Q-CHECK

The authenticity of an image is difficult to guarantee. Indeed, image retouching and modification techniques are becoming increasingly effective and difficult to detect, thanks in particular to the emergence and adoption of generative AI solutions. Whereas software-generated modifications could be detected by conventional methods using image pixels, format, geometry or lighting, the detection of images generated and/or manipulated by Deep Fakes generative AI models requires a far greater effort. The most common solutions for identifying DeepFakes use deep learning models, which are effective but costly, not least because of their training requirements and the need for adequate, voluminous training databases.

Q-CHECK combines the speed of pixel analysis techniques, namely Error Level Analysis(ELA), the precision of deep learning, namely a pre-trained convolutional neural network model, and the efficiency of emerging quantum technologies to identify manipulated regions efficiently in near-real time. More specifically, Q-CHECK leverages the potential of quantum optimization algorithms, namely quantum annealing, to analyze the precious details that nestle in ELA images to reveal the manipulated regions of an image.

The benefits of Q-CHECK are manifold: Q-CHECK not only assesses the reliability of an image, our module goes beyond this by enabling the user to precisely identify manipulated areas of the image. In fact, our quantum segmentation layer quickly and efficiently analyzes the subtleties of the image to pinpoint any tampered areas. This segmentation makes it easier for experts to interpret results, understand and justify accusations of retouching, and assess the extent of retouching carried out, as well as the seriousness of the fraud.

A wide field of action: Q-CHECK effectively combats all types of image retouching, from simple copy-paste and splicing to manipulations generated by recent generative AI models, the famous DeepFakes. Q-CHECK lets you guarantee the authenticity of your images using a single tool.

Speed of calculation: Q-CHECK enables fast, efficient image verification. By intelligently combining the finesse and precision of neural networks with the speed of analog quantum machines to solve optimization problems, Q-CHECK enables fast and efficient identification of tampered areas in your images, enabling you to cope with the growing volumes of images to be analyzed in reasonable time.

Simplified scaling: Q-CHECK can analyze large and/or high-resolution images in reasonable time.

Improved image and reliability: Q-CHECK effectively combats image falsification and guarantees the authenticity of images distributed and used.

Examples of Q-CHECK use cases

Detecting manipulated photos of car accidents

Motor insurance fraud is a growing scourge for motor insurers. According to the ALFA 2023 report, the cost of insurance fraud jumped by +48 M€ between 2022 and 2023. Unfortunately, this is likely to intensify further with the mass adoption of generative artificial intelligence, facilitating the preparation of fraudulent claims that are ever more realistic and difficult for adjusters to detect. Mass digitization now enables motor insurers to ask their policyholders for detailed snapshots of claims to facilitate their verification. However, the explosion in DeepFakes and their availability to the general public poses a problem in terms of the authenticity of the images transmitted. One insurance company wants to authenticate images of automobile claims in order to combat fraud and minimize losses.

Detail

Q-CHECK detects manipulations in even the most realistic photos, as well as those manipulated by generative artificial intelligence.

What's more, the quantum segmentation included in Q-CHECK makes it possible to analyze images attached to claims to detect any altered regions in the images, making it easier for adjusters to justify, qualify and quantify the severity of fraud, and enabling them to make informed decisions about the nature of claims.

Advantage: by virtue of its sobriety in terms of computing resources, Q-CHECK reduces the costs of image analysis in economic and environmental terms, and enables a greater volume of declarations to be processed in a shorter time. What's more, thanks to quantum annealing, the operation of segmenting a doctored image is particularly fast and efficient, since the mathematical model chosen is ideally suited to quantum annealing.

Detecting fraud in scientific articles

Today, around 2% of researchers admit to having falsified data presented in an article submitted to a scientific journal. The phenomenon is more widespread in medicine, biology and the natural sciences, where images are a frequently used source of data to support results or justify observations. A scientific journal therefore wishes to strengthen its tools for detecting scientific fraud, in order to protect its reputation and stem the spread of false information.

Detail

Q-CHECK allows you to distinguish a genuine image from one that has been manipulated by retouching and copy-and-move techniques, often used to improve the cleanliness of an observation and mask undesirable areas of the results that could weaken a prefabricated argument.

Advantage: although the subjects covered are varied, our hybrid algorithm is able to highlight these fraudulent operations on any type of image and covers a wide spectrum of falsifications from simple copy-and-drop operations to subtle DeepFake. What's more, the quantum segmentation layer enables the identification of manipulated areas, making it possible to assess the extent and severity of the fraud, as well as its impact on the study results.

Detecting counterfeits in digital commerce

Image falsification is a scourge affecting online sales platforms. Indeed, subtle modifications are sometimes made to counterfeit product images to make them look more like the originals on e-commerce platforms, undermining the platform's credibility and exposing customers to scams. In 2019, counterfeits accounted for around 2.5% of global trade and 5.8% of European Union imports.
To curb this phenomenon, a digital commerce company wants to implement a tool to identify ads containing doctored photos, thereby improving customer satisfaction.

Detail

Details

Q-CHECK can quickly identify whether a product photo has been altered, even with the help of generative artificial intelligence.

Manipulated details are identified quickly and efficiently thanks to our segmentation algorithm based on quantum annealing.

Advantage: by virtue of their computational sobriety, Q-CHECK algorithms lend themselves well to execution on large image sets. What's more, identifying the areas manipulated by the quantum optimization layer facilitates the work of experts and the substantiation of accusations of counterfeiting and fraudulent offers.