Photos generated by artificial intelligence are becoming increasingly realistic from week to week. The boundaries between reality and AI-created images are becoming less visible. All of this is made possible by Midjourney Inc., a leader in the field of AI-generated graphics.
Midjourney Inc Development
Midjourney Inc. was founded in San Francisco by David Holz, co-founder of Leap Motion, and has gained tremendous popularity since its inception. Its GenAI platform, which was launched in beta in July 2022, has become one of the most important tools for creating this type of graphics. However, the real breakthrough came in 2023 when the company introduced version 6 of its solution, which set entirely new standards for image generation.
Features of the Latest Version
The latest version offers a range of new features and improvements aimed at facilitating, speeding up, and enhancing the efficiency of the art creation process. Its advantages include higher resolution (2048×2048 pixels), advanced natural language processing for more precise and realistic images, greater control over graphic variations, improved hand generation, the ability to create 3D models, and text-to-film capabilities.
Top graphic – v 6
Bottom graphic – v 5.2
Source: https://android.com.pl/tech/672882-midjourney-v6-ai-obrazy/
The cheapest package from Midjourney is $10 per month, or $8 with annual billing (as of January 2024 data). It allows for the generation of 200 minutes of work, equivalent to approximately 100 images.
The tool’s creator also announces enhancements such as 3D and video capabilities, which are likely to be introduced in version 7. The company also plans improvements to its web and mobile platforms.
How to distinguish between a photo generated by AI and a real photo?
However, amid this fascinating technology, a problem arises. How to distinguish between a photo generated by AI and a real photo? This question becomes increasingly significant, especially in the context of the potential for misuse of this technology. This year, a photo generated on Midjourney V6 with the prompt “phone photo of man holding post-it note, posted to Reddit in 2018” was published – v 6.0 -s 0 -style raw -ar 9:16 w.
What should you pay attention to?
Here is a brief analysis of this photo. Several elements that immediately catch the eye:
- Three white dots on the forehead form a perfect triangle.
- The skin on the fingers lacks texture, the hand looks ‘plastic’, waxy.
- The face has an unnatural shine.
- There is a noticeable difference in the size of the ears.
- The collar on the left side (from the viewer’s perspective) also stands out, its left point magically becomes a seam on the shoulder.
- The hair is smoothed out, almost airbrushed.
- Additionally, the jawline seems to be mirrored.
- It is worth noting that most faces are quite asymmetrical, especially the nose.
- The size of the hand, the size of the note, the angle of the note.
The text on the note was added using Photoshop.
Of course, these features can be subtle, and sometimes it can be difficult to notice them, so it’s worth examining the details carefully, such as:
- Symmetry and proportions: Check if the facial proportions are realistic. AI sometimes struggles to maintain facial symmetry, which can result in minor but noticeable asymmetries.
- Details and textures: Take a close look at the texture of the skin, hair, and clothing. Generated images may contain unnatural patterns or texture errors, such as repeating elements.
- Background: Backgrounds in AI-generated photos may appear blurry or contain illogical elements. Pay attention to whether the background is consistent.
- Lighting and shadows: Inconsistencies in lighting and shadows may indicate that the photo was generated. Shadows should correspond to the direction and source of light.
- Graphic elements and text: AI may have difficulty generating text or other small graphic elements correctly. Check if the text is legible and if all characters are correct.
The use of AI for fraud
It is also worth mentioning cases where generated images are used for fraud. A Reddit user used the Stable Diffusion tool to generate realistic photos to deceive the identity verification system. This opens the door to potential fraud and abuse, especially in the context of Know Your Customer (KYC) procedures.
Source: https://www.facebook.com/story.php?story_fbid=880657874064466&id=100063607833505
KYC is a standard identity verification process that involves submitting a photo of the user holding their identity document, such as a passport or driver’s license. However, AI generative tools like Stable Diffusion now allow for the creation of fake verification photos in just a few minutes. This poses a serious threat to financial institutions, online platforms, and other entities relying on KYC to prevent fraud and maintain regulatory compliance.
The advancement of AI generative technology brings new security challenges. Therefore, developers and security experts must collaborate to develop AI-resistant verification systems. One such solution is TimeEvidence.com, which offers a live verification system that eliminates the possibility of AI fraud by enforcing a specific user position in real-time.
It is important for us to be aware of the potential threats associated with the growing power of generative artificial intelligence and work on developing precautionary measures. In the development of AI technology, we will need to stay ahead of it to safeguard our data and identity.
Analyst’s comment:
Just six months ago, people didn’t need to scrutinize photos closely. Now, even if they do so with zoom, it’s not always obvious. For me as an OSINT analyst, distinguishing between AI-generated photos and authentic ones is incredibly crucial. False or manipulated images can be tools of disinformation or fraud, which is why meticulous graphic analysis and utilizing available OSINT tools and techniques become key elements in verifying sources and content in cyberspace. An OSINT analyst must be prepared to detect subtle signals in images that may indicate their artificial origin and contextualize them to understand potential threats or objectives associated with a given image.
Author: Alicja Pawłowska
Sources:
https://www.facebook.com/story.php?story_fbid=880657874064466&id=100063607833505
https://openaijourney.com/stable-diffusion-fake-user-verification/
