AI Research

Sandia National Laboratories’ Research Team Develop AI Algorithms to Detect Physical Issues, Cyberattacks within Grid

Published

4 days ago

September 11, 2025

Researchers at Sandia National Laboratories have developed AI algorithms to detect physical problems, cyberattacks and both at the same time within the grid.

“As more disturbances occur, whether from extreme weather or from cyberattacks, the most important thing is that operators maintain the function and reliability of the grid,” said Shamina Hossain-McKenzie, a cybersecurity expert and leader of the project. “Our technology will allow the operators to detect any issues faster so that they can mitigate them faster with AI.”

Adrian Chavez, a cybersecurity expert involved in the project, added that as the neural network runs on single-board computers, or existing smart grid devices, it will be able to protect older equipment as well as the latest equipment lacking only cyber-physical coordination.

The package of code works at the local, enclave and global levels. At the local level, the code monitors for abnormalities at the specific device where it is installed, at the enclave level, devices in the same network share data and alerts to provide the operator with better information on whether the issue is localized or happening in multiple places and at the global level, only results and alerts are shared between systems owned by different operators.

Thus, operators will receive early alerts of cyberattacks or physical issues their neighbours are seeing and protect proprietary information. The Sandia team collaborated with experts at Texas A&M University to create secure communication methods, between grids owned by different companies, Hossain-McKenzie added.

According to Logan Blakely, a computer science expert who led development of the AI components, the challenge in detecting cyber-physical attacks is combining the constant stream of physical data with intermittent packets of cyber data.

Physical data such as the frequency, voltage and current of the grid is reported 60 times a second, while cyber data such as other traffic on the network is more irregular, Blakely said. The team used data fusion to extract the important signals in the two different kinds of data.

The team also used an autoencoder neural network, which classifies the combined data to determine if it fits with the pattern of normal behavior or there are abnormalities with the cyber data, physical data or both, Hossain-McKenzie said. For example, an increase in network traffic is predicted to indicate a denial-of-service attack while a false-data-injection attack can include atypical physical and cyber data, Chavez added.

Autoencoder neural networks do not need to be trained on data labelled with every type of issue coming up as compared to other kinds of AI, Blakely said. However, the network only requires huge amount of data from normal operations for training.

After the team constructed the autoencoder neural network, they put it to the test in three different ways. First, they tested the autoencoder in an emulation environment, which includes computer models of the communication-and-control system used to monitor the grid and a physics-based model of the grid itself, Hossain-McKenzie said.

The team used the environment to model a variety of cyberattacks or physical disruptions, and to provide normal operational data for the AI to train on. The collaborators from Texas A&M University assisted with the emulation testing.

Secondly, the team incorporated the autoencoder onto single-board computer prototypes that were tested in a hardware-in-the-loop environment, Hossain-McKenzie said. In hardware-in-the-loop testing, researchers connected a real piece of hardware to software simulating various attack scenarios or disruptions.

When the autoencoder is on a single-board computer, it can read the data and implement the algorithms faster than a virtual implementation of the autoencoder can in an emulation environment, Chavez said. Hardware implementations are a hundred or thousand times faster than software implementations, he added.

The team is working with Sierra Nevada Corporation to test Sandia’s autoencoder AI on the company’s existing cybersecurity device called Binary Armor, Hossain-McKenzie said.

The team is testing both formats, single-board prototypes interfaced with the grid and the AI package on existing devices, in the real world at the Public Service Company of New Mexico’s Prosperity solar farm as part of a Cooperative Research and Development Agreement, Hossain-McKenzie said. The tests started in summer 2024, Chavez said.

The project expanded upon a previous R&D 100 Award-winning project called the Proactive Intrusion Detection and Mitigation System, which focused on detecting and responding to cyber intrusions in smart inverters on solar-panels, Hossain-McKenzie said. The team is expanding upon the autoencoder AI in similar projects, she added.

The team filed a patent on the autoencoder AI and is looking for corporate partners to deploy and hone the technology in the real world, Hossain-McKenzie said. The autoencoder is expected to be used to protect other critical infrastructure systems such as water and natural gas distribution systems, factories, even data centers, Chavez said.

The project is funded by Sandia’s Laboratory Directed Research and Development program.

Source link

AI Research

Reuters hires Seetharaman to cover artificial intelligence

Published

19 minutes ago

September 15, 2025

Chris Roush

Reuters global tech editor Kenneth Li sent out the following on Monday morning:

All,

I’m very pleased to announce that Deepa Seetharaman is returning to Reuters as a Tech Correspondent, based in San Francisco.

For Deepa, this marks a homecoming. She began her career at Reuters in New York and covered the U.S. autos in Detroit before moving to San Francisco to report on Amazon, building a reputation for breaking news and delivering ambitious stories at the heart of America’s biggest companies. She went on to spend a decade at the Wall Street Journal, where she covered some of the most consequential developments in technology, politics, and society.

At the Journal, Deepa was the lead reporter on Facebook (now Meta), where her coverage explored the company’s business, culture, and influence. Her reporting included coverage of Instagram’s impact on teenage girls and investigations into how AI systems falter in moderating racist and hateful content. More recently, she turned her focus to artificial intelligence, chronicling how advances in the technology are reshaping business models, political discourse, and cultural norms.

