Connect with us

Tech

LLMs can unmask pseudonymous users at scale with surprising accuracy

Published

on



Recall at various precision thresholds.

Advertisement

Recall at various precision thresholds.

In a third experiment, the researchers took 5,000 users from the Netflix dataset and added another 5,000 “distraction” identities of people not in the results. They then added to the list of 10,000 candidate profiles 5,000 query distractors comprising users who appear only in a query set, with no true match in the candidate pool.

Compared to a classical baseline that mimics the Netflix Prize attack to LLM deanonymization, the latter far outperformed the former.

Advertisement


Screenshot

Advertisement

The researchers wrote:

(a) The precision of classical attacks drops very fast, explaining its low recall. In contrast, the precision of LLM-based attacks decays more gracefully as the attacker makes more guesses. (b) The classical attack almost fails completely even at moderately low precision. In contrast, even the simplest LLM attack (Search) achieves non-trivial recall at low precision, and extending it with Reason and Calibrate steps doubles Recall @99% Precision.

The results show that LLMs, while still prone to false positives and other weaknesses, are quickly outstripping more traditional, resource-intensive methods for identifying users online.

The researchers went on to propose mitigations, including platforms enforcing rate limits on API access to user data, detecting automated scraping, and restricting bulk data exports. LLM providers could also monitor for the misuse of their models in deanonymization attacks and build guardrails that make models refuse deanonymization requests.

Of course, another option is for people to dramatically curb their use of social media, or at a minimum, regularly delete posts after a set time threshold.

Advertisement

If LLMs’ success in deanonymizing people improves, the researchers warn, governments could use the techniques to unmask online critics, corporations can assemble customer profiles for “hyper-targeted advertising,” and attackers could build profiles of targets at scale to launch highly personalized social engineering scams.

“Recent advances in LLM capabilities have made it clear that there is an urgent need to rethink various aspects of computer security in the wake of LLM-driven offensive cyber capabilities, the researchers warned. “Our work shows that the same is likely true for privacy as well.”

Source link

Advertisement
Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Tech

Google Chrome shifts to two-week release cycle for increased stability

Published

on

Google Chrome shifts to two-week release cycle for increased stability

Google Chrome will shift from a four-week to a two-week release cycle to roll out new features, bug fixes, and performance improvements more frequently.

With the release of Chrome 153 on September 8, Google will start shipping two new stable versions of the browser every month, breaking the long-standing schedule developers followed since 2021.

The new model applies to both beta and stable releases across Desktop, Android, and iOS. The new development cycle is reflected in the table below:

Table
Source: Google

However, the Dev and Canary channels for early development and testing will continue on the current schedule. Also, the ‘Extended Stable’ branch will remain on its existing eight-week cycle for enterprise customers who need longer update timelines.

Google says the smaller, more frequent releases will reduce disruption and simplify debugging while maintaining stability due to recent process improvements.

Advertisement

“While releases will be more frequent, their smaller scope minimizes disruption and simplifies post-release debugging,” Google says in a press release.

“And thanks to recent process enhancements, we are confident this shift will maintain our high standards for stability.”

For Chrome users, the impact of this change should not be dramatic, but they can expect to see new feature rollouts more frequently. Chrome updates occur silently in the background, but restart prompts may appear more often now.

Although security fixes will still arrive as part of milestone releases, Chrome will receive weekly security updates according to the current model announced in August 2023.

Advertisement

According to it, Chrome receives security fixes every week to reduce the “patch gap,” shortening the opportunity window for hackers to exploit vulnerabilities in one of the world’s most popular web browsers.

With the latest announcement, Google increases the new release cadence for Chrome, building on the previous change, but with a broader, more structural goal this time.

Chrome had a relatively calm start to the year so far, with only one zero-day vulnerability (CVE-2026-2441) reported as actively exploited in the wild. In 2025, hackers leveraged eight Chrome zero-days in attacks.

Malware is getting smarter. The Red Report 2026 reveals how new threats use math to detect sandboxes and hide in plain sight.

Download our analysis of 1.1 million malicious samples to uncover the top 10 techniques and see if your security stack is blinded.

Advertisement

Source link

Continue Reading

Tech

Amid new competition, Chrome speeds up its release schedule

Published

on

As AI-powered browsers from companies like OpenAI, Perplexity, and others attempt to carve out a space for themselves in a market that’s long been dominated by Chrome, Google says that it’s now speeding up the pace of Chrome releases. Starting this September, Chrome will move from a four-week release schedule to a two-week schedule, the tech giant announced on Tuesday.

