# Skyld

> Skyld secures embedded AI models against extraction, copying, theft, and manipulation through mathematical transformation. Our solution enables precise deployment control and license compliance. 

## About us

    Founded in 2023 by cryptography doctor Marie Paindavoine
    She has over 12 years of experience in cybersecurity
    French deeptch AI startup
    AI models are valuable intellectual property
    Deployed models can be stolen or reverse-engineered
    Traditional security fails for on-device AI
    Target customers deploy AI at the edge
    Skyld secures AI models directly on-device
    Protects against extraction and reverse engineering
    Defends models from adversarial attacks
    Compatible with ONNX, PyTorch, and TFLite
    Works on mobile, embedded, and IoT devices
    Preserves model accuracy and performance
    Skyld secures AI at runtime, not just data
    Built for edge and embedded AI
    Cloud security does not protect on-device models
    Minimal performance overhead
    Designed for real-world AI deployments

## Index

### Pages (EN)
* [Skyld](https://skyld.io/) : Skyld provides solutions to secure AI models and regularize AI licensing.
* [About Skyld: Pioneering AI Model Protection](https://skyld.io/company/about-us/) : Skyld is a pionneer in protecting AI models from theft, attacks and misuse. Discover our mission, values, and why leading companies trust our expertise.
* [Harden Your AI Against Adversarial Examples](https://skyld.io/solutions/adverscan) : Fortify Your AI model Against Adversarial Examples.Even high-performing AI can be deceived. Test your resilience in the lab, before your model gets deployed.
* [Control and Monetize Your Innovation](https://skyld.io/solutions/ai-licensing) : Secure, manage, and license your AI models with Skyld. Maximize profits and prevent unauthorized usage with AI licensing solutions.
* [A SDK To Fight Model Extraction](https://skyld.io/solutions/ai-protection) : Secure your on-device AI models against reverse engineering with a low computing footprint. Protect your competitive edge and technological advance.
* [Contact Us](https://skyld.io/contact) : Do you have a question integrating our protection solution? Don't hesitate to contact our experts on this page.
* [Cookie policy](https://skyld.io/cookie-policy) : Skyld's cookie policy
* [Skyld](https://skyld.io/jobs) : Skyld provides solutions to secure AI models.
* [Legal notice](https://skyld.io/legal-notice) : Skyld's legal notice
* [Privacy policy](https://skyld.io/privacy-policy) : Skyld's privacy policy
* [Request A Demo](https://skyld.io/request-a-demo) : Schedule a meeting with one of our technical experts to define your use case and find the right product to answer your needs.
* [Your Mind is Mine: How Android Apps Expose Deep Learning Models to Theft](https://skyld.io/ressources/scientific-articles) : The SSTIC 2023 talk reveals how deep learning models in Android apps are exposed to theft. Discover how ModelHunter identifies weak AI model protections.
* [Solutions](https://skyld.io/solutions/) : Discover our solutions for AI protection, licensing and our new product Adverscan.
* [Success Stories](https://skyld.io/success-stories) : Explore our success stories to see how we drive business growth and innovation. Discover real-life case studies showcasing our impactful solutions.
* [Website plan](https://skyld.io/website-plan) : Skyld's website plan
* [article](https://skyld.io/category/article/) 
* [Edge AI](https://skyld.io/category/edge-ai/) 
* [ML security](https://skyld.io/category/ml-security/) 
* [News](https://skyld.io/category/news/) 
* [Model Extraction](https://skyld.io/category/model-extraction/) 
* [ML Optimization](https://skyld.io/category/ml-optimization/) 
* [technical](https://skyld.io/category/technical/) 
* [Skyld Blog: Insights on AI Model Protection, Threats & Cybersecurity](https://skyld.io/ressources/blog/) : Read the latest insights on AI model protection, AI model licensing, and adversarial threats. The Skyld blog delivers expert content to secure your AI models.
* [Skyld Blog: Insights on AI Model Protection, Threats & Cybersecurity](https://skyld.io/ressources/blog/page/2/) : Read the latest insights on AI model protection, AI model licensing, and adversarial threats. The Skyld blog delivers expert content to secure your AI models.
* [Skyld News: Awards, Innovations, and Key Industry Events](https://skyld.io/company/news/) : Follow Skyld's latest achievements : awards, product launches, industry recognition, and our presence at top cybersecurity and AI events worldwide.

