Weak cyber defenses are exposing critical infrastructure — how enterprises can proactively thwart cunning attackers to protect us all - Related to poses, social, critical, reconstructing, all

MELON: Reconstructing 3D objects from images with unknown poses

We leverage two key techniques to aid convergence of this ill-posed problem. The first is a very lightweight, dynamically trained convolutional neural network (CNN) encoder that regresses camera poses from training images. We pass a downscaled training image to a four layer CNN that infers the camera pose. This CNN is initialized from noise and requires no pre-training. Its capacity is so small that it forces similar looking images to similar poses. Providing an implicit regularization greatly aiding convergence.

The second technique is a modulo loss that simultaneously considers pseudo symmetries of an object. We render the object from a fixed set of viewpoints for each training image. Backpropagating the loss only through the view that best fits the training image. This effectively considers the plausibility of multiple views for each image. In practice, we find N=2 views (viewing an object from the other side) is all that’s required in most cases, but sometimes get more effective results with N=4 for square objects.

Moving to another aspect, these two techniques are integrated into standard NeRF training, except that instead of fixed camera poses. Poses are inferred by the CNN and duplicated by the modulo loss. Photometric gradients back-propagate through the best-fitting cameras into the CNN. We observe that cameras generally converge quickly to globally optimal poses (see animation below). After training of the neural field, MELON can synthesize novel views using standard NeRF rendering methods.

We simplify the problem by using the NeRF-Synthetic dataset, a popular benchmark for NeRF research and. Common in the pose-inference literature. This synthetic dataset has cameras at precisely fixed distances and a consistent “up” orientation, requiring us to infer only the polar coordinates of the camera. This is the same as an object at the center of a globe with a camera always pointing at it, moving along the surface. We then only need the latitude and longitude (2 degrees of freedom) to specify the camera pose.

As an illustrative case study, we applied the framework to a dermatology model, which utilizes a convolutional neural network similar to that describe...

Google DeepMind researchers have built an AI weather forecasting tool that makes faster and more accurate predictions than the best system available t...

It’s been a great year for the Dutch startup ecosystem.

Venture capitalists have. So far, invested $ into Netherlands-based early-stage companie...

Weak cyber defenses are exposing critical infrastructure — how enterprises can proactively thwart cunning attackers to protect us all

Direct attacks on critical infrastructure get a lot of attention, but the bigger danger often lies in something less visible: The poor cybersecurity practices of the businesses that keep these systems running. , a staggering 84% earned a “D” grade or worse for their cybersecurity practices, with 43% falling into the “F” category. Only 6% of companies got an “A” for their efforts. What’s more troubling is that industries at the heart of critical infrastructure — like energy, finance and. Healthcare — are among the weakest links.

Corporate cybersecurity failures can’t be separated from national security risks. The strength of the ’ critical infrastructure relies on solid digital defenses, and when businesses fail to secure their networks, they leave the entire country vulnerable to potentially devastating attacks.

A mismatch between risks and. Preparedness.

Furthermore, the World Economic Forum’s latest research reveals a worrying disconnect. Two-thirds of organizations are counting on AI to shape cybersecurity this year, but. Only 37% have processes in place to check if their AI tools are secure before using them. It’s like putting all your trust in a high-tech gadget without reading the manual — risky and potentially asking for trouble. While businesses are grappling with preparation, AI is being leveraged by cybercriminals to orchestrate offensive campaigns against them. For instance, corporate executives are facing a surge of highly targeted phishing attacks created by AI bots.

Cyberattacks of any type are getting harder to repel. Take the finance and insurance sectors, for example. These industries manage sensitive data and are key to our economy, yet 63% of companies in these sectors earned a “D” and 24% failed entirely. It’s no surprise that, last year, LoanDepot, one of the country’s biggest mortgage lenders, was hit by a major ransomware attack that forced them to take some systems offline.

Ransomware continues to be a major issue due to weak cybersecurity measures. Crowdstrike found that cloud environment intrusions surged by 75% from 2022 to 2023, with cloud-conscious incidents rising by 110% and cloud-agnostic incidents by 60%. Despite advances in technology, email remains one of the main methods for cybercriminals to target companies. nearly 37% of all emails in 2024 were flagged as “unwanted,” a slight increase from the previous year. This points to that businesses are still struggling to address fundamental vulnerabilities through proactive measures.

Weak cybersecurity isn’t merely a corporate issue — it’s a national security risk. The 2021 Colonial Pipeline attack disrupted energy supplies and exposed vulnerabilities in critical industries. Rising geopolitical tensions, especially with China, amplify these risks. Recent breaches attributed to state-sponsored actors have exploited outdated telecommunications equipment and other legacy systems, revealing how complacency in updating technology can put national security in danger.

For instance, last year’s hack of and international telecommunications companies exposed phone lines used by top officials and. Compromised data from systems for surveillance requests, threatening national security. Weak cybersecurity at these companies risks long-term costs, allowing state-sponsored actors to access sensitive information, influence political decisions and. Disrupt intelligence efforts.

It’s critical to recognize that vulnerabilities don’t exist in isolation. What happens in one sector — be it telecommunications, energy or finance — can have a domino effect that impacts national security at large. Now, more than ever, it’s essential to collaborate with IT and DevOps teams to close any gaps, and prioritize timely updates. To stay one step ahead of evolving cyber threats.