Chrome has committed to shipping a new milestone of some sort with each release, in areas like stability, speed, or ease of use. (This is in addition to the weekly security updates that were introduced back in 2023.) As a result, those milestones will now come twice as frequently.

Officially, Google says that the new schedule reflects the ever-changing web platform, and it wants to ensure developers have immediate access to the latest tools and improvements. However, the move comes at a time when Chrome is finally facing what could one day become real competition from AI model providers, who are trying to rebuild the browser for an agentic web, where more tasks are automated on users’ behalf.

ChatGPT Atlas, OpenAI’s web browser, offers its AI assistant built in and is experimenting with various automations. Perplexity’s Comet, meanwhile, includes a sidecar AI assistant for all, and other tools like an email assistant and meeting scheduler for itspaid customers.

Advertisement

In response to the emerging threats, Google has been rapidly rolling out deeper Gemini integrations into Chrome, including its own set of agentic features for autonomous tasks.

Google told TechCrunch this latest move isn’t AI-related, but it’s hard to see how the need to compete at a faster pace isn’t playing a role.

The new release schedule starts with the beta and stable versions (ver. 153) of Chrome on September 8, 2026, and will apply to all platforms, including desktop, Android, and iOS. No changes are being made to other early release platforms, like the Dev and Canary channels.

Image Credits:Google

The Extended Stable release, designed for enterprise admins and Chromium embedders who need more time to manage updates, will continue to be on an eight-week cycle, as before. This option will also still be available to Chromebook users, Google notes.

Source link

Advertisement
Continue Reading

Tech

AI-Generated Art Can’t Be Copyrighted After Supreme Court Declines To Review the Rule

Published

on

The Supreme Court of the United States declined to review a case challenging the U.S. Copyright Office’s stance that AI-generated works lack the required human authorship for copyright protection, leaving lower court rulings intact. The Verge reports: The Monday decision comes after Stephen Thaler, a computer scientist from Missouri, appealed a court’s decision to uphold a ruling that found AI-generated art can’t be copyrighted. In 2019, the U.S. Copyright Office rejected Thaler’s request to copyright an image, called A Recent Entrance to Paradise, on behalf of an algorithm he created. The Copyright Office reviewed the decision in 2022 and determined that the image doesn’t include “human authorship,” disqualifying it from copyright protection.

After Thaler appealed the decision, U.S. District Court Judge Beryl A. Howell ruled in 2023 that “human authorship is a bedrock requirement of copyright.” That ruling was later upheld in 2025 by a federal appeals court in Washington, DC. As reported by Reuters, Thaler asked the Supreme Court to review the ruling in October 2025, arguing it “created a chilling effect on anyone else considering using AI creatively.” The U.S. federal circuit court also determined that AI systems can’t patent inventions because they aren’t human, which the U.S. Patent Office reaffirmed in 2024 with new guidance. The UK Supreme Court made a similar determination.

Source link

Continue Reading

Tech

Optimizing a Battery Electric Vehicle Thermal Management System

Published

on

Optimizing a Battery Electric Vehicle Thermal Management System

This webinar looks at a Battery Electric Virtual Vehicle Model of a mid-size BEV, and uses Simulink and Simscape to facilitate design exploration, component refinement, and system-level optimization. The virtual vehicle comprises five subsystems: Electric powertrain, driveline, refrigerant cycle, coolant cycle, and passenger cabin. The model will be tested using different drive cycles, cooling, and heating scenarios.  The results will be analyzed to determine the impact of the different design parameters on vehicle consumption.

The resulting virtual vehicle will be used to:

  • Test different drive cycles and environmental conditions
  • Perform sensitivity analysis
  • Optimize model to improve thermal performance and consumption

Click ‘Watch Now’ to explore this webinar.

Source link

Continue Reading

Tech

What’s New And How Much They Cost

Published

on





Apple has just announced a bunch of new products as part of its early 2026 launch cycle. The updated products include the MacBook Air and the MacBook Pro — which incorporate the newer M5 series Apple Silicon, offering big performance upgrades from the M4 series, and add several other major upgrades to boot. These updates don’t come cheap, but if you need that kind of power, read on.

While the MacBook Pro models claim better performance with the newer chips, they also offer 2x read/write speeds thanks to an upgraded SSD. The MacBook Pro models also start with 1TB of base storage now, which is great for folks seeking colossal storage space without the need for external drives. As for the new MacBook Air, Apple’s bread-and-butter notebook lineup also gets a storage upgrade enabling much faster speeds and now starts with twice the storage capacity on the base models (512GB) compared to the previous generation. That’s not all Apple is offering with this generation, though.

