but those rules are ineffectual against the current technologies of surveillance and control. When people can be identified and their data correlated at a speed and scale previously unseen, we need new rules.
Today, facial recognition technologies are receiving the brunt of the tech backlash, but focusing on them misses the point. We need to have a serious conversation about all the technologies of identification, correlation and discrimination, and decide how much we as a society want to be spied on by governments and corporations—and what sorts of influence we want them to have over our lives.
This essay previously appeared in the New York Times.
EDITED TO ADD: Rereading this post-publication, I see that it comes off as overly critical of those who are doing activism in this space. Writing the piece, I wasn’t thinking about political tactics. I was thinking about the technologies that support surveillance capitalism, and law enforcement’s usage of that corporate platform. Of course it makes sense to focus on face recognition in the short term. It’s something that’s easy to explain, viscerally creepy, and obviously actionable. It also makes sense to focus specifically on law enforcement’s use of the technology; there are clear civil and constitutional rights issues. The fact that law enforcement is so deeply involved in the technology’s marketing feels wrong. And the technology is currently being deployed in Hong Kong against political protesters. It’s why the issue has momentum, and why we’ve gotten the small wins we’ve had. (The EU is considering a five-year ban on face recognition technologies.) Those wins build momentum, which lead to more wins. I should have been kinder to those in the trenches.
If you want to help, sign the petition from Public Voice calling on a moratorium on facial recognition technology for mass surveillance. Or write to your US congressperson and demand similar action. There’s more information from EFF and EPIC.
EDITED TO ADD (3/16): This essay has been translated into Spanish.
The New York Times has a long story about Clearview AI, a small company that scrapes identified photos of people from pretty much everywhere, and then uses unstated magical AI technology to identify people in other photos.
His tiny company, Clearview AI, devised a groundbreaking facial recognition app. You take a picture of a person, upload it and get to see public photos of that person, along with links to where those photos appeared. The system—whose backbone is a database of more than three billion images that Clearview claims to have scraped from Facebook, YouTube, Venmo and millions of other websites—goes far beyond anything ever constructed by the United States government or Silicon Valley giants.
Federal and state law enforcement officers said that while they had only limited knowledge of how Clearview works and who is behind it, they had used its app to help solve shoplifting, identity theft, credit card fraud, murder and child sexual exploitation cases.
But without public scrutiny, more than 600 law enforcement agencies have started using Clearview in the past year, according to the company, which declined to provide a list. The computer code underlying its app, analyzed by The New York Times, includes programming language to pair it with augmented-reality glasses; users would potentially be able to identify every person they saw. The tool could identify activists at a protest or an attractive stranger on the subway, revealing not just their names but where they lived, what they did and whom they knew.
And it’s not just law enforcement: Clearview has also licensed the app to at least a handful of companies for security purposes.
EDITED TO ADD (1/23): Twitter told the company to stop scraping its photos.
This essay discusses the futility of opting out of surveillance, and suggests data obfuscation as an alternative.
We can apply obfuscation in our own lives by using practices and technologies that make use of it, including:
- The secure browser Tor, which (among other anti-surveillance technologies) muddles our Internet activity with that of other Tor users, concealing our trail in that of many others.
- The browser plugins TrackMeNot and AdNauseam, which explore obfuscation techniques by issuing many fake search requests and loading and clicking every ad, respectively.
- The browser extension Go Rando, which randomly chooses your emotional “reactions” on Facebook, interfering with their emotional profiling and analysis.
- Playful experiments like Adam Harvey’s “HyperFace” project, finding patterns on textiles that fool facial recognition systems not by hiding your face, but by creating the illusion of many faces.
I am generally skeptical about obfuscation tools. I think of this basically as a signal-to-noise problem, and that adding random noise doesn’t do much to obfuscate the signal. But against broad systems of financially motivated corporate surveillance, it might be enough.
Public Voice Launches Petition for an International Moratorium on Using Facial Recognition for Mass Surveillance
You can sign on as an individual or an organization. I did. You should as well. No, I don’t think that countries will magically adopt this moratorium. But it’s important for us all to register our dissent.
MIT Technology Review is reporting about an infrared laser device that can identify people by their unique cardiac signature at a distance:
A new device, developed for the Pentagon after US Special Forces requested it, can identify people without seeing their face: instead it detects their unique cardiac signature with an infrared laser. While it works at 200 meters (219 yards), longer distances could be possible with a better laser. “I don’t want to say you could do it from space,” says Steward Remaly, of the Pentagon’s Combatting Terrorism Technical Support Office, “but longer ranges should be possible.”
Contact infrared sensors are often used to automatically record a patient’s pulse. They work by detecting the changes in reflection of infrared light caused by blood flow. By contrast, the new device, called Jetson, uses a technique known as laser vibrometry to detect the surface movement caused by the heartbeat. This works though typical clothing like a shirt and a jacket (though not thicker clothing such as a winter coat).
Remaly’s team then developed algorithms capable of extracting a cardiac signature from the laser signals. He claims that Jetson can achieve over 95% accuracy under good conditions, and this might be further improved. In practice, it’s likely that Jetson would be used alongside facial recognition or other identification methods.
Wenyao Xu of the State University of New York at Buffalo has also developed a remote cardiac sensor, although it works only up to 20 meters away and uses radar. He believes the cardiac approach is far more robust than facial recognition. “Compared with face, cardiac biometrics are more stable and can reach more than 98% accuracy,” he says.
I have my usual questions about false positives vs false negatives, how stable the biometric is over time, and whether it works better or worse against particular sub-populations. But interesting nonetheless.
Sidebar photo of Bruce Schneier by Joe MacInnis.