Research Summary - Convergence of Blockchain, IoT, and AI


The authors provide a non-technical overview of the benefits of blockchain, Artificial Intelligence (AI), and the Internet of Things (IoT), and discuss their complementary properties and interactions.

They present the role of these technologies using a layered framework: IoT is used for data generation, the blockchain provides an infrastructural backbone via immutable smart contracts, and AI can be used to optimize rules and processes.

The authors argue that as these systems intersect, the use of blockchains may prove beneficial. Smart contracts can provide an instrumental mechanism for IoT devices to interact with one another predictably.


Sandner P, Gross J and Richter R (2020) Convergence of Blockchain, IoT, and AI. Front. Blockchain 3:522600. doi: 10.3389/fbloc.2020.522600


Core Research Question

How can the convergence of blockchain and AI make up for the drawbacks of traditional IoT?


AI, IoT, and blockchain are innovations with massive implications for business and industry.

  • Blockchains: increase trust, security, and transparency.
  • IoT: increases automation and user-friendliness of industries.
  • AI: improves processes through pattern detection and outcome optimization

The interconnection between these fields has been neglected as a field of study.

  • In the future, they will converge.

Connection: IoT provides data, blockchains provide infrastructure and rules of engagement, while AI optimizes and processes rules.

Convergence may allow for optimized data management and business process automation.


  1. There are two general storage options for blockchain-based data: on-chain storage and off-chain storage. On-chain data can be restored from full nodes at any time, but storage requirements are significant, which can lead to “blockchain bloating” (i.e. large quantities of on-chain-stored data that hinder throughput and scalability). Off-chain storage stores data off-chain and only keeps aggregated metadata on-chain; it is considerably more scalable than on-chain solutions but decreases data transparency.
  2. Blockchain may improve the infrastructure of IoT devices via smart contracts, especially when it comes to data immutability.
  3. The true potential of AI, IoT, and blockchain may only be accessed if and when combined.
  4. As these technologies intersect, when it comes to data management, the design goal is to improve standardization, enable privacy, and increase the security and scalability of systems where this technology is relevant.
  • A considerable challenge for traditional IoT is the massive amount of data it collects.

  • This data is traditionally stored on centralized servers, which face a lack of standardization, privacy, security, or scalability. Cross-platform data is hard to access and difficult to explain.

  • Blockchain offers access for multiple parties and stores data in one format, increasing interoperability.

  • Another challenge is that data submitted to the cloud is not encrypted and therefore does not ensure privacy.

  • Privacy technologies being pursued in the context of blockchains may lead to better privacy if implemented correctly. If implemented properly in a combined system, this may increase the privacy assurances of IoT devices.

  • Security: Cryptography and Consensus mechanism

    • Trade-off: privacy vs. control of illicit activities.
    • Security may be improved by AI performing data analysis (IoT is a source of big data).
  • AI benefits from big data produced by IoT devices; the larger the quantity of data, the better the AI performance.

  • IoT can’t store big data effectively, but blockchain combined with AI can.

    • Certain blockchains lack scalability but may be increased with different consensus algorithms and with AI.
  1. Data Management: Authentication via a Blockchain-Based Identity
  • Blockchains can manage IoT device identity and authenticate IoT participants on the network.
    • Identity: individuals, companies, devices, or machines
    • Transactions between two entities may be processed efficiently (high speed, low cost).
    • 2025 estimate: more than 20 billion IoT devices will be connected to the internet.
    • Identity management will play an extremely important role.
    • Blockchain allows protection of data, organization of ownership, and monetization of data
    • Identities are secure as blockchain data is extremely difficult to forge
  1. Automatization via Smart Contracts
  • The convergence of blockchain, AI, and IoT is promising for automation.
  • Smart contracts are the main connection between these three innovations.
  • The main limitation of smart contracts is their crypto prerequisite that companies are often unwilling to leverage.
    • Typical crypto value too volatile
    • Stablecoins: unregulated
    • IT and accounting systems are typically based in fiat
    • Solution for smart contracts: fiat-denominated cryptocurrencies that flow through the smart contract, e.g. DLT-based digital Euro
      • IoT devices could make micropayments
      • Enterprise Resource Planning (ERP) systems could record every blockchain transaction
      • Cheap or no cost of currency conversion
      • Compliant with regulatory requirements
      • Issued by banks, e-money institutes, unregulated institutions, or central banks
      • The distinction between central-bank-issued Euro and e-money-provider-issued Euro is important in times of crisis, hence why an official digital currency may be needed.

