How will the emergence of 5G affect AI federated learning?

As development teams scramble to build AI tools, training algorithms on edge devices is becoming more common. Federated learning, a subset of distributed machine learning, is a relatively new approach that allows companies to improve their AI tools without explicit access to raw user data.

[[330792]]

Federated learning, conceived by Google in 2017, is a decentralized learning model through which algorithms can be trained on edge devices. Regarding Google's "on-device machine learning" approach, the search giant pushes its predictive text algorithm to Android devices, aggregates the data, and sends a summary of the new knowledge back to a central server. To protect the integrity of user data, this data is passed through homomorphic encryption or differential privacy, which is the practice of adding noise to the data to obscure the results.

In general, through federated learning, AI algorithms can be trained without specific data that identifies any individual user. In fact, the raw data never leaves the device itself, only aggregated model updates are sent back. These model updates are then decrypted after delivery to the central server. Then, a test version of the updated model is sent back to the selected devices. After repeating this process thousands of times, the AI ​​algorithm is significantly improved, while never compromising user privacy.

This technology is poised to make waves in the healthcare space. For example, medical startup Owkin is currently exploring federated learning. In order to leverage patient data from multiple healthcare institutions, Owkin uses federated learning to build a C algorithm with data from different hospitals. This could have far-reaching implications, especially since hospitals will be able to share disease progression data with each other while maintaining the integrity of patient data and complying with HIPAA regulations, which is invaluable. Healthcare is by no means the only sector adopting this technology; federated learning will increasingly be used by self-driving car companies, smart cities, drones, and fintech organizations. Several other federated learning startups are set to go public, including Snips, S20.ai, and Xnor.ai, the latter of which was recently acquired by Apple.

Potential Problems

Man-In-The-Middle Attacks

Given that these AI algorithms are worth a lot of investment, it is expected that these models will become lucrative targets for hackers. Nefarious hackers may attempt to conduct man-in-the-middle attacks. However, as mentioned earlier, by adding noise and aggregating data from various devices and then encrypting this aggregated data, companies may make it difficult for hackers to do this.

Model Poisoning

Perhaps more worrisome are attacks that poison the model itself. A hacker could conceivably compromise the model, either through their own device or by taking over the devices of other users on the network. Ironically, because federated learning aggregates data from different devices and sends an encrypted summary back to a central server, hackers entering through a backdoor are somewhat masked. As a result, it is difficult, if not impossible, to identify the location of the anomaly.

Bandwidth and processing limits

While on-device machine learning effectively trains algorithms without exposing raw user data, it does require a lot of local power and memory. Companies try to circumvent this by only training their AI algorithms on the edge when the device is idle, charging, or connected to Wi-Fi; however, this is an everlasting challenge.

The impact of 5G

As 5G expands around the world, edge devices will no longer be limited by bandwidth and processing speed limitations. According to a recent Nokia report, 4G base stations can support 100,000 devices per square kilometer. The upcoming 5G base stations will support up to 1 million devices in the same area. Through enhanced mobile broadband and low latency, 5G will provide energy efficiency while facilitating device-to-device communications (D2D). In fact, it is predicted that 5G will bring a 10-100 times increase in bandwidth and a 5-10 times reduction in latency.

As 5G becomes more prevalent, we will experience faster networks, more endpoints, and a larger attack surface, which could attract an influx of DDoS attacks. 5G also features slicing capabilities, which will allow slices (virtual networks) to be easily created, modified, and deleted based on user needs. It remains to be seen whether this network slicing component will alleviate security concerns or bring about a host of new problems, according to a study on the disruptive power of 5G.

In summary, new concerns arise from a privacy and security perspective; however, the fact remains: 5G is ultimately a boon to federated learning.