Author:
Adam Duke, stacuity Developer

IoT – the Internet of Things – is a broad term meaning a device that isn’t a human that needs connectivity. The number of IoT devices continues to increase as we use technology to unlock new capabilities. AI allows IoT devices – and the systems they communicate with – to be more intelligent and better able to understand the world. The combination of IoT and AI will allow us to solve some of our greatest challenges with applications across every industry.

New opportunities

In agriculture, AI is powering computer vision models able to detect crop health, helping to reduce food waste. In medicine, AI models are now used to diagnose diseases remotely and to discover new classes of drugs and treatments. The list of applications is endless: AI truly gives the human race new superpowers and has been called the new electricity.

AI can be applied to create IoT solutions in many different ways, including uploading sensor data to the cloud for later processing by machine learning models, or directly running models on-device to enable things like object, face or voice recognition.

Developing a new AI model to solve a specific problem usually requires immense computing power and energy. However, once a model is trained it can be deployed to run with much less compute power and energy. The process of asking questions of an already-trained AI model is known as inference.

For IoT devices, running an AI model (inference) usually works in one of two ways: the AI model runs in the cloud, or on-device (also known as Edge AI).

For remotely located devices, cloud-based models may not work due to their reliance on connectivity, which may not be guaranteed. For example, if your voice assistant stops working when you are low on reception then that’s not a great experience. However, as small, low-powered computing devices become more powerful and gain the ability to run AI models locally then the shift towards on-device models will gain momentum. A mixture of the two different approaches can also be used where on-device models are combined with more powerful models running in the cloud.

As AI and IoT become more commonplace, the ability to securely update models running at the edge and provide stable connectivity to cloud services is a must. The stacuity core network aims to help solve some of these problems and simplify and empower these new opportunities in AI and IoT.

Security concerns

We believe that the recent press surrounding the dangers of AI is likely overblown. Andrew Ng, one of the foremost experts in the field of AI and machine learning, recently gave a more positive view on the future impact of AI.

  • To watch the full TED talk follow this link

However, as with any new technology there are some remaining concerns, which cannot be ignored.

For example, AI may allow bad actors to create new tools, specifically engineered to disrupt, attack and cause harm. Future AI bots and agents may be able to learn from all known existing network weakness and exploits and formulate new and previously unseen exploits and strategies. This may pose an increasing threat to the network security of IoT and cloud services.

Against this backdrop, new counter-offensive AI models and systems may be used to detect attacks in real-time and even provide counter-attack capability.

What can we do to help mitigate the risks?

Network isolation
The ability to isolate devices at a network level – potentially disconnected from the internet – allows for the attack surface to be much reduced.

The stacuity core network allows devices to belong to discreet network VSlices with isolated boundaries of communication. By utilising a strategy of network segregation, it is possible to keep the ‘internet of bad actors’ – including any future threats – away from IoT devices and the systems they rely on. Traditional core networks utilise APNs to achieve this layer of network segregation, which can be costly to set up and configure. In contrast, the stacuity network can be configured instantly on a per-device basis.

Logging and Observability
To detect new and novel attacks it is essential to have the ability to inspect, log and and respond to attacks as they happen. AI systems, such as those provided by Darktrace, or Cloudflare can be integrated into network pipelines to observe, analyse and detect unusual traffic patterns before they present an issue. Equally, the ability to observe live data usage patterns also helps to discover and mitigate DDoS-style attacks.

Towards the future

At stacuity, we see AI as an extremely important and transformational technology. The combination of IoT and AI has enormous potential to solve many of the world’s problems and we look forward to helping our customers use the stacuity core network platform to build new and exciting solutions.

Discover more

To learn more about Stacuity’s unique software defined mobile core network and how it can address many of the challenges discussed reach out to Sales@stacuity.com

Published On: March 18th, 2024 / Categories: Blog / 4.3 min read /

Want to find out more? 

Our mission is empowering IoT connectivity providers and customers with control of mobile connectivity, so that they can build a better future with IoT technology.