Advertisement

Everything new with the 2026 MacBook Air

As outlined earlier, the biggest change to the 2026 edition of the MacBook Air over the 2024 MacBook Air is the replacement of the older M4 chips with the faster M5 chips. These machines also feature 512GB of base storage, up from 256GB offered on the outgoing models. The SSDs used on the 2026 MacBook Air are rated for faster read/write speeds as well. The unified memory offered on the machine is now faster and rated for 153GB/s.

Advertisement

In addition to these, also new for 2026 is the addition of Apple’s N1 wireless chips that support Wi-Fi 7 as well as Bluetooth 6. The M5 architecture claims better performance with AI-focused work. Thanks to the upgraded processor, the MacBook Air now better handles AI tasks locally, such as running large language models (LLMs) on-device. Also guaranteed for 2026 is faster graphics performance, translating to better ray-tracing and GPU throughput.

Pricing for the 13-inch MacBook Air starts at $1,099, while the 15-inch model will set you back by $1,299. Both models cost $100 less for purchases made through the education channel. Color options on offer include sky blue, midnight, starlight, and silver.

Advertisement

Everything new with the 2026 MacBook Pro

As for Apple’s professional-grade MacBook Pro models, the biggest addition again centers around the new chips: the M5 Pro and the M5 Max. Both these chips use the same 18-core CPU configuration, which is also claimed to be the world’s fastest CPU core. The GPU on the M5 Pro gets 20 cores, while the same for the M5 Max is 40 cores, a not insignificant upgrade even compared to the M4 Max.

In addition to notable gains in CPU and GPU performance, the new 2026 MacBook Pro models also boast of a 2x increase in SSD read and write speeds, while also offering higher baseline storage space. While the base MacBook Pro models with the M5 Pro chip get 1TB storage, models with the M5 Max chip ship with a 2TB SSD. The new chips also boast of notable performance gains in AI-focused tasks. Like with the MacBook Air models, the Pro models also feature Apple’s N1 wireless chip, which enables Wi-Fi 7 and Bluetooth 6 support. Even with the additional performance, the 2026 MacBook Pro models offer 24 hours of battery life as well.

Pricing for the 14‑inch MacBook Pro powered by the M5 Pro chip starts at $2,199 while the 16‑inch MacBook Pro with M5 Pro chip is priced at $2,699. As for the M5 Max models, the 14-inch iteration of this model starts at $3,599 while the 16‑inch MacBook Pro with M5 Max will set you back by $3,899. Color options on offer include space black and silver.

Advertisement



Source link

Advertisement
Continue Reading

Tech

Chasing the Blue as One Maker Attempts to Create Liquid Oxygen at Home

Published

on

Homemade Liquid Oxygen
Liquid oxygen is used in rockets and treated with care in laboratories, but it’s not something that people think about in everyday life. Except for one maker who was interested as to why this was the case, so he decided to make some at home and examine what makes people so cautious about it.



Electron Impressions began by employing a proton-exchange membrane electrolyzer to convert water to pure oxygen. The device neatly splits the water molecules, allowing the hydrogen to escape into the air while the oxygen flows through a drying tube filled with calcium chloride to remove any remaining moisture, because moisture can significantly reduce the effects he was attempting to study.


UNGLINGA 150 Experiments Science Kits for Kids, S.T.E.M Educational Project Toys for Boys Girls Birthday…
  • 150 EXCITING EXPERIMENTS FOR KIDS: DIY projects to get kids’ minds humming, try one of these science experiments, which cover topics like earth…
  • EASY-TO-FOLLOW SCIENTIFIC MANUAL: Well-illustrated in a step-by-step format, which makes the experiments easy to follow. it is easy and fun to…
  • ALMOST TOOLS & MATERIALS NEEDED INCLUDED: high-quality lab science tools and kids-friendly materials. kids can wear goggles to do experiments like…

The next hurdle was to reach the point of liquefaction, which he accomplished using a Stirling cryocooler. This compact device uses the reverse Stirling cycle to attain temperatures as low as 60 Kelvin at the cold end. He placed a Dewar flask within a really nifty 3D printed double-walled PETG sleeve to ensure that the cooling was as efficient as possible. As the oxygen went through, the system was able to condense approximately 30 to 40 milliliters each hour, which isn’t a large amount, but enough to produce a very little but perceptible amount of pale blue liquid within the Dewar.

