Toggle light / dark theme

Compute-in-memory chip shows promise for enhanced efficiency and privacy in federated learning systems

In recent decades, computer scientists have been developing increasingly advanced machine learning techniques that can learn to predict specific patterns or effectively complete tasks by analyzing large amounts of data. Yet some studies have highlighted the vulnerabilities of some AI-based tools, demonstrating that the sensitive information they are fed could be potentially accessed by malicious third parties.

A machine learning approach that could provide greater data privacy is federated learning, which entails the collaborative training of a shared neural network by various users or parties that are not required to exchange any raw data with each other. This technique could be particularly advantageous when applied in sectors that can benefit from AI but that are known to store highly sensitive user data, such as health care and finance.

Researchers at Tsinghua University, the China Mobile Research Institute, and Hebei University recently developed a new compute-in-memory chip for federated learning, which is based on memristors, non-volatile electronic components that can both perform computations and store information, by adapting their resistance based on the electrical current that flowed through them in the past. Their proposed chip, outlined in a paper published in Nature Electronics, was found to boost both the efficiency and security of federated learning approaches.

Turning captured carbon into natural gas could provide cost-competitive energy storage

Solar and wind energy are highly variable, dependent on the day, weather and location of the facilities. At times, they can generate more electricity than is needed, but they can also fall short when demand is at its peak. Unfortunately, any extra energy created by these sources is often wasted, as there are few methods that adequately store it long-term. To improve energy security in the United States, the nation requires both sources of energy and novel ways to store and distribute it.

In a new study, published in Cell Reports Sustainability, researchers from Lawrence Livermore National Laboratory (LLNL) have explored how a reactive capture and conversion (RCC) process could be used to produce synthetic renewable natural gas—a chemical form of long-duration energy storage.

“Rather than sourcing carbon from below-ground, RCC enables the use of above-ground carbon as a resource,” said LLNL scientist and lead author Alvina Aui. “Synthetic renewable natural gas, when used as an energy-storage option, can reduce grid instability caused by the intermittency of energy sources like wind and solar.”

True single-photon source boosts secure key rates in quantum key distribution systems

Quantum key distribution (QKD), a cryptographic technique rooted in quantum physics principles, has shown significant potential for enhancing the security of communications. This technique enables the transmission of encryption keys using quantum states of photons or other particles, which cannot be copied or measured without altering them, making it significantly harder for malicious parties to intercept conversations between two parties while avoiding detection.

As true single-photon sources (SPS) are difficult to produce, most QKD systems developed to date rely on attenuated light sources that mimic single photons, such as low-intensity . As these laser pulses can also contain no photons or more than one photon, only approximately 37% of pulses employed by the systems can be used to generate secure keys.

Researchers at the University of Science and Technology of China (USTC) were recently able to overcome this limitation of previously proposed QKD systems, using a true SPS (i.e., a system that can emit only one photon on demand). Their newly proposed QKD system, outlined in a paper published in Physical Review Letters, was found to outperform techniques introduced in the past, achieving a substantially higher secure key rate (SKR).

Webinar: Stolen credentials are the new front door to your network

Cybercriminals no longer need zero-day exploits or other vulnerabilities to breach your systems—these days, they just log in.

On July 9th at 2:00 PM ET, BleepingComputer and SC Media will co-host a live webinar with identity security expert Darren Siegel of Specops Software (part of Outpost24), exploring how threat actors are increasingly breaching networks by simply logging in with stolen credentials.

The webinar “Stolen credentials: The New Front Door to Your Network” will unpack the real-world mechanics behind credential-based attacks and how to stop them before damage is done.

Nanofibers yield stronger, tougher carbon fiber composites

Researchers at the U.S. Department of Energy (DOE)’s Oak Ridge National Laboratory (ORNL) have developed an innovative new technique using carbon nanofibers to enhance binding in carbon fiber and other fiber-reinforced polymer composites—an advance likely to improve structural materials for automobiles, airplanes and other applications that require lightweight and strong materials.

The results, published in the journal Advanced Functional Materials, show promise for making products that are stronger and more affordable, opening new options for U.S. manufacturers to use in applications such as energy and national security.

“The challenge of improving adhesion between carbon fibers and the that surrounds them has been a concern in industry for some time, and a lot of research has gone into different approaches,” said Sumit Gupta, the ORNL researcher who led the project. “What we found is that a hybrid technique using to create chemical and mechanical bonding yields excellent results.”

MiniMax-M1 is a new open source model with 1 MILLION TOKEN context and new, hyper efficient reinforcement learning

From a data platform perspective, teams responsible for maintaining efficient, scalable infrastructure can benefit from M1’s support for structured function calling and its compatibility with automated pipelines. Its open-source nature allows teams to tailor performance to their stack without vendor lock-in.

Security leads may also find value in evaluating M1’s potential for secure, on-premises deployment of a high-capability model that doesn’t rely on transmitting sensitive data to third-party endpoints.

Taken together, MiniMax-M1 presents a flexible option for organizations looking to experiment with or scale up advanced AI capabilities while managing costs, staying within operational limits, and avoiding proprietary constraints.