Discussion & Key Takeaways

The authors argue that convergence of IoT, AI, and Blockchain can and will happen. When it comes, it will begin a new age of digitization. Most critiques of blockchain’s limitations from ten years ago have already been remedied. Scalability limitations may be addressed with newer architectures. There are still challenges that have yet to be overcome.

  • General Data Protection Regulation’s (GDPR) enumerated right for data to be forgotten
  • Integration with legacy systems

The authors argue that these issues will sooner or later be overcome.


The authors provide a case study with a network of IoT lamps. Each lamp has a blockchain-based identity. It operates using a digital Euro. Micropayments made to the lamp allow for it to turn on. AI may leverage the data about the lamps’ usage and automatically perform needed maintenance based on blockchain-provided metrics. It may also optimize downtime by optimizing the maintenance process. Lamps may be tokenized and availed to investors, allowing for building and autonomous maintenance for a network of lamps

Tokenization may be performed for all IoT devices (e.g. sensors, cars, or cameras) that are connected to the Internet and a blockchain network.

Thanks to @Cindy for helping to create this summary.


@Gearlad what was the purpose of the IoT lamp case study? Were they making a case for an autonomous business model – is that what happens at the convergence of IoT, AI, and blockchain? It’s an amazing field, one application being investigated in California is wildfire detection. What kind of work are you doing with IoT, AI and blockchain at NTU?


Neat summary @Gearlad. I’m completely sold on the idea of tokenizing the ownership of IoT devices. It’ll enable a host of niche IoT applications, like the one @jmcgirk mentioned, to gain access to funding. Using blockchains to protect data, manage device identities, and automate business processes (via smart contracts) also makes a lot of sense. However, I’m skeptical of the role AI plays outside of data analytics. What exactly do the authors mean when they claim AI can be combined with blockchain to improve scalability and data storage?

Finally, I didn’t quite understand the point about companies shying away from smart contracts because they are fueled by cryptocurrencies. I get that official DLT-based digital currencies would be less volatile than typical cryptocurrencies and more compliant than stablecoins, but I don’t see how that affects the experience of interacting with a smart contract. As far as I know, governments aren’t building their CBDCs using blockchains [1], let alone the ERC-20 tokens required for Ethereum smart contracts [2]. That means government digital currencies would need to be tokenized (i.e. baked into stablecoins) in order to be used with smart contracts. One way or another companies will have to buy and sell crypto tokens. Am I missing something here? Anyone feel free to correct me If I’m wrong.

[1] D. Shah, R. Arora, H. Du, S. Darbha, J. Miedema and C. Minwalla, “Technology Approach for a CBDC”, Bank of Canada, 2021. [Online]. Available: Technology Approach for a CBDC - Bank of Canada.

[2] N. Reiff, “What Is ERC-20 and What Does It Mean for Ethereum?”, Investopedia, 2020. [Online]. Available: What Is ERC-20 and What Does It Mean for Ethereum?.


This research definitely covers new grounds and makes for a compelling investigation due to its novelty. I personally find the authors’ lamp analogy ambiguous as to its real-world applicability and impact. What are your thoughts? I like the example that @jmcgirk gave here.

With regards to the core research question, it’s notable that IoT devices and AI systems stand to gain through integrating blockchain infrastructure in many respects. Centralized IoT systems have lots of inefficiencies. With the quantity of IoT devices continuing to surge, data inflation can place restrictions on scalability and drive up maintenance overhead, just to name a few of the problems [2]. Blockchain systems do offer decentralization that can remedy nuances that arise in centralized IoT systems, but I side with the authors’ position - merging just these two technologies will not grant optimization. Something more is needed.

