What is it about?
Today’s connected devices, such as sensors in smart cities and factories, often need to work together to train Artificial Intelligence (AI) systems without exposing their private data. This method, called Federated Learning, is fast and private, but it is also vulnerable: attackers can inject false data, manipulate the training, or add fake devices, making the results unreliable. Our research solves this problem by adding blockchain, which keeps a secure and unchangeable record of the process. This ensures that only trusted devices can join and that the shared model cannot be secretly tampered with. Our tests show that this approach blocks attacks while keeping the system quick and efficient, making collaborative AI safer and more reliable for real-world Internet of Things environments.
Featured Image
Photo by FlyD on Unsplash
Why is it important?
Cyberattacks on artificial intelligence models are a growing threat, especially in the Internet of Things, where billions of small devices are now connected worldwide. Current solutions often fail to stop advanced attacks or are too resource-intensive for these lightweight devices. Our work is important because it introduces a blockchain-based defence that both secures the learning process and remains efficient for large-scale use. By preventing poisoned data, fake devices, and tampered updates, this approach makes collaborative AI much safer and more reliable. These advances are timely, as smart cities, healthcare, and industrial systems increasingly depend on trustworthy AI to operate securely.
Perspectives
Working on this article has been a rewarding experience, as it brought together colleagues with whom I share both research interests and long-standing collaborations. For me, the most exciting part was bridging two fields, Federated Learning and blockchain, to propose a practical solution to a pressing problem. I hope that this work not only contributes to academic discussion but also inspires further applied research, especially as secure and trustworthy AI becomes increasingly critical for our everyday technologies. More broadly, I hope readers find in this article a reason to reflect on how we can build digital systems that are not only powerful but also safe, reliable, and designed with resilience in mind.
Luis Miguel García-Sáez
University of Castilla-La Mancha
Read the Original
This page is a summary of: Poisoning-Resilient Federated Learning for MEC-IoT Environments Using Blockchain, ACM Transactions on Internet Technology, September 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3767740.
You can read the full text:
Resources
Contributors
The following have contributed to this page