### Pages (FR)
* [Skyld](https://skyld.io/fr/) : Skyld propose des solutions pour sécuriser les modèles d'IA et régulariser les licences d'IA.
* [A propos](https://skyld.io/fr/company/about-us/) : Découvrez Skyld : une startup à la pointe spécialisée dans la protection et la monétisation des modèles d'IA.
* [Renforcez votre IA contre les exemples adversariaux](https://skyld.io/fr/solutions/adverscan) : Même les IA les plus performantes peuvent être trompées. Testez la résilience de votre modèle avant qu'il ne soit déployé.
* [Contrôlez et monétisez votre innovation](https://skyld.io/fr/solutions/ai-licensing) : Sécurisez, gérez et licencez vos modèles d'IA avec Skyld. Maximisez vos profits et évitez les usages non autorisés grâce à nos solutions de licence IA.
* [Un SDK contre l'extraction de modèles d'IA](https://skyld.io/fr/solutions/ai-protection) : Sécurisez vos modèles d'IA embarqués ou on-premise contre la rétro-ingénierie. Préservez votre avance technologique, sans compromis sur la performance.
* [Nous contacter](https://skyld.io/fr/contact) : Vous avez une question à propos de notre protection ? N'hésitez pas à contacter nos experts sur cette page.
* [Politique de cookies](https://skyld.io/fr/cookie-policy) : Politique de cookies de Skyld
* [Skyld](https://skyld.io/fr/jobs) : Skyld provides solutions to secure AI models.
* [Mentions légales](https://skyld.io/fr/legal-notice) : Mentions légales de Skyld
* [Politique de confidentialité](https://skyld.io/fr/privacy-policy) : Politique de confidentialité de Skyld
* [Demander une démonstration](https://skyld.io/fr/request-a-demo) : Prenez rendez-vous avec l'un de nos experts techniques pour définir votre cas d'utilisation et trouver le bon produit pour répondre à vos besoins.
* [Articles scientifiques](https://skyld.io/fr/ressources/scientific-articles) : Découvrez des articles scientifiques rédigés par nos experts.
* [Solutions](https://skyld.io/fr/solutions/) : Découvrez nos solutions pour protéger vos modèles d'IA, pour les monétiser via un système de licence ainsi que notre nouveau produit Adverscan.
* [Témoignages clients](https://skyld.io/fr/success-stories) : Découvrez nos témoignages clients : croissance et innovation grâce à nos solutions.
* [Plan du site web](https://skyld.io/fr/website-plan) : Plan du site web de Skyld.
* [Blog](https://skyld.io/fr/ressources/blog/) : Consultez le blog de Skyld, vous en apprendrez beaucoup sur les enjeux et les solutions de cybersécurité de l'IA on-device.
* [Blog](https://skyld.io/fr/ressources/blog/page/2/) : Consultez le blog de Skyld, vous en apprendrez beaucoup sur les enjeux et les solutions de cybersécurité de l'IA on-device.
* [News](https://skyld.io/fr/company/news/) : Jetez un oeil aux évènements et compétitions auxquelles nous avons participé récemment.

### Blog Articles
* [Minister Anne Le Hénanff Visits the Cyberplace: A Shared Vision for AI Cybersecurity, and Digital Sovereignty](https://skyld.io/Minister-Anne-Le-H%C3%A9nanff-Visits-the-Cyberplace) - Published on : 11-06-2025 : Skyld affirms its commitment to secure, ethical, and sovereign AI at the heart of the French Tech ecosystem.
* [When On-Device AI Becomes a Security Flaw: The SafetyCore Case Study](https://skyld.io/when-on-device-ai-becomes-a-security-flaw-safetycore-case) - Published on : 10-17-2025 : We recently conducted a case study on SafetyCore, a Google systems service that embed an on-device AI model for sensitive image content detection. Our analysis demonstrates how attackers can extract and manipulate the AI model, effectively bypassing its utility.
* [Marie Paindavoine Recognized Among the Top 100 Women in Tech in Europe 2025](https://skyld.io/Marie-Paindavoine-Recognized-Among) - Published on : 10-09-2025 : Skyld joins the first SISTA AI program
* [Skyld, French Deeptech Startup, Raises €1.5 Million to Secure AI Models](https://skyld.io/Skyld-raises-15million) - Published on : 10-01-2025 : We are pleased to announce our first round of fundraising. The round was backed by French and Swiss investors: Auriga Cyber Ventures, BNP Développement, Bloomhaus, Sodero, and seasoned industry business angels.
