Blog

Blockchain and Swarm Learning to Improve Big Data IoT Security

29 Oct 2019

In the age of data, the Internet of Things (IoT) and artificial intelligence (AI) technologies are changing just about every industry you can think of. Media may have filled your imagination with apocalyptic images of robots and super AI computers overtaking humans on earth, but the future outlook of such technologies is to continue complementing our work.

Security and privacy issues are major reasons why many hesitate to fully embrace such Big Data tech, so what can we possibly do to ensure more security?

Despite blockchain technology having stirred up global controversy with its popular use for cryptocurrency, it offers a sound structure in sharing encrypted data that could be applied to other tech. Using blockchain as a basis for encrypting information opens up possibilities for other advanced methods, such as swarm learning, to secure big data flows.

What is blockchain?

Blockchain is an encrypted virtual ledger – listing records of transactions that are linked in a ‘chain’ using cryptography. Each ‘block’ in the ‘chain’, includes transaction data, a timestamp and a cryptographic hash of the previous block.

In a private blockchain, only the network admin(s) can grant permissions to participants and control their access (opposed to a public blockchain where anyone can be a validator and send transactions via the Internet). Participants on the network will each have a copy of the blockchain on their own computer meaning it’s decentralised. The structure of blockchain prevents tampering with historical records.

diagram of hash functions in blockchain, data blocks, nonce and new hash value generation

It’s practically impossible for hacker attempts because:

  • each block has its own cryptographic hash and contains one of the previous blocks, and if any of the data is edited then the hash code would change.
  • altering a previous block would break the chain and require updating the hash codes in all following blocks including the copies on all of the other participants’ computers.
  • recalculating hashes would require improbable amounts of computing power and energy.
  • ‘consensus model’ tests/‘proof of work’ are security measures that can be added on, requiring a computer to solve extremely complex mathematical problems in order to be permitted to add a block. The odds of solving one is about 1 in 5.8 trillion! And would take up enormous amounts of energy and computing power.

IoT Security Challenges Can Be Solved With Blockchain

IoT Security (IoTSec) worldwide, faces a future of protecting exponential growth to billions of connected devices and hundreds of zettabytes of data.

  • The global scale of infrastructure to secure IoT devices worldwide would require capabilities to continually identify, authenticate and secure IoT devices;
  • Identifying when sensitive data has been leaked or a device compromised is challenging because device control can be done remotely;
  • It would be very costly and challenging to maintain a centralised security model.
    Implementing on an industrial scale, in the case where edge nodes are spread out internationally would be difficult to manage;
  • Any single point cyberattack or failure to the single model would compromise security, meaning a DDoS attack could easily be carried out.

These key challenges revolve around dependency on a centralised security model, which is why a blockchain-based approach could be a promising solution.

How IoTSec can Leverage Blockchain Tech

Integrating blockchain with IoT could increase the level of encryption used for identifying and authenticating devices with cryptographic hash functions. Any tampering with the defined processes or data would immediately be flagged and all participants in the network could be notified. The processes could also be programmed to automatically stop as soon as there is a breach.

Sensitive data can be allocated unique signatures to safely be shared between IoT devices and controllers using hashes with close to no possibility of being intercepted.

With increased integrity, a secure mesh network can be created to enable reliable connections between IoT devices. Impersonation or spoofing threats are then almost impossible.

Where Blockchain Falls Short for IoT Use

Blockchain may very well be brilliant for strong cryptography and decentralised replication, but such processes increase latencies; which is not practical for services requiring real-time processes and control. For this reasoning, cloud computing could not be solely used for an IoT network depending on real-time communication/responses.

Edge Computing Complements IoT

Edge computing has been developed to complement cloud computing and improve IoT usage by managing data storage and computing physically closer to the required location. It
improves efficiency in bandwidth architecture, increases privacy and regulation of data, and ultimately decreases the occurrence of latency effects.

Edge Computing Complements AI

Current AI models are collecting data at ‘the edge’ and then sending it all to a centralised point e.g. cloud. IoT and AI work together in models where IoT devices/sensors collect raw data which is buffered at the edge and then sent to a centralised database for the AI. With advanced programming, AI can mature and self-learn by recognising patterns and automating processes.

What is Swarm Learning?

Swarm Learning, coined by Dr. Eng Lim Goh, is derived from the concept of swarm intelligence. It combines the use of private blockchain, edge computing and AI. From the illustration below, IoT data is generated and buffered at the edge, then only the learnings from the edge are distributed through the ‘peer network’, similarly to blockchain, in an encrypted form. The raw data is never shared to the cloud and is stored locally for efficient security and broadband connection.

Swarm Learning diagram of edge computing and iot managed with the cloud via blockchain security

The ‘Gravity Concept’

Examples of Swarm Learning can be seen in various ecosystems in our world. From birds that flock together for flying further distances, to fish that swarm together for defences, and bees that swarm to forage efficiently. Individually these small animals may not be the smartest, but when they gather close and collectively work in a systematic way, there are invaluable behaviours to learn.

Using the example of bees, individually each bee may specialise in only single tasks e.g. foraging, scouting, defence, feeding the queen etc. but collectively they move intelligently and make better decisions. For example, with a collective goal to find bountiful food sources some bees will leave the hive and disperse to randomly search for good food sources. Then:

  • Once a bee finds a good food source, it will gather some to return to the hive
  • After unloading its discovery in the hive it will inform the other bees about the source by performing a certain dance.
  • Onlooker bees will watch the dances and help decide on the best food source worth sending employee bees to.

Overall, their intelligent survival processes are based on many individual findings. Each bee does not have to carry out every kind of task and know every detail, rather they share learnings to improve decision-making. Hence, the more bees there are to sense, discover and share information for a queen bee’s hive, the more likely the colony is to prosper.

Swarm Learning for Future IoT and AI Security

In the foreseeable future, it’s clear that we’ll continue using IoT for its many benefits in providing immediate recording and access to data for use on about anything; and use edge computing for security and reduced latency benefits.

IoT models may be satisfactory for small scale situations and our current needs, but if we are to continue working smarter we should leverage self-learning AI to automate processes – saving us time to spend on developing more intelligent and secure AI models.

AI improves as it’s fed more raw data and importantly with humans to program it to specialise in certain processes. To take IoT and AI to the next level, Swarm Leaning can make these necessities, privacy, security and more, incredibly efficient.

With Swarm Learning, latency is significantly reduced so that peers in the network can learn quicker by collaborating in this way and feedback human inference to continue improving the AI. In the future, it’s possible that through swarm learning, we will develop an even more optimal and secure system for Big Data processes. Until then, businesses should not be afraid to leverage these technologies and make beneficial digital transformations. The best we can do is:

  • take extra steps to manage access controls,
  • ensure stakeholders understand what security risks may be involved,
  • and have effective response processes in the case one security is breached.

Looking to smart enable your building, increase occupier engagement or just generally interested in the Smart Spaces offering and would like to know more? Fill in the form below and we will get back to you as soon as possible.


+44(0)20 7481 1655

Browser Update Required

This website does not support your current version of Internet Explorer, Please download the recent version from one of the links provided.

Update to Google Chrome Update to Internet Edge