This will be the most important blog post you, a job seeker, will read this week. Possibly this month. Let's review Generative Adversarial Network (GAN) generated faces and fake personas in the recruitment scams space: How to identify them, how to confirm they aren't real, and how to avoid the traps they set. (The long and short of it is: Ignore them. Block them.)
The inciting incident for this post: I received an email "noticing" I hadn't uploaded my resume for a job with an agency I never applied to:
The sender's address, Jennifer Young, at first appears legitimate. Umail is a general University Mail service. University of Utah uses it. UC Santa Barbara too. Red flag 🚩: Why is a recruiter for a job not yet defined using a university domain?
🚩🚩 Furthermore... why isn't that email domain formatted like a legitimate University email like this random faculty member I found at the University of Utah.
Even if the address Jennifer Young used is a routing email for workers safety... email@umail.jobcase.com 🚩🚩🚩 They're preying on you through the surface level legitimacy granted by a UMail account.
Next I look for her on LinkedIn. There is a Jennifer Young without a photo. However, she looks inactive with three followers. A long abandoned profile.
This email from Jennifer Young isn't even the the red flag that raised my eyebrow. But starting here by checking their email itself is the element you should check & verify first!
The element that caught my eye first is Jennifer Young herself. It's specifically her face. This face does not exist.
Literally. See for your self by making your own GAN faces here: https://thispersondoesnotexist.com/ This website is perfectly harmless.
Refresh the page to see a new face.
These generated faces are downloadable by anyone and useable for anything- but often they're used for scams. Like, fake reviews on Yelp, Amazon, or Etsy or disinformation campaigns ran through twitter to inflate the popularity of unpopular policies through fake engagement.
🚩 Jennifer Young's headshot demonstrates all the telltale signs of artificially created unmodified GAN-generated faces. 🚩 One telltale trait is that the main facial features (especially the eyes) are in the same position on every image. Seeing one isn't enough for the novice to pick up on. Here's an infographic from a botnet identified on Twitter in 2020 to help deconstruct what you're looking at. A better example follows!
Merely stealing profile pics can be back traced. When superimposed on one another they create an indistinguishable blob. GAN faces do not.
The important take away you need to understand is this: GAN technology operates the same way and produces similar output whether it's 2019 or 2024 because the goals of the project making them, and the goals of the people using them couldn't be more dramatically misaligned. Here's a video of GAN-generated faces from 2024 superimposed atop one another to really clarify what's going on here:
Identifying GAN-generated faces on sight takes time. It's not a skill I thought I needed to cultivate but it's one I'm glad I gained over time. Here's a few easy to spot ways they stand out.
🚩Look for signs of nonsensical and inauthentic clothing or jewelry.
Our inciting incident Jennifer Young is wearing one earing on her (facing us) left ear. 🚩 Hair placement over her right ear looks like there could be an earring there but isn't if you look close. If she were a person, she doesn't strike me as the sort of person inclined to fashionable asymmetry. I am. I am 100% about asymmetrical style. She would not be.
🚩Next look at the necklace Jennifer Young's wearing. It's lacking substantial detail and appears to be cutting in to her skin. If that were a chain it'd be light enough to rest comfortably on her neckline. Unless maybe it's tungsten. And who wears that?
Other tells you're looking at GAN-generated faces include vestigial heads near the edges of the image (sometimes referred to as "side demons") and malformed glasses or goggles that sometimes blend into the face or are unrealistically asymmetrical. All are also telltale signs of GAN-generated pics.
Other potential indicators of GAN-generated face pics include nonsensical clothing and hats made of colorful abstract blobs rather than any real material. GAN-generated "sports caps" will often have random colorful debris where the team logo would be if it were a real hat.
GAN-generated face pics also sometimes have trippy backgrounds that often superficially resemble buildings or outdoor scenes but are composed of abstract shapes and blobs of varying colors when looking closer.