At Reuters, Deepa will focus on AI and OpenAI at a time when the technology is at an inflection point. With breakthroughs harder to achieve and investors pressing for returns, her work will span cutting-edge research, the strategies of the most powerful tech companies, and the global implications of AI’s rise. She will report to me and work closely with our global technology team as well as Steve Stecklow and the enterprise team. Her return also reunites her with Jeff Horwitz, who joined our San Francisco bureau in June. She starts today.

Deepa’s work has earned some of journalism’s most prestigious awards. She was part of a team that won the George Polk Award for Business Reporting and the Gerald Loeb Award in Beat Reporting.

Please join me in welcoming Deepa back to Reuters.

Ken

Source link

AI Research

How to Turn Early Adoption into ROI

Published

55 minutes ago

September 15, 2025

Christine Livingston

To realize AI’s full potential, organizations must be in it for the long game; a pursuit that requires patience, persistence, and strategic alignment. While quick wins are important, they won’t stand alone in delivering meaningful value; agile experimentation is a necessity, execution requires iteration, and early challenges are inevitable.

Protiviti’s inaugural global AI Pulse Survey highlights a compelling correlation between AI maturity and return on investment (ROI) as well as a disconnect between expectations and performance for many organizations in the early stages of AI adoption. The survey, which had more than 1,000 respondents, categorizes organizations from more than a dozen industry sectors into five maturity stages:

Stage 1: Initial — Recognizing AI’s potential but lacking strategic initiatives.

Stage 2: Experimentation — Running small-scale pilots to assess feasibility.

Stage 3: Defined — Integrating AI into business processes.

Stage 4: Optimization — Enhancing performance and scalability with data feedback.

Stage 5: Transformation — AI drives significant business transformation.

Expectations from AI Investments

As organizations progress through these stages, their satisfaction with AI investments improves. In fact, of the 50% of survey respondents who indicated that they are in the early stages (initial or experimentation) of AI adoption, about 26% reported that AI investment returns fell below expectations.

Of course, not all AI experimenters are experiencing poor returns. Indeed, a majority report ROI meeting expectations, but the results showed a higher concentration of slightly exceeded or significantly exceeded ROI expectations among groups in the middle to advanced stages of AI adoption.

In reviewing what differentiates successful experimenters — those in the experimentation stage of AI adoption who reported exceeding ROI expectations — from those that did not, we find three compelling attributes:

Focus on balanced key performance indicators (KPIs) and measuring success using a mix of financial and operational indicators, such as employee productivity, cost savings and revenue growth;

Report fewer challenges with skills and integration, as they tend to invest in training, upskilling and cross-functional collaboration;

Seek diverse support, including strategic planning assistance and data management tools, not just training.

One more thing: These successful experimenters also emphasized financial and operational outcomes more evenly, while others focused more narrowly on cost savings.

Challenges AI Experimenters Face

Many AI experimenters are struggling not because of unrealistic expectations, but more likely due to unclear objectives or misunderstood value potential. This challenge and difficulties with integrating AI into existing systems are the two biggest hurdles faced by organizations in the early stages of adoption (stages 1 and 2).

Integration issues peak in the middle stages of AI adoption, but they begin in the early stages. Interestingly, the challenge related to understanding the most impactful use cases is most acute in the earliest stage, dips in the middle stages, and resurfaces even at the highest levels of maturity, albeit for different reasons.

The AI experimenters, of course, are unsure how to apply AI strategically and technical compatibility remains a hurdle, unlike the more mature companies. Compounding these issues are unclear or conflicting regulatory guidance and difficulties with data availability and access, a foundational issue for effective AI deployment.

It is the lack of structured approaches, unclear project objectives, and unreliable data that often lead to underwhelming ROI for these companies in the early stages.

Redefining AI Success

In another interesting finding from the survey, we see that as organizations progress to stages 3 to 5, their success metrics evolve from cost savings and process efficiency to revenue growth, customer satisfaction and innovation.

The good news is that organizations starting out on their AI journey can course-correct by focusing on these success metrics. It starts with redefining AI success, which means moving beyond short-term wins to sustainable transformation. 

Having a clear understanding of what you’re trying to accomplish with AI is critical from the outset. Without clarity on what AI is meant to achieve, and how value will be measured, they will struggle to unlock its full potential.

Early experimenters should seek to build a solid foundation by:

Asking Why? Why are you adopting AI? What specific problems are you solving?

Investing in data infrastructure is critical. This step should involve auditing existing data systems and implementing robust data governance frameworks. Organizations will be well served in considering cloud-based platforms for scalability.

Developing a robust integration strategy early. Many existing systems were not originally designed to support AI. To overcome this deficiency, organizations should be proactive in assessing and modernizing infrastructure to handle AI workloads in the initial phases. They are likely to find greater success if IT, data and business teams collaborate and there’s shared ownership of AI initiatives to ensure alignment and adoption.

Aligning AI strategies with business objectives and organizational culture: This is not just a technical step. It involves ensuring organizational readiness and managing cultural and operational changes effectively. 

Turning AI Trials into ROI Triumphs

The research is clear: there’s tremendous ROI potential for early-stage companies that can test, learn and scale AI use cases swiftly. Yet, while speed is crucial to capturing value, it’s important to recognize that AI experimentation is ongoing, requiring continuous iteration.

To win, think big, act swiftly, and continuously evolve — never stop.

Source link