IoT-blockchain systems satisfy the need for transparency, accountability, secure transactions, lower costs (central auditors are essentially taken out of the picture), etc. Throughput still remains an issue, though. AI addresses this concern by automating processes and enabling self-sufficient models. According to Sandner, Gross, and Richter, AI algorithms advance through data availed by IoT [1]. Further, “the more data is used to train the AI algorithm, the better the performance of the algorithm” [1]. Based on this, the more that IoT expands, the better AI is able to meet growing demands, increase throughput, maximize efficiency, and fuel economic growth.

It seems that the three technologies will inevitably exist as symbiotic cohorts. Still, why is there widespread hesitance to adopt new frameworks like these if the benefits ultimately outweigh drawbacks? In your summary, you state that “smart contracts are the main connection between these three innovations.” Can you elaborate on this?

[1] P. Sandner, J. Gross, and R. Richter, “Convergence of Blockchain, IoT, and AI.” Frontiers in Blockchain, vol. 3, pp. 1-5, Sep. 2020, doi: 10.3389/fbloc.2020.522600. [Online]. Available:
[Accessed Jul. 22, 2021].

[2] S. Verma. “How blockchain and IoT is making supply chain smarter.” IBM. Available: [Accessed Jul. 22, 2021].


It’s great to point to convergence AI, IoT, Blockchain. Blockchain is like a bridge to connect IoT and AI. I also believe in the future is coming soon. I think all the technical solutions are provided already to companies. As the summary mention that

I read through one of the board of governors of the federal reserve system thinks about stablecoins from this paper Taming Wildcat Stablecoins. This paper point out three commnets:
First, the use of private bank notes was a failure because they did not satisfy the no-questions-asked (NQA) principle.
Second, the U.S. government took control of the monetary system under the National Bank Act and subsequent legislation in order to eliminate the private bank note system in favor of a uniform currency—namely, national bank notes.
Third, runs on demand deposits only ended with deposit insurance in 1934.

I think the most difficult part is how to convince the rule makers.


The author is convinced with the idea of the Convergence of Blockchain, IoT, and AI, and his lamp analogy is interesting and visionary. The definition of IoT is a system of interrelated computing devices or machines that are provided with unique identifiers (UIDs) equipped with the ability to transfer data over a network without requiring human-to-human or human-to-computer interaction. According to the Internet, there are 2.5 quintillion bytes of data being created every day, it is expected that the volume of data is going to double every two years. Additionally, we are entering the area of AIOT, therefore; the estimation of the data generated every other year will be one time more might be underestimated. Trained with such a huge amount of data, artificial intelligence will undoubtedly thrive, but it’s impossible to store the volume of data on this scale with traditional method; left alone each device should be equipped with the ability to communicate with each other and make the appropriate decision in time, which is the essence of AIOT. This is when blockchain comes into play, not only because of the scalability it provides, but also most importantly its immutability feature.

Just imagine years after now, you’re sitting in an autonomous car, each car can acquire data via ways of means lidar, traffic data, communication with others vehicles, etc., and then make the best decision to arrive at the destination in the shortest time. The communications between agents become essential, however, it also becomes a vulnerability of the framework. If someone sends the target car some fake traffic data making the vehicle takes a detour, therefore they can hijack it. This scenario points out blockchain is an inevitable part of the development of AIOT.


The IoT lamp case was actually just an example to show that truly any arbitrary IoT device may leverage the proposed AIoT framework in a beneficial way.
I personally feel that the authors could have provided more use cases. The one you describe in this article, PANTHER, is great - in which wildfires are detected by an artificial nervous system, of sorts, where an assortment of sensors - heat, humidity, wind, fuel, and cameras - are connected in a node that connects to the cloud, the data of which can subsequently be used by machine learning tools.

Our team of Deane @Albert @fmendoz7 @TurtleHead @eleventh and @Cindy are now in the beginning stages of our AIoT Blockchain research project. I plan to work on the research summaries “Multi-Layer Aggregate Verification for IoT Blockchain” and “Efficient Attribute-Based Smart Contract Access Control Enhanced by Reputation Assessment” with Cindy. Albert is working on the summary for “SoK: Applying Blockchain Technology in Industrial Internet of Things”. Francis is tackling the summary “An analysis and evaluation of lightweight hash functions for blockchain-based IoT devices”.