To tackle these growing cyber threats, businesses need to step up their security game. Taking action in these key areas can make a big difference:

If not yet, implement AI-based cybersecurity tools that continuously monitor for suspicious activities. Including AI-powered phishing attempts. These tools can automate the detection of emerging threats, analyze patterns and respond in real-time. Minimizing potential damage from cyberattacks such as ransomware.

Establish a comprehensive system to evaluate the security of AI tools before deployment. This should include rigorous AI security audits that test for vulnerabilities such as susceptibility to adversarial attacks, data poisoning or model inversion. Companies should also implement secure development lifecycle practices for AI tools, conduct regular penetration testing and ensure compliance with established frameworks like ISO/IEC 27001 or the NIST AI Risk Management Framework.

As cloud-based attacks increase, especially with the surge in ransomware and. Data breaches, companies should adopt advanced cloud security measures. This includes robust encryption, continuous vulnerability scanning and the integration of AI to predict and. Prevent future breaches in cloud environments.

Let me remind you that legacy systems are a hacker’s favorite target. Keeping systems updated and applying patches promptly can help close the door on vulnerabilities before attackers exploit them.

No organization can face today’s cyber threats on its own. Collaboration between private businesses and government agencies is more than helpful — it’s imperative. Sharing threat intelligence in real-time allows organizations to respond faster and stay ahead of emerging risks. Public-private partnerships can also level the playing field by offering smaller companies access to resources like funding and. Advanced security tools they might not otherwise afford.

The aforementioned World Economic Forum’s analysis makes it clear: Resource constraints create gaps in cyber resilience. By working together, business and the government can close those gaps and build a stronger, more secure digital environment — one that’s superior equipped to prevent increasingly sophisticated cyberattacks.

The business case for proactive security.

Some businesses may argue that implementing stricter cybersecurity measures is too expensive. However, the price of doing nothing could be much higher. , the average cost of a data breach rose to $ million in 2024, up from $ million in 2023, marking a 10% increase — the highest since the pandemic in 2020.

Businesses that have already taken steps towards more secure systems benefit from faster incident response times and greater trust from clients and. Partners who want to keep their data safe. For instance, Mastercard developed a real-time fraud detection system that uses machine learning (ML) to analyze transactions globally. It has reduced fraud, boosted customer trust and improved security for clients and merchants through instant suspicious activity alerts.

Such companies also save costs. two-thirds of organizations are now integrating security AI and automation into their security operations centers. When widely applied to prevention workflows — such as attack surface management (ASM) and posture management — these organizations saw an average reduction of $ million in breach costs compared to those not using AI in their prevention strategies.

America’s critical infrastructure is only as strong as its weakest link — and right now. That link is business cybersecurity. Weak private-sector defenses pose a serious risk to national security, the economy and public safety. To prevent catastrophic outcomes, decisive action is needed from both businesses and the government.

Fortunately, progress is underway. Former President Biden’s executive order on cybersecurity, requires companies working with the federal government to meet stricter cybersecurity standards. This initiative encourages business leaders, investors and policymakers to enforce stronger safeguards, invest in resilient infrastructure and foster industry-wide collaboration. By taking these steps, the weakest link can become a powerful line of defense against cyber threats.

The stakes are too high to ignore. If businesses — government partners or not — fail to act, the systems everyone relies on could face more serious and devastating disruptions.

Vincentas Baubonis leads the team at Cybernews.

AutoBNN is based on a line of research that over the past decade has yielded improved predictive accuracy by modeling time series using GPs with learn...

Differential privacy (DP) is a property of randomized mechanisms that limit the influence of any individual user’s information while processing and an...

Due to the complexity of understanding and. Solving various tasks solely using instructions, the size of multi-task LLMs typically spans from several b...

Social learning: Collaborative learning with large language models

Large language models (LLMs) have significantly improved the state of the art for solving tasks specified using natural language. Often reaching performance close to that of people. As these models increasingly enable assistive agents, it could be beneficial for them to learn effectively from each other, much like people do in social settings, which would allow LLM-based agents to improve each other’s performance.

To discuss the learning processes of humans, Bandura and. Walters described the concept of social learning in 1977, outlining different models of observational learning used by people. One common method of learning from others is through a verbal instruction (, from a teacher) that describes how to engage in a particular behavior. Alternatively, learning can happen through a live model by mimicking a live example of the behavior.

Given the success of LLMs mimicking human communication, in our paper “Social Learning: Towards Collaborative Learning with Large Language Models”. We investigate whether LLMs are able to learn from each other using social learning. To this end, we outline a framework for social learning in which LLMs share knowledge with each other in a privacy-aware manner using natural language. We evaluate the effectiveness of our framework on various datasets, and propose quantitative methods that measure privacy in this setting. In contrast to previous approaches to collaborative learning, such as common federated learning approaches that often rely on gradients, in our framework, agents teach each other purely using natural language.

In December 1972, at the American Association for the Advancement of Science meeting in Washington, . MIT meteorology professor Ed Lorenz gave a t...

ScreenAI’s architecture is based on PaLI, composed of a multimodal encoder block and an autoregressive decoder. The PaLI encoder uses a vision transfo...

Amsterdam-headquartered Nebius, which builds full-stack AI infrastructure for tech firms, has secured $700mn in a private equity deal led by Nvidia. A...