Homemade Liquid Oxygen
That obvious blue tint was a strong indication that something strange was going on, and it was due to the fact that liquid oxygen has a light sky blue color, which differs from the clear liquid nitrogen you may be more familiar with in cryogenics. To show it off, he poured some into a beaker. Then came the experiments, which demonstrated why liquid oxygen has such a nasty reputation.

Homemade Liquid Oxygen
When he dropped a burning piece of paper into a test tube of liquid, the flame erupted in a rapid and fierce blaze. When he lighted a piece of paper soaked in the liquid, it’s an understatement to say the results were even more remarkable, as it burned more hotter and faster than it would in air. He also bubbled hydrogen gas through a sample and ignited it just above the surface, producing explosive results. Liquid oxygen is essentially a powerful oxidant, converting a wide range of everyday materials into fuel that can burn at an alarming rate.

Homemade Liquid Oxygen
Aside from the fire and flash, the experiment discovered an interesting attribute of oxygen molecules: they contain unpaired electrons, which give them a magnetic attraction. It’s a property that gets lost in the mix when they’re floating around as gas at room temperature, but when they’re cooled down to a liquid form, the effect is much more noticeable.
[Source]

Advertisement

Source link

Continue Reading

Tech

2025 was the year AI grew up. How will AI evolve in 2026?

Published

on

In 2025, AI crossed an important threshold. After years of experimentation, generative AI moved decisively into enterprise workflows, while agentic systems and long-term memory capabilities began to take shape in real-world deployments.

Sarah Hoffman

Director of AI Thought Leadership, AlphaSense.

Advertisement

Source link

Advertisement
Continue Reading

Tech

These Awesome Concept Gadgets Make MWC an Exciting Place to Be

Published

on

Mobile World Congress in Barcelona is where the tech world’s biggest companies get together to show off their latest gadgets. At MWC 2026 we’ve seen some amazing products, including the Leica Leitzphone by Xiaomi and the super skinny Honor Magic V6 foldable phone. But the show always provides a wealth of quirky concept devices and this year is no exception. 

From wild cars to transforming phones, these are the most exciting concepts we’ve seen on the MWC show floor so far. 

Advertisement
xiaomi-vision-gt-concept-hypercar-08

The Vision GT looks amazing but I don’t think I’ll ever get to drive one.

Andrew Lanxon/CNET

Xiaomi Vision GT

Xiaomi is no stranger to EVs, but this is the first time the company has designed a hypercar specifically to be used as a digital asset in the PlayStation 5 racing game Gran Turismo. But Xiaomi didn’t stop there — it actually built the car for real and gave it pride of place on its enormous booth at the conference center. 

The Vision GT, as it’s called (GT stands for Gran Turismo, obvs) is an all-electric hypercar that Xiaomi says is “sculpted by wind.” By which it means, it’s designed with all kinds of swooping lines and flowing inlets that allow it to pass through the air with minimal resistance. It’s got an enormous rear… umm…section? Whatever it is, it’s basically one massive hole to allow for airflow. 

The car looks incredible and I’d love to have been able to sit inside the LED-strewn cockpit but sadly the doors remained firmly closed. This is a concept model designed for the game, and the company has made no statement on whether it ever plans to put something like this into production. 

Advertisement

One thing’s for sure though: It sure as hell won’t come cheap. 

tecno-modular-camera-phone-03

This modular really “lens” itself to photography

Andrew Lanxon/CNET

Tecno modular camera phone

I may have been bowled over this year by Xiaomi and Leica’s incredible camera powerhouse of a phone but Tecno’s concept may even be able to take things further. At its heart is essentially a skinny Android phone but the series of electric contact pins on the back allow you to slap on a variety of modular accessories to completely change what the phone can do. 

Advertisement

One module that particularly caught my eye was a camera unit, that added not just a whopping great zoom lens to the phone, but actually had its own larger camera sensor too. It basically turned the phone into a fully fledged camera that just used the display as the viewfinder. 

Hopefully that larger image sensor would also allow the phone to take some pretty awesome photos, though I’ll have to reserve judgement on what its images look like for at such point Tecno puts it into production. 

Honor Robot Phone

The robot peeks out of the back of the Honor phone.

Advertisement

Katie Collins/CNET

Honor Robot phone

Honor first showed off its concept Robot Phone back at CES in Las Vegas but we’ve been able to get much closer up with the thing this year. It looks sort of like the love child of an Android phone and a DJI Osmo Pocket 3, with a gimbal-stabilized camera unit folding out from inside the phone. 

