AI in OT: Opportunities and risks you need to know

Artificial intelligence (AI), significantly generative AI apps corresponding to ChatGPT and Bard, have dominated the information cycle since they turned broadly obtainable beginning in November 2022. GPT (Generative Pre-trained Transformer) is commonly used to generate textual content educated on giant volumes of textual content knowledge.

Undoubtedly spectacular, gen AI has composed new songs, created photos and drafted emails (and rather more), all whereas elevating reputable moral and sensible considerations about the way it may very well be used or misused. However, when you introduce the idea of gen AI into the operational know-how (OT) house, it brings up important questions on potential impacts, how to greatest check it and how it may be used successfully and safely. 

Impact, testing, and reliability of AI in OT

In the OT world, operations are all about repetition and consistency. The objective is to have the identical inputs and outputs in order that you can predict the end result of any scenario. When one thing unpredictable happens, there’s at all times a human operator behind the desk, prepared to make selections shortly primarily based on the potential ramifications — significantly in important infrastructure environments.

In Information know-how (IT), the implications are sometimes a lot much less, corresponding to dropping knowledge. On the opposite hand, in OT, if an oil refinery ignites, there’s the potential value of life, adverse impacts on the surroundings, important legal responsibility considerations, in addition to long-term model harm. This emphasizes the significance of creating fast — and correct — selections throughout instances of disaster. And that is in the end why relying solely on AI or different instruments just isn’t good for OT operations, as the implications of an error are immense. 

AI applied sciences use loads of knowledge to construct selections and arrange logic to present applicable solutions. In OT, if AI doesn’t make the suitable name, the potential adverse impacts are severe and wide-ranging, whereas legal responsibility stays an open query.

Microsoft, for one, has proposed a blueprint for the general public governance of AI to tackle present and rising points via public coverage, regulation and regulation, constructing on the AI Risk Management Framework not too long ago launched by the U.S. National Institute of Standards and Technology (NIST). The blueprint requires government-led AI security frameworks and security brakes for AI programs that management important infrastructure as society seeks to decide how to appropriately management AI as new capabilities emerge.

Elevate pink group and blue group workout routines

The ideas of “red team” and “blue team” refer to completely different approaches to testing and bettering the safety of a system or community. The phrases originated in army workout routines and have since been adopted by the cybersecurity group.

To higher safe OT programs, the pink group and the blue group work collaboratively, however from completely different views: The pink group tries to discover vulnerabilities, whereas the blue group focuses on defending towards these vulnerabilities. The objective is to create a sensible state of affairs the place the pink group mimics real-world attackers, and the blue group responds and improves their defenses primarily based on the insights gained from the train.

Cyber groups might use AI to simulate cyberattacks and check ways in which the system may very well be each attacked and defended. Leveraging AI know-how in a pink group blue group train can be extremely useful to shut the abilities hole the place there could also be a scarcity of expert labor or lack of funds for costly sources, and even to present a brand new problem to well-trained and staffed groups. AI might assist establish assault vectors and even spotlight vulnerabilities that will not have been discovered in earlier assessments. 

This kind of train will spotlight numerous ways in which may compromise the management system or different prize property. Additionally, AI may very well be used defensively to present numerous methods to shut down an intrusive assault plan from a pink group. This might shine a lightweight on new methods to defend manufacturing programs and enhance the general safety of the programs as a complete, in the end bettering total protection and creating applicable response plans to shield important infrastructure. 

Potential for digital twins + AI

Many superior organizations have already constructed a digital duplicate of their OT surroundings — for instance, a digital model of an oil refinery or energy plant. These replicas are constructed on the corporate’s complete knowledge set to match their surroundings. In an remoted digital twin surroundings, which is managed and enclosed, you might use AI to stress check or optimize completely different applied sciences.

This surroundings gives a protected manner to see what would occur if you modified one thing, for instance, tried a brand new system or put in a different-sized pipe. A digital twin will permit operators to check and validate know-how earlier than implementing it in a manufacturing operation. Using AI, you might use your individual surroundings and info to search for methods to enhance throughput or reduce required downtimes. On the cybersecurity facet, it gives extra potential advantages. 

In a real-world manufacturing surroundings, nonetheless, there are extremely giant risks to offering entry or management over one thing that may consequence in real-world impacts. At this level, it stays to be seen how a lot testing in the digital twin is enough earlier than making use of these adjustments in the actual world.

The adverse impacts if the check outcomes will not be fully correct might embody blackouts, extreme environmental impacts and even worse outcomes, relying on the trade. For these causes, the adoption of AI know-how into the world of OT will possible be sluggish and cautious, offering time for long-term AI governance plans to take form and threat administration frameworks to be put in place. 

Enhance SOC capabilities and reduce noise for operators

AI will also be used in a protected means away from manufacturing gear and processes to help the safety and progress of OT companies in a safety operations middle (SOC) surroundings. Organizations can leverage AI instruments to act nearly as an SOC analyst to evaluate for abnormalities and to interpret rule units from numerous OT programs.

This once more comes again to utilizing rising applied sciences to shut the abilities hole in OT and cybersecurity. AI instruments may be used to reduce noise in alarm administration or asset visibility instruments with really useful actions or to evaluate knowledge primarily based on threat scoring and rule constructions to alleviate time for employees members to give attention to the best precedence and biggest influence duties.

What’s subsequent for AI and OT?

Already, AI is shortly being adopted on the IT facet. That adoption might also influence OT as, more and more, these two environments proceed to merge. An incident on the IT facet can have OT implications, because the Colonial pipeline demonstrated when a ransomware assault resulted in a halt to pipeline operations. Increased use of AI in IT, due to this fact, might trigger concern for OT environments. 

The first step is to put checks and balances in place for AI, limiting adoption to lower-impact areas to make sure that availability just isn’t compromised. Organizations which have an OT lab should check AI extensively in an surroundings that isn’t linked to the broader web.

Like air-gapped programs that don’t permit exterior communication, we need closed AI constructed on inner knowledge that is still protected and safe inside the surroundings to safely leverage the capabilities gen AI and different AI applied sciences can supply with out placing delicate info and environments, human beings or the broader surroundings in danger.

A style of the long run — right this moment

The potential of AI to enhance our programs, security and effectivity is sort of limitless, however we need to prioritize security and reliability all through this fascinating time. All of this isn’t to say that we’re not seeing the advantages of AI and machine studying (ML) right this moment. 

So, whereas we need to concentrate on the risks AI and ML current in the OT surroundings, as an trade, we should additionally do what we do each time there’s a new know-how kind added to the equation: Learn how to safely leverage it for its advantages. 

Matt Wiseman is senior product supervisor at OPSWAT.