Despite the controversy surrounding the Poland-based facial recognition software PimEyes, a comprehensive test of the search engine shows that it has trouble identifying normal people.
Of the 25+ searches performed by DailyMail.com, the AI system had various problems with 70 percent of the photos, including rotated or slightly blurred images.
Journalists and celebrities seemed pretty accurate, but only 25 percent of the results were completely accurate for the average person.
For this reason, however, security experts consider PimEyes to be a “serious security risk” – the website provides information for social media accounts.
Some of the matches included URLs to the person’s Instagram, TikTok, Tumblr, and Facebook, as well as personal blogs.
Those looking to “chase” someone with PimEyes may find their target but have to search through a ton of pornographic images.
Approximately 15 percent of the search results were displayed with explicit images pointing to the original adult content website.
Scroll down for video
Despite the controversy surrounding the Poland-based facial recognition software PimEyes, a comprehensive test of the search engine shows that it has trouble identifying normal people. Most of the search results were wrong, with the exception of one from Facebook more than 10 years ago.
Other outcomes varied in success – one return gave a correct picture of Facebook more than a decade ago, but the rest of the returns were inaccurate.
Only 26 percent of the returns were completely accurate for no apparent reason, as most of the subjects had social media profiles that were either sparse or private
The website, which uses more than 900 million images to find people, enables people to perform reverse image searches and find images on the internet that they believe are the same person.
A generic site name on which the previous image was found is offered for free search. However, it should be noted that it has privacy features, including protection from “scammers, identity thieves, or those who use your picture illegally”.
Only 26 percent of the returns were completely accurate for no apparent reason, as most of the subjects had social media profiles that were either sparse or private. Journalists (pictured) and celebrities seemed pretty accurate
A search for DailyMail.com journalist Jen Smith tended to pull up accurate images of Daily Mail links, all of which point to DailyMail content
A paid subscription that can cost as little as $ 23.99 per month includes additional details such as: B. the link through which the picture was found.
Despite the lackluster results of DailyMail.com’s recent testing, Digital Warfare lead analyst James Knight said the service was “very open to abuse”.
“Although it is marketed for individuals to search for their own picture, anyone can search for your picture,” Knight said via email on Friday. “The main fear is that this will be used maliciously by stalkers to get more information about you.”
Knight continued, “A quick photo of someone could be uploaded to potentially reveal hundreds of photos of them and, from there, their name, address, phone numbers, and email addresses. This tool is another stage in the privacy erosion. ‘
Although it offers a free service that uses artificial intelligence and machine learning algorithms, the company’s paid offering is described in more detail and includes the link that the photo was taken from.
Also, users can set up alerts that a matching picture they have uploaded will be added to the internet.
Despite the controversy surrounding the Poland-based facial recognition software PimEyes, a recent test of the search engine shows that it has trouble identifying normal people. The original photo is uploaded on the left and PimEyes results are on the right
Of the more than 20 searches performed by DailyMail.com, it struggled with 12 even if images were rotated on the page or were slightly fuzzy. Some search results even showed up with pornographic images. The original photo is uploaded on the left and PimEyes results are on the right
A paid subscription that can cost as little as $ 23.99 per month includes additional details such as: B. the link through which the image was found – bottom left in the result images. The original photo is uploaded on the left and PimEyes results are on the right
With a Premium account, users – individuals and businesses – can pay $ 29.99 per month for one month of access and 25 searches per day. A more expensive plan at $ 299.99 per month allows unlimited searches and month-long access.
The company also gives a 20 percent discount when the plans are billed annually.
In order for photos to be removed from the website, a person can request that they be removed using an online form.
In order to really get removed from the search engine, PimEyes is offering a “PROtect service” for US $ 79.99 per month, which gives users “4 hours per month of professional services during which our agent tries to remove your photos from the Remove source sites “.
London-based technology researcher Stephanie Hare told the Washington Post that it is possible for strangers to keep tabs on other people’s lives.
ORIGINAL: Shown is another image that DailyMail uses to test how inaccurate PimEyes was in finding normal people on the web
RESULTS: Although the results do not show the man in the original, PimEyes was able to create images of people wearing masks
‘What is stopping you? Literally nothing, ”Hare told the news agency.
“The people who put these images on the Internet – with their children, their parents, the people who might be vulnerable in their lives – weren’t doing this to feed a database that businesses could monetize,” she continued. “I can leave my phone at home. What I can’t leave is my face. ‘
DailyMail.com has contacted PimEyes with a request for comment.
PimEyes was used by people on 4chan and other sites who paid subscriptions to “Stalk Women” and created threads by them, citing an interview with Aaron DeVera, a security researcher in New York, according to The Post.
PimEyes has also come under fire in the past for potentially violating the European Union’s General Data Protection Regulation (GDPR) on facial recognition.
Daily Mail reporters tested the website and found that it had limited and spirited success for a photo of a member of the general public. In the picture two pictures that have been uploaded. The selfie on the left didn’t match. The image to the right of Facebook created matches
FAIL: Daily Mail tested the website to see how effective it was in finding non-famous people. When a selfie was inserted into the system, no correct matches were found (image)
MATCH: If a Facebook profile picture and not a new selfie was used, more pictures were found on the website that are available online (figure, results). It should be noted, however, that two of the results included the exact same picture as the profile photo (top left).
However, according to the company, PimEyes was designed as a “multipurpose tool” that people can use to find their face online, reclaim image rights, and monitor their online presence.
That is, when a person enters a picture of a stranger on the website and sets up a warning on their face, the system will tell the customer when they can find that stranger in a picture posted on Twitter, for example.
With PimEyes Premium, users can set up to 25 notifications. These can be used for 25 photos of a single person to increase the chances of success, or for photos of 25 people.