* [Can TensorRT AI Models Be Reverse-Engineered?](https://skyld.io/nvidia-reverse) - Published on : 09-26-2025 : This article provides details on how to extract or reverse-engineered a Nvidia TensorRT AI model
* [Skyld joins the first SISTA AI program](https://skyld.io/kyld-joins-the-first-SISTA-AI-program) - Published on : 09-16-2025 : Skyld joins the first SISTA AI program
* [Adversarial Patches in the Wild](https://skyld.io/adversarial-patches-in-the-wild) - Published on : 07-06-2025 : We created and tested an adversarial patch that tricks real-world object detection models into seeing stop signs even when none exist. Learn how it works, and how to defend your models.
* [How to Build an Adversarial Patch?](https://skyld.io/how-to-build-an-adversarial-patch) - Published on : 07-05-2025 : Create adversarial patches on YOLOv5 using PGD and ART. Fool object detectors with code and test your model’s robustness.
* [How to Build an Adversarial Example?](https://skyld.io/how-to-build-an-adversarial-example) - Published on : 07-04-2025 : Create adversarial examples on YOLOv5 using PGD and ART. Fool object detectors with code and test your model’s robustness.
* [How AI Models can be Fooled: Adversarial Examples?](https://skyld.io/how-ai-models-can-be-fooled-adversarial-examples) - Published on : 07-03-2025 : Learn how simple modifications to inputs can fool AI models and cause unexpected behaviors. Understand the security risks and vulnerabilities in modern artificial intelligence systems.
* [How to Accurately Measure VRAM Usage](https://skyld.io/vram-profiler) - Published on : 06-23-2025 : This article provides a practical guide on how to accurately measure VRAM usage in Python for NVIDIA GPU.
* [Innovation and Recognition: Skyld among the 100 most promising AI startups according to RAISE Summit 2025](https://skyld.io/100-Most-Promising-AI-Startup) - Published on : 06-20-2025 : Innovation and recognition for Skyld: ranked among the 100 most promising AI startups at RAISE Summit 2025. A proud moment for the whole team!
* [TV Show – AI and Cybersecurity: New Threats, New Protections](https://skyld.io/Tv-Show-AI-And-Cybersecurity) - Published on : 06-16-2025 : We had the pleasure of being invited by Thales to take part in a special live TV show during the Forum InCyber. The roundtable focused on a crucial concern: AI as a target for cyberattacks
* [VivaTech 2025: Go Behind the Scenes of Our Participation!](https://skyld.io/VivaTech-2025) - Published on : 06-14-2025 : From June 11 to 14, 2025, Paris Expo Porte de Versailles opened its doors again for the biggest innovation event in Europe: VivaTech 2025.
* [In The Press: Skyld in the spotlight in ‘La Jaune et la Rouge’!](https://skyld.io/Skyld-in-the-spotlight-in-La-Jaune-et-la-Rouge) - Published on : 05-20-2025 : the École Polytechnique alumni magazine devotes an article to Marie Paindavoine, founder and CEO of Skyld.
* [Google Photos' AI Models: The Secret Sauce That Can Be Stolen](https://skyld.io/google-photos-model-extraction) - Published on : 03-31-2025 : Even Google Photos is vulnerable to AI model extraction. Learn how we extracted Google's AI models, and why classical software protection are not adapted to AI model.
* [Skyld Among the Têtes 2025: A Bold and Inspiring Vision Hailed by API and Ouest-France](https://skyld.io/skyld-among-the-tetes-2025) - Published on : 03-20-2025 : Skyld joins the Têtes 2025 list with a bold vision for AI. Praised by API and Ouest-France, the startup stands out for its innovation and ambition
* [Season 7 Of The Women Entrepreneurs Program by Orange Kicks Off](https://skyld.io/Season7-of-the-women-entrepreneurs) - Published on : 03-14-2025 : Skyld is proud to be selected for the 2025 Women Entrepreneurs Program by Orange, supporting 100 women-led startups across France.
* [Skyld joins the Female Founders Fellowship at STATION F](https://skyld.io/Skyld-joins-the-female-founder) - Published on : 03-07-2025 : Skyld has been selected for the prestigious Female Founders Fellowship 2025 at STATION F, joining 15 other top women-led startups in tech.
