A blockchain is a digital record-keeping system that is shared across many computers. It helps securely track transactions, ensuring no one changes past records unless everyone in the network agrees and updates all connected records.
Although blockchain is not yet a widely used technology, it has many potential uses. Blockchain can make money transfers faster and cheaper by reducing banking fees. Businesses can use it to track products in supply chains and improve quality control. It can help people manage their digital identities and securely share data between industries.
Blockchain can also protect music and software rights by ensuring fair payments to creators. In health care, it can help store medical records and manage clinical trial data safely.
Sara Rouhani is an assistant professor of computer science in the U of M’s faculty of science who currently works mainly on distributed systems, centralized systems and blockchain technology. She was introduced to blockchain technology during her postdoctoral program.
“My supervisor just introduced the idea, and then at that time there were very few, maybe less than 10 academic papers on this topic,” she said about blockchain technology. “I had to go through a comprehensive exam. We had to write a huge literature review with hundreds of references, and it didn’t sound feasible at that time, but I gave myself a few weeks and months. I dug into the technology and the directions.”
While conducting research, Rouhani gained hands-on experience with advanced technologies such as distributed systems, documentation, virtualizations and programming languages like ROSS.
Initially, her focus was distributed and decentralized access control using blockchain.
“I started looking into different research lines, including blockchain interoperability,” she said. “We can [then] connect different blockchain platforms — we look into digital identities and self-serving identities, so implementing the identity systems that the participants can control their identity information.”
Rouhani is exploring decentralized artificial intelligence (AI), focusing on how to use blockchain-based federated learning. The goal is to ensure the entire process is fully decentralized, without a central server or a single party controlling the data and model training.
“We use blockchain technology to build a secure and reliable infrastructure for various applications in different sectors such as health care, smart grid, supply chain and more,” she said. “We’re also working toward enhancing the underlying technology by, for example, focusing on problems such as blockchain and interoperability and utilizing machine learning to build more efficient consensus mechanisms for the blockchains.”
Rouhani is also examining how traditional training methods work, and how blockchain could improve them.
“In most of the AI-based systems,” she explained, “the data is located and collected from multiple participants, but eventually it would be transferred in one centralized server. So it means that the ones who are in charge for the model training would have access to the data and would control the data. And then any kind of privacy or security breaches can happen through that single point that the data is maintained and trained.”
Federated learning is a privacy-preserving machine learning method that trains models directly on users’ devices instead of sending raw data to a central server. This approach allows multiple clients to perform local training, and only the model updates are shared with a centralized server.
As a result, personal data remains on the client side, enhancing privacy and security. This method is a significant advancement in protecting user data while enabling machine learning.
“What we are doing here is replacing the centralized server with a blockchain network, which is a distributed and decentralized network,” she said.
This removes the need for a single central server, because traditional federated learning systems can still be vulnerable to certain attacks.
Rouhani added that blockchain can be used to verify participants and improve privacy through encryption and cryptographic methods.
When asked about the benefits, Rouhani noted that “one of them is definitely additional security.”
She included that security and reliability are important because a network with multiple nodes is more stable than a single centralized server.
Additionally, incentivization plays a key role, as more participants are likely to join the network if they can earn rewards by participating in training. Instead of sharing raw data, they train models locally on their data, which then helps improve global training.
“Keep working and engage with the science and technology,” Rouhani said when asked if she had any final messages to share with students. “Just be curious.”
