Fraunhofer SIT at it-sa 2026

AI systems can be compromised through so-called backdoors, which can cause their behavior to change in harmful ways. At the booth, you can interact with compromised and harmless chatbots and learn about common detection mechanisms for backdoors.

More info about the project

Counterfeit products and product piracy jeopardize brands’ reputations, cause significant economic damage, and jeopardize the safety and health of consumers. SmartID enables reliable authenticity verification throughout the entire supply chain: The system uses the unique characteristics of the product’s physical surface or packaging to generate a unique, counterfeit-proof QR code. Manufacturers, distributors, and end customers can use it to quickly verify whether a product is genuine or counterfeit.

For more information, click here

Until now, despite ever-advancing automation, attacks on IT systems have depended primarily on the attacker’s expertise and resources. AI techniques now make it possible to automate attacks almost entirely and scale them massively, rendering existing risk assessments ineffective. This exhibit demonstrates how an AI-powered attack can take over an entire network of standard office IT systems—a topic of interest to IT leaders (e.g., CISOs), penetration testers, and researchers.

Realistic, AI-generated faces and voices make deepfake attacks a growing threat to businesses and individuals. Criminals use manipulated video calls and phone calls to carry out CEO fraud and identity theft, spread fake news, misinform the public, and deliberately damage corporate reputations. See firsthand how such manipulated media can be identified in real time using modern forensic analysis methods—for information security, fraud prevention, and the trustworthiness of digital communication and media.

AI-generated images, videos, and voiceovers are everywhere—from social media to advertising clips to journalistic visualizations. Article 50 of the EU AI Regulation establishes clear transparency and labeling requirements for these. Our researchers demonstrate how AI providers and operators can reliably label and track synthetic media. To this end, best practices from multimedia security—such as invisible digital watermarking and robust perceptual hashing—are being further developed to ensure AI operations comply with the law.

More info