AI-Powered Disease Prevention Without Privacy and Security Trade-offs

July 30, 2025

Federated learning enables AI to help predict disease outbreaks by training models on hospital data without compromising patient privacy. This technique keeps sensitive health information secure while building powerful disease surveillance systems for better public health responses.

Rainy seasons in Brazil bring more than floods and humidity. They are often associated with dengue, a mosquito-borne illness that causes extremely high fevers, rash and can lead to death.

More than 18 million Brazilians have contracted dengue over the past 25 years, but in 2024, cases of the disease skyrocketed — with 6 million cases and 4,000 deaths by June of that year. It’s not just Brazil that is affected by dengue. About half the world’s population faces risks from dengue, with an estimated 100 million to 400 million cases each year, according to the World Health Organization.

Since the COVID-19 pandemic, researchers around the world have turned to artificial intelligence to monitor potential disease outbreaks. But those efforts face a major hurdle: access to sensitive health data. Federated learning enables hospitals and agencies to build predictive models without sharing patient data.

IEEE Member Márcio Teixeira is part of the Dengue Alert project, which aims to train machine learning models capable of predicting dengue severity, identifying outbreak patterns and optimizing resource allocation.

Here, Teixeira discusses federated learning, and how it’s used to prevent outbreaks of dengue.

What is federated learning and how is it different from other AI techniques?

Federated learning is a technique that allows researchers to train AI models without moving the training data from its original location. For example, imagine several hospitals or health centers, each with data about dengue patients. A shared AI model is sent to each hospital. Then, the model is trained locally using only the data available at that hospital. It then sends back only the learning (the model’s updates), and these updates are aggregated to improve the global model. This process is repeated several times until the global model is fully trained — without ever accessing the actual patient data.

Traditional AI gathers all data in one place to train the model — which can risk privacy. Federated learning keeps data where it is, in this case hospitals, and trains models locally. This makes federated learning better suited for sensitive data.

What motivated your research in this area?

Our motivation came from the urgent need to improve dengue surveillance and treatment using innovative and privacy-preserving technologies. Dengue remains a major public health issue, especially in regions like Brazil, where outbreaks cause significant strain on health care systems.

Building accurate predictive models requires access to sensitive medical data, which poses ethical and legal challenges. Brazil and many other countries have laws that strictly regulate how health data can be used, stored and shared.

Health data is extremely valuable because it contains personal information, such as full names, birth dates, addresses, medical history and other information. If criminals have access to this data, they can steal the patient’s identities and use it for billing scams or fake treatments, create fake insurance claims or launch targeted phishing attacks.

How could AI help predict and prevent dengue outbreaks, and why is this disease a good test case for this technology?

AI can analyze large amounts of health, environmental and behavioral data to identify patterns that humans might miss. In the case of dengue, machine learning might be used to identify high-risk patients early and predict outbreaks based on weather conditions because mosquitoes use pools of standing water to breed.

What other ways could federated learning protect our privacy while still giving us better technology?

Federated learning offers a new way to develop smart technologies without compromising user privacy. Beyond healthcare, it’s used in a number of sectors. For example, banks can collaborate to detect fraud patterns without exposing customer records. It’s also used on your smartphone. Your phone learns locally, but it contributes to a global model, without messages leaving your device.

Learn More: Digital medical devices enable more personalized medical treatments and the hope of improved health outcomes. But they can be a target for cybercriminals, especially when wirelessly connected to the internet. A new IEEE standard, recognized by the FDA, helps device-makers identify and counter relevant threats and describes the necessary requirements of a wireless security program. Check out this blog from the IEEE Standards Association to learn more.