Teamwork makes the dream work!


The technical questions of crypto that have come about and that are bound to come in the future will always eventually have a solution. But answering this-

-remains the key in the future of crypto!


What are some immediate business use cases for the convergence of IoT, blockchain and AI? @Albert @fmendoz7 @TurtleHead @eleventh and @Cindy – would this be a military technology primarily? I can imagine swarming drone detection lines or smart mines. What other things come to mind?


There are already some examples that startup companies are applying AIoT and blockchain technology to the agriculture industry. With the IoT sensors, producers can gather data on a range of metrics and send back information for decision-making. Using the data, artificial intelligence (AI) can improve the growing and selling processes with the parameters put in their model, help farmers determine which crops to grow to maximize the profit or anticipate potential threats by cross-fitting with historical information. Then, the producer can use robots to get rid of the weak plants to create more space for the healthy plants and deploy drone spreading capsules with the eggs of natural enemies targeting certain types of pests before they cause damage. Meanwhile, the data will be uploaded to the blockchain, creating secure, transparent records, which is valuable for the agriculture industry to create smart contracts tracking the food from its origins to grocery stores. The data can also be uploaded to the blockchain via oracle to create an insurance smart contract. Farmers can purchase agricultural derivatives. If the condition goes south, the derivative will compensate for the losses.


Although the trend is just starting, it’s guaranteed that it will gain popularity in the recent future.


One good application scenario might be with surveillance systems that use cameras or motion detectors in an IoT network. All data is uploaded and stored (at least for a specified amount of time) to a database (off-chain storage). Suspicious activity would be detected by AI and flagged data could be uploaded as on-chain storage for both immutability and a longer-term (perpetual) storage. Essentially in this case we have AI to select which data goes onto the blockchain.


Smart contracts are what governs autonomous processes in an AIoT blockchain system. The authors describe a pay-per-use scenario in which, based on the terms of the smart contract, lamp(s) in the network turn on for a specified time after receiving a micro-payment. Moreover, the lamp’s system status data such as power or time of usage may be uploaded to the blockchain and AI could use the uploaded data to optimize business processes (periodic maintenance, minimization of downtime, et cetera).

I personally think that a better example than the authors gave would be in the process of renting a car. In the same way, each car would then have its own wallet and micropayments would allow for it to be driven for a specified amount of time. Due to a constant stream of data uploaded to the blockchain, if the driver speeds or if for example one of the parts malfunctions, the timestamp of when this event occured could be used to prove misuse or misconduct.


That’s a great point on tokenization and shared ownership. More and more infrastructure in an AIoT blockchain network could be publically owned, rented out for daily use, and auto-monitored/regulated.
In terms of AI increasing blockchain scalability, the authors cite a case from another paper titled “Performance optimization for blockchain-enabled industrial internet of things (IIoT) systems: a deep reinforcement learning approach”. In this paper, a DRL-based system allows for higher level of throughput by using dynamic selection of certain blockchain parameters (block producers, block size, consensus algorithm, and block interval).


This is another great example. Decisions made by AIoT devices do need to be explainable, especially when there are such immense safety concerns as in the case of autonomous vehicles. We can have blockchain to prove specific data existed and to justify the AI’s course of action. This could then be accessed by law enforcement entities or car insurance companies, for example.


I really enjoyed this summary, I think this convergence is going to have huge implications on the future of industry. The fact that it will behoove a business to have every aspect be connected via IoT from the whole supply chain to the coffee maker in the break room and be logged, evaluated, and optimized.

I like to think the more and more sensors added are only going to give further insights and novel discoveries of how to benefit from the system whether it be large or small. For example, on a global scale, the more data we have about the climate will crack open new models that will help to predict what might happen better than something we might have never been able to predict ourselves/without AIoT.

These researchers are looking into Wave-induced Atmospheric Variability and have only recently been able to get some interesting results due to the availability of more sensors/data.

Systems using blockchain data will be able to make leaps and bounds as they can communicate effectively and efficiently interesting correlations due to the standardized datasets.

It is only a matter of time before blockchain AIoT is going to take over as they are perfect for each other.