* [Skyld Wins the #WomenInTech Award: A Big Step for AI Cybersecurity and Women in Tech](https://skyld.io/Skyld-wins-the-women-in-tech-award) - Published on : 03-06-2025 : Marie Paindavoine, CEO of Skyld, wins the first Women in Tech Award for her breakthrough in AI model security.
* [Marie Paindavoine Has Been Selected As One Of The 100 Innovators Of 2025 By Le Point!](https://skyld.io/Marie-has-been-selected) - Published on : 02-18-2025 : Marie Paindavoine, CEO of Skyld, is included in the list of the 100 innovators of 2025 by Le Point.
* [Skyld on the Red Carpet: Review of WAICF 2025](https://skyld.io/Skyld-on-the-Red-Carpet) - Published on : 02-13-2025 : Skyld took part in WAICF 2025. Our cybersecurity expert Mathis Mauvisseau gave a presentation on the risks of AI model extraction and how to protect yourself.
* [Skyld at AI Business Day: The Summit for Action on Artificial Intelligence](https://skyld.io/Skyld-at-AI-Business-Day) - Published on : 02-11-2025 : Edge AI took center stage with Qualcomm and STMicroelectronics, but Skyld raises a vital concern: AI model security remains dangerously underdiscussed.
* [Marie Paindavoine Joins Women TechEU Program: A Recognition For Women In Deep Tech](https://skyld.io/Marie-Paindavoine-joins-women-techeu) - Published on : 02-10-2025 : Marie Paindavoine, CEO of Skyld, is among 40 winners of the prestigious EU-funded Women TechEU program.
* [3rd Edition of the Winter Research School by CyberSchool](https://skyld.io/3rd-edition-of-the-research) - Published on : 02-06-2025 : Skyld took part in the 3rd edition of the Winter Research School. Marie Paindavoine, CEO of Skyld, gave a presentation on model extraction.
* [13 French Startups to Watch And Skyld Is On the List!](https://skyld.io/13-French-Startups-to-Watch-And-Skyld-Is-On-the-List) - Published on : 12-12-2024 : Skyld is proud to be featured in Sifted’s list of top French startups beyond Paris. We help companies protect and monetize their edge AI models with confidence.
* [In The Spotlight: Marie Paindavoine Wins The Cyber Woman Award!](https://skyld.io/In-The-Spotlight-Marie-Paindavoine-Wins-the-Cyber-Woman-Award) - Published on : 12-10-2024 : Skyld CEO Marie Paindavoine was honored with the European Cyber Woman Award in Paris, recognizing her leadership at the intersection of AI and cybersecurity.
* [Imagine Summit: A Day Focused on Innovation](https://skyld.io/Imagine-Summit-A-Day-Focused-on-Innovation) - Published on : 12-09-2024 : Skyld shared its mission for trustworthy AI, met Mayor Nathalie Appéré, and joined a panel on startup collaboration with the Ministry of Armed Forces.
* [GenerationAI: A Vision of AI Beyond Algorithms](https://skyld.io/GenerationAI-A-Vision-of-AI-Beyond-Algorithms) - Published on : 12-03-2024 : Skyld's CEO Marie Paindavoine gave a presentation on the risks of AI model extraction and how to defend against them.
* [New Deployments, New Threats: How To Protect Local AI Models From Reverse Engineering?](https://skyld.io/New-Deployments-New-Threats-How-To-Protect-Local-Ai-Models-From-Reverse-Engineering) - Published on : 09-16-2024 : Discover effective strategies to protect on-device and on-premise AI models from extraction and abusive reuse. Skyld secures your AI models against reverse-engineering and allow a strong AI licensing strategy.
* [Attack On AI Models: What You Need to Know!](https://skyld.io/Attack-On-AI-models-What-You-Need-to-Know) - Published on : 08-28-2024 : AI models face growing threats such as IP theft, adversarial attacks, and unauthorized copy. Explore how to safeguard your models with SKYLD’s solutions and stay ahead of emerging risks.
* [AI Learning Types : Supervised, Unsupervised & Reinforcement](https://skyld.io/What-Are-the-Three-Types-of-Artificial-Intelligence-Learning-and-Their-Specific-Features) - Published on : 08-22-2024 : Explore the three fundamental types of AI learning : supervised, unsupervised, and reinforcement. What are their features and differences?