Jennifer Young's background looks nonsensical to me. It's blurred enough to allow for humanity's innate pattern seeking skills to imply a planter box next to a cement pillar in front of a building with various bricks façade. Unlikely. I'm familiar enough with mixed-use natural living plant integrated façade decoration to know that's not it. Additionally the sidewalk behind her doesn't have a fixed perspective. It's different on each side of her head. There's no continuity here to speak of. A wisp of hair at the center top of her head serves as a delimitator between the fake façade and out-of-focus nature scape to the right.
Not convinced yet? I got you. I ran a reverse image search on her photo. Let me take you down the rabbit hole - all it'll take is 1 minute and thirty seconds. This comes with a few other GAN-generated recruiters who, on the surface, look like they could be legit representatives of Jobcase.
Notice how Jennifer Young ONLY exists in Jobcase listings? And when clicked in to those listings she's not the "Jobcase Community Specialist" listed on those job listings? A generic listing exists instead.
Found another person I thought was a GAN entry. Kai Dickerson. She shows up on 1 listing. Clicked in to the listing and she's not assigned to it just like Jennifer Young. But Kai Dickerson has a LinkedIn. She's the Director of Communications and Advocacy at Jobcase, Inc. Jobcase is, if nothing else, a legitimately certified corporation registered with the Better Business Bureau.
Perfect timing to bring up the list of complaints filed against Jobcase, Inc.
At The Better Business Burea: https://www.bbb.org/us/ma/cambridge/profile/electronics-and-technology/jobcase-inc-0021-145334/complaints
The complaint filed with the BBB 12/31/2023 catches my eye for it's my experience with one key difference. I never applied for a job through Jobcase, Inc. I did apply for a job through CareerBoutique via Talentfy.io.
Needless to say I will continue to ignore all communiques from Jobcase, Inc.
So- you see a face you're suspicious of. What would be positive indications the person is real? I'll borrow my friend's headshot. A public figure who should be easy to find online in a professional capacity!
See! Her photo shows up at multiple links and has a real listing on a real website demonstrating she's a real person.
GAN faces first appeared in 2014. In a GAN, two neural networks contest with each other in the form of a zero-sum game, where one agent's gain is another agent's loss. They're actions are similar to mimicry in evolutionary biology, moving towards an evolutionary arms race to improve; and they are always improving. For the technically inclined check out the Wikipedia: https://en.wikipedia.org/wiki/Generative_adversarial_network
Although the GAN technique existed since 2014. It wasn't until 2019 these technology was made publicly available.
Here's how they're used out in the wilds of the open internet: Click to see how they inflated engagement in a 2021 election race in Texas: https://x.com/conspirator0/status/1457049273971916802?s=20
They've been a bit of a problem ever since. Now there's a new variant out there to terrorize job seekers driving the narrative of "I've sent in 1000 applications and NOTHING". Yeah. In addition to factors like Application Tracking Systems, an overwhelming number of well qualified applicants, finite attention from hiring managers, cut Recruiter staff, and a risk adverse culture that's only eating through company budgets. Companies like Jobcase require companies to pay them to see their talent pool. Why would any company do that? Jobcase isn't demonstrating quality to their talent, nor in their business operations. And hiring companies already pay for LinkedIn's backend and likely a few others like Monster.com (?) Idealist.org, Indeed.com among other big names.
Here's 1081 accounts identified in a botnet by dedicated researchers in November 2021.
Here's 110 networked generated faces in a botnet identified in 2024. These were generated by GAN's successor StyleGAN. For the technically inclined: https://en.wikipedia.org/wiki/StyleGAN
So hey- watch out job seekers! People who figured out how to monetize wasting your time for themselves are out there capitalizing on you. The attention economy is a growth market when few other growth markets exist. Woe unto us.
So what can you do to navigate this? The solution is thankfully simple!
Just ignore these suspicious accounts and move on.
Unsubscribe from their mailing lists. It'll depreciate their trust rating to Google, etc. automatically triggering spam filters.
If the job is real apply directly with the company offering it who should have posted it on their website.
If they haven't yet posted it on their website reach out to anyone you can find at that company who might point you towards the right person to speak with.
This post would not have taken the form it did without the time, effort, and wisdom of the following OSINT researchers:
댓글