Dave Gershgorn recently reported for OneZero that the website also offers law enforcement contracts.
When MailOnline health editor Stephen Mathews’ Twitter ad photo (pictured) fed into PimEyes, the right matches were found
He reports that the facial recognition software can scan both darknet websites and the surface web, and says that PimEyes is built into the app from at least one company called Paliscope.
Paliscope is aimed at law enforcement agencies so they can use facial recognition on images and files to identify suspects in a case.
Paliscope recently partnered with the 4theOne Foundation, an organization dedicated to finding trafficked children.
PimEyes claims to enable people to protect their privacy and prevent their images from being misused. However, there are no safeguards in place to ensure that a person uploads their own face.
MATCH: Pictured is PimEyes’ first page when MailOnline’s health editor Stephen Matthews used his Twitter profile picture. It correctly matched his face to more than ten results, however six of the results contained the exact same photo (picture)
MATCH: In the picture, more results for Stephen Matthews (in the picture). These matches are the right person, but they are variations of two individual photos – smiling after receiving a journalist award and a front-on picture used in another MailOnline article. No surprising or hidden images were found
Nothing prevents people from uploading a photo of a stranger taken without permission and scouring the internet for more details about their life.
The ability to find more images and information about a person in a single photo poses dangers to all types of people.
PimEyes isn’t the first website to bring facial recognition technology to the public and search for a person’s face online.
In 2020, the Russian search engine Yandex was accused of providing an unregulated facial recognition system and invading privacy.
Yandex, which claims to perform more than 50 percent of Russian searches on Android, allows users to enter a picture and see results from the same person.
Another Russian company called NtechLab started FindFace in 2016. This worked similarly to PimEyes and was extremely popular.
It was eventually shut down and the technology adapted for government surveillance. The company’s founders at the time said they developed the technology to help women find someone who was up to date.
The photo of Piers Morgan holding up a sausage roll provided various results of the journalist and television personality from elsewhere on the internet (picture)
Last year, Comparitech’s security specialist Brian Higgins told MailOnline that PimEyes had “very serious and obvious privacy implications”.
“I seriously doubt the developers are naive enough to believe it will only be used for its intended purpose,” he added.
‘Unfortunately, the Internet is not subject to comprehensive protection of privacy. Each website or platform asks users to accept their terms and conditions.
‘These pages are almost always many pages long and contain references to “picture ownership”, which often gives the provider rights to store and use pictures.
‘FindFace’s terms and conditions were a prime example of this, but global adoption was quick and substantial.
‘Unfortunately, nobody reads the terms and conditions, so platforms like PimEyes can use new and developing technology to deliver potentially unethical products and services.
“The only solution to the problem is for individual Internet users to take their privacy seriously and take steps to protect themselves.”
As recently as June 2020, MailOnline reporters tested the website and found that it had only limited and spirited success for a photo of a member of the public.
However, when used by celebrities, the website was surprisingly far more effective.
Images by Amanda Holden, Piers Morgan and Boris Johnson provided hundreds of correct matches and demonstrated the effectiveness of the technology for a face that is regularly in the public domain.
It seems to be more effective on high resolution pictures that are already online than on new selfies that have not been shared before.
In the picture is the picture of the MailOnline health reporter Connor Boyd, which was entered into the PimEyes system
This feature sets it apart from popular search engines like Google, which don’t use face recognition when searching for images.
Instead, Google tech looks for features similar to clothing and surroundings.
For example, if a white man with brown hair wearing a blue tie against a white background uses Google image search to take a selfie, it will provide images that match the description. No data from a person’s face is used.
However, PimEyes pulled up pictures of me wearing different items of clothing in different lighting and environments, which suggests that face recognition is actually being used.
Stephen Matthews, Health Editor at MailOnline, used his Twitter profile picture and found the website to be moderately successful.
MailOnline health reporter Connor Boyd also tested the PimEyes technology. After entering a clear picture, the software could only correctly identify one more picture of Boyd. That picture – a five-year-old vacation photo (top left) – came as a surprise as it’s a photo that gets very little attention online or on social media
It correctly matched his face on more than ten results, however six of the results were exactly the same photo on different domains.
The other results were variations of two individual photos – smiling after receiving a journalism award and a front-on image that was used in another MailOnline article. No surprising or hidden images were found.
MailOnline health reporter Connor Boyd also tested the PimEyes technology.
After entering a clear picture, the software could only correctly identify one more picture of Boyd.
This image – a vacation photo from 2015 – was surprising as it’s a photo that is little known online or on social media.
However, more readily available images of Boyd were not identified by the software, although his face appeared in previous articles on MailOnline, one of the most widely read websites in the world. All other results were wrong.
For celebrities and well-known people like Tom Brady, PimEye provides accurate results
WHAT CAN USERS DO TO PROTECT THEMSELVES FROM FACIAL DETECTION ERRORS ONLINE?
Felix Rosbach, product manager at the German software development company comforte AG, says there is only so much that an individual can do to protect himself from this technology.
He told MailOnline: “As a private person, you can only protect your data by making sure that your social media profiles are not publicly available and by only disclosing data to trustworthy parties.”
But Mr. Rosbach adds that unfortunately there is very little users can do when their peers post pictures of you publicly.
He calls the search engines himself to make sure the public is protected.
He said: ‘Search engines should make sure that these features cannot be misused.
However, with the wide availability of machine learning software, there will always be a page or app that can offer this service.
“Instead, companies should and can ensure that sensitive consumer data is protected at all times.
“And it’s not just about securing access to data, but also about strong data protection to ensure that data becomes unusable in the event of data loss, misconfiguration or data breach.”