* [Skyld Joins The New Season Of Thales Cyber@Station F !](https://skyld.io/Skyld-Joins-The-New-Season-of-Thales-Cyber@StationF!) - Published on : 06-25-2024 : Skyld is proud to join the Thales Cyber@Station F acceleration program, gaining access to expert coaching and strategic partnerships.
* [We're Now Part Of The NVIDIA's Inception Program For Startups.](https://skyld.io/Skyld-Joins-NVIDIA's-Inception-Program-For-Startups) - Published on : 06-10-2024 : Skyld is now part of NVIDIA’s Inception Program, gaining access to cutting-edge AI tools, cloud resources, and networking opportunities with investors.
* [Skyld At VivaTech 2024](https://skyld.io/Vivatech-2024) - Published on : 05-22-2024 : Skyld took part in VivaTech 2024, showcasing its AI security tech at the DFKI booth, meeting INRIA’s CEO, and pitching at the European Deeptech Meetup.
* [On Wavestone's Radar!](https://skyld.io/Skyld-In-Wavestone's-Radar) - Published on : 05-22-2024 : Skyld is proud to be recognized by Wavestone and Bpifrance among the leading French cybersecurity startups in AI for 2024.
* [Skyld, The Nominated Startup, Wins The Audience Award At The Rennes Creators' Morning!](https://skyld.io/Skyld-Creator-Morning) - Published on : 04-15-2024 : Skyld, founded by Marie Paindavoine, PhD, was nominated to pitch at Creators' Morning in Rennes and won the Audience Award.
* [How Do You Reverse-Engineer an Android App?](https://skyld.io/reverse-ingeneer-android-app) - Published on : 04-15-2024 : Using static and dynamic reverse-engineering, we show how to extract AI models from Android applications.
* [What Are the Applications of On-Device Machine Learning?](https://skyld.io/Ml-models-hidden) - Published on : 04-15-2024 : Understanding where AI models reside across various applications and their vulnerability to model theft, model inversion and adversarial attacks or intellectual property theft.
* [Skyld At The ECSO Startup Award Finals Which Took Place On March 21st And 22nd In Germany](https://skyld.io/Skyld-at-the-ESCO-Startup-award-finals) - Published on : 03-23-2024 : Skyld participated in the finals of the European Cyber Security Organisation (ECSO) Startup Award.
* [Skyld In The Spotlight for The 38th Edition Of The Ille-et-Vilaine Oscars, Broadcast Live On TV Rennes](https://skyld.io/Skyld-in-the-spotlightfor-broadcost-live-on-TV-Rennes) - Published on : 02-13-2024 : Skyld secured second place at the Ille-et-Vilaine Oscars, celebrating innovation in AI security.
* [Flashback To Our Participation At WAICF At The Palais des Festivals Et Des Congrès De Cannes](https://skyld.io/WAICF-2024) - Published on : 02-08-2024 : Skyld joined global AI leaders at WAICF 2025 in Cannes to present its cutting-edge solution for protecting embedded AI models.
* [Artificial Intelligence Model Extraction](https://skyld.io/model-extraction) - Published on : 02-06-2024 : Discover how easy AI model theft can be. Be conscious of protecting your AI models against extraction, malicious attacks, and intellectual property theft.
* [Expert Series: Understanding and Defending Against Adversarial Attacks on AI by Thibault Maho](https://skyld.io/Expert-series) - Published on : 01-31-2024 : Adversarial attacks can manipulate AI model behavior and compromise reliability. Discover how to spot and defend against these threats from a leading AI security researcher.
* [Model Inversion Attacks in Machine Learning: Are Your AI Models at Risk?](https://skyld.io/model-inversion-attacks) - Published on : 01-31-2024 : Model inversion attacks can expose private or confidential data from AI training sets. Learn how these attacks work and how to defend your AI models against privacy leaks and confidentiality breaches.
* [Protect Your On-Device Artificial Intelligence Algorithms: Encryption Is Not Enough!](https://skyld.io/Protect-your-on-device-AI-Algorithms) - Published on : 01-31-2024 : Can encryption protect on-device or on-premise AI algorithms from reverse engineering?
* [Edge AI: Benefits, Applications and Risks](https://skyld.io/edgeAI) - Published on : 05-02-2023 : Explore the pros and cons of Edge AI, its key benefits and uses cases. Uncover the associated risks to your AI models, including adversarial attacks, intellectual property theft, and AI model theft.