AI-Generated Selfies: What are they and How Your Business can Counter them?

Artificial Intelligence (AI) is quickly growing to become an integral part of our personal and professional lives. You might have often come across apps or websites that create “professional selfies” using AI. These platforms use generative AI to create these images.

Now while such an application of generative AI might be pretty useful, there is a flipside to this coin. Fraudsters and bad actors have started using generative AI to create fake images of people to commit identity fraud and circumvent the KYC processes of businesses. According to the FBI’s Internet Crime Report 2023, losses from identity theft stood at $126 million. With the integration of AI into the process, these numbers could increase in the future. Let’s find out how these AI-generated selfies work and how your business can counter them.

What are AI-generated Selfies?

AI-generated selfies are hyper-realistic selfies created using generative AI. These images can either be modifications to the faces of existing people, or brand new faces that the AI comes up with. Typically these images are created using text prompts. Bad actors will use prompts to create images of people striking certain poses and with specific facial features to make the image appear more realistic.

Some of the most commonly used models for creating fake selfies include generative adversarial networks (GANs), variational autoencoders (VAEs), diffusion models, and neural radiance fields (NeRFs). These models have functions like decreasing detectability, generating multiple outputs from the same inputs, adding realistic lighting, and more. When used together, these models can make finding fake selfies difficult.

Now that you know what these AI-generated selfies are and how they’re made, let’s get to the meat of the matter, and find out how these selfies can pose a threat to your business.

Why are AI-generated Selfies a Threat to Businesses?

AI-generated selfies are most commonly used by bad actors to create fake profiles and accounts. These fake profiles are used to apply for and avail loans, launder money, fund illegal activities, and more.

Failure to detect these fake selfies and profiles can lead to your business getting involved in illicit activities, which can lead to legal penalties. TD Bank, one of the USA’s largest banks, was fined $1.3 billion by FinCEN for the violation of the Bank Secrecy Act (BSA). BSA is one of the most prominent anti-money laundering regulations in the US. Now, your business might not be implicated in any illegal activities itself, but any legal proceedings around it can lead to loss of patronage from your customers. Now that you know the threat that these fake images can pose against your business, let’s see how to counter this problem.

How to combat the rise of AI-generated selfies

Here are some of the ways your business can detect and prevent the use of fake selfies:

1. Integrate Liveness Checks

Selfie verification alone can often fail in verifying the identity of the user submitting them. This is because selfie verification works solely off of a picture that is uploaded by the user, and this image could be a fake one generated through AI. This weakness was initially mitigated by making users click live selfies instead of uploading an image file.

However, fraudsters caught up to this change and have since been able to hijack the camera of the device to emulate a camera being used to upload an image. Thankfully, the problem does not end on a bad note. Liveness checks have improved the state of identity verification. Signzy’s Liveness Verification API helps businesses integrate robust liveness checks into their verification processes without causing any customer drop-offs due to a poor user experience.

2. Add Complexity to the Selfie Verification Process

Another way to reinforce your business’ consumer onboarding process is to make users strike certain poses when they upload their selfies. This is a somewhat “low-tech” method to avoid fraud using AI-generated selfies and may generate false positives. However, this is a good starting point in selfie verification if your business is just starting out, and does not need third-party software that can scale.

3. Go Beyond Selfies and Liveness Checks

While making your customers upload selfies and conducting liveness checks on them works, limiting your identity verification to just these two processes can lead to vulnerabilities in your system. You can improve the state of your business’ identity verification by integrating hardware and behavior-specific signals from the user’s end.

Hardware-based verification can include the user’s IP address, location data, fingerprint data, and more. These hardware signals are also called passive signals or device signals.

Behavior-based verification includes validating the authenticity of a selfie based on the subtle signs that the human face makes. It can include the user’s blinking frequency, facial twitches, and more. Essentially, these signals make it so that not only AI-generated images but also videos can be verified.

4. Add Multiple Layers of Verification Steps

Similar to the previous step, you can increase the number of verification processes that a user has to go through to ensure the security of your business ecosystem. After all, selfie verification is only one side of the identity verification dice. Security and verification systems need to go beyond and include processes like DL verification, SSN verification, passport verification, and more.

How Signzy Helps with Identity Verification

When working with security and user experience, a balance needs to be struck. While more verification techniques may improve the state of your business’ security, too many steps could lead to customer dissatisfaction and even drop-offs. To ensure that both checkboxes of security and customer satisfaction are met, you need to ensure that your business’ verification processes are as streamlined as possible. Thankfully with Signzy, your business’ Identity Verification can be reinforced with a variety of APIs that go beyond the ones mentioned above and avoid user drop-offs. From Face Match and Liveness Check to a plethora of other APIs, Signzy has your verification needs covered. To see which of our other offerings work for your business, book a call with us.

Conclusion

AI-generated selfies could pose a threat to a business’ verification processes and security. However, most prudent businesses have multiple checks in place. So even in the unlikely scenario that a bad actor slips through one of the many checks, it is nearly impossible for a bad actor to circumvent all the measures put in place by a business.