Hey @willcon, thanks for the comment. I just checked out the WAVE project. Climate prediction is just another great example of what AIoT is capable of. It just so happens that these three innovations - AI, IoT, and blockchain - are set to change everything. The converged fields of AIoT and IoT Blockchain have already become major fields of research. In the same way, the converged fields of AI Blockchain and AIoT Blockchain are likely to become major topics of future research.

Incidentally on some news, just today Facebook released its first set of smart glasses (Ray-Ban Stories), another great example of AIoT technology. I’ve pre-ordered Dr. Kai-Fu Lee’s book “AI 2041: Ten Visions for Our Future” which holds some astounding predictions for the future. One interesting one is that AR and VR technologies will allow for virtual and physical meetings to become practically indistinguishable from each other.


This has been a great thread to read - a lot of research has focused on more incremental improvements and points regarding to mechanism or design, but real world use cases are not isolated, research that focuses on connecting the dots and utilizing all elements of emerging tech such as the articles cited in this thread will be the applications we’ll see.

A couple points that may be of interest to this thread

  • Helium, a telecom based project may be a good example of connecting IoT with blockchain. Traditionally IoT networks have been difficult to develop because of the startup and incentive costs to maintain a network, Helium has used blockchain economics and incentive structures to help with this (miners and validators process and host their local network for IoT device in exchange for token rewards)

  • Other shifts to think about that all are developing in parallel - the 5G rollout will enable far higher throughput of information transfer / compute and lower latency and the transition onto the Cloud for most applications will complement AI / IoT use cases (and many would argue is the infrastructure needed to enable the next generation of industrial or city-wide AI/IoT)

  • Although the use cases of blockchain are exciting, one small critique in the paper would be that the researchers don’t exact answer the question of why a blockchain (as opposed to just a normal database) would be the best solution to some of these problems. We’re seeing for example the development of private LTE networks for smart cities or for utilities to monitor their service deliver with IoT, with blockchain potentially being a slower and more expensive solution due to consensus needs and node hosting. One would argue with the need for security that instead of a public blockchain a permissioned blockchain would be utilized, but at that point the line blurs between what one may define a blockchain to be

An interesting paper regarding 5G and Blockchain is this one here for anyone interested in a read:


Great summary. The topic brought by this paper is interesting yet somewhat vague. Despite all the difficulties yet to be overcome, there are still a lot of research potentials with systems that already exist. Introducing (new) IoT devices combining existing Blockchain and AI systems could be a great commercial achievement; Pulishing a recommandation map for different IoT devices to different Blockchain protocols using different AI applications in between along with all the tradeoff discussed could be a great academic achievement. It’s good to know inspiring paper like this one.


Very nice summary.

Hence reading through the comments ebbs and flows, I can’t help but to ask whether “standardizing” incumbent IoT data and providing “authenticity” would be the appropriate approach to automate data collection and preprocessing for AI applications. Forcefully making end users or consumers abide by a specific blockchains Smart contract might as well de-incentivize adoption of blockchain technology on the IoT layer of data collection.

In my opinion, blockchain can provide data authenticity and basic preprocessing via the Smart contract and that would be a huge advantage for streamlining the end to end of an AI applications development. Along side with block chain providing a more secure decentralized but also centralized data infrastructure, at the end of the day, providers of such blockchain infrastructure are still there to make money. Denying a means to obtain data and enforcing a more complex interfacing by the end user would be the life and death of said budding blockchain maintainers.

Then there is the issue of what it truly means to be “on chain”. With such large volumes of inflow, inevitably, there needs to be a form of a single source of truth. To my knowledge, a purely on chain approach to that “single source of truth” is yet to be available. I know of several Graph based technologies that may facilitate such “single source of truths” but even then, there are still drawbacks and compromises.

I also do concur that there needs to be further improvement on the existing network layer architectures. Lightning network for bitcoin, cardano being quite dismissive on doing anything more than networking on chain, and IOTA tangle being an aspiring resolution to the “on chain” data scalability issue.

I have this presentation for a past partnership proposal I tried to pitch to a blockchain company on the integration of Data science and Blockchain for aspiring Paas companies.

Please do tell me your thoughts about it.