As a YouTuber myself, I love the idea of having a compact way to shoot my photography videos. Honor has actually had to develop its own tiny motors — based on the technology it uses in the hinges for its folding phones — and CNET’s own Katie Collins was impressed when the camera’s built-in AI complimented her “soft and shiny hair.”

While the robot phone is still in the concept stage right now, Honor has said that it will go into full production and we may even be able to buy it in the second half of the year. 

Advertisement
A man stands in front of the Yoga Book 3D concept using the clip-on tools to alter the picture on the dual-screen computer. The bottom screen shows a 2D image of a mouse, while the top screen shows a 3D render of the same image.

We got hands-on experience with the Yoga Book 3D concept — this picture of a mouse turned into a fully rendered 3D model before our very eyes.

Josh Goldman/CNET

Lenovo Yoga Book 3D concept

Lenovo and sister brand Motorola have frequently shown off some fun concepts at the show, my favorite being Motorola’s wrist-worn phone from 2024. This year Lenovo is leading the way with its concept Yoga Book 3D laptop display, which shows images in 3D — and you don’t even need to wear those stupid glasses to see it. 

Like any tech item launched recently, the device leans on AI to achieve its goals. In this case, the AI goes to work in helping transform 2D drawn objects into full 3D renderings. It has two displays, with the bottom display being your “working screen” where you’ll draw and interact with your creations while the top one uses stereoscopic screen technology to render your images in a way that makes them look actually three dimensional.

We tested it at a hands-on event ahead of the show and CNET’s Tyler Graham remarked that “if you aren’t standing directly in front of the computer, the projection feels less impressive and more headache-inducing.” This has been my experience using any glasses-free 3D technology so I don’t see this kind of tech being deployed in a mass-market product just yet. But it’s nice to see it being experimented with.

Advertisement
man shaking hands with robot

Honor CEO James Li greets the company’s robot on stage at MWC

Katie Collins/CNET

Honor Humanoid robot 

Did you think Honor was done with robots after the aforementioned Robot Phone? Oh no, the company has much bigger plans with robotics and laid its cards out clearly on the table at this year’s show. Its first humanoid robot took to the stage during the company’s press conference, dancing, moonwalking and even backflipping to show off how easily it can move around versus, say, a 38-year-old tech journalist whose knees struggle with standing up, let alone backflipping. 

The robot will be packed with AI smarts, of course, and rather than focusing on industrial applications, Honor is aiming its robot firmly towards the consumer world. It says it’ll be able to help us in the workplace, as a humanoid companion in the home and for assisting with shopping. Though if I hear one word from it about how I maybe don’t need to buy a second pack of biscuits I’m kicking it straight into the sea. 

Advertisement

Source link

Continue Reading

Tech

The MacBook Air M5 starts at $1,099, up $100 from the M4

Published

on

Reports had suggested that Apple wasn’t going to reveal a refreshed MacBook Air with an M5 chip during its week of announcements, but whaddyaknow? Apple just went ahead and did that exact thing.

Let’s deal with the bad news first: Apple is reverting a price cut from last year. It dropped the M4-powered MBA’s starting price down to $999, but for the M5-equipped model, you’ll need to shell out at least $1,099.

The company claimed the M5 MacBook Air will be able to deliver four times faster performance in AI tasks than the M4 MBA. Compared with the M1 MacBook Air, you’ll get up to 9.5 times faster performance, the company said.

Along with swapping in a more powerful chip, Apple has upgraded the starting storage by doubling it from 256GB to 512GB. The company says the SSD has “2x faster read/write performance compared to the previous generation.” You can kit out the MacBook Air with 4TB of internal storage if you have the will and the means.

Advertisement

You’ll also get 16GB of RAM for starters. The memory has 153GB/s of bandwidth, which Apple said is a 28 percent improvement over the M4 MBA. The latest MacBook Air can be equipped with up to 32GB of memory.

Just like it did with the latest iPad Air, Apple has upgraded the connectivity hardware. Thanks to the inclusion of the company’s N1 wireless chip, the M5 MacBook Air supports Wi-Fi 7 and Bluetooth 6.

As you’d imagine, the M5 MacBook Air runs on macOS Tahoe and it supports Apple Intelligence features. It has a Liquid Retina display, 12MP Center Stage camera, a sound system with Spatial Audio support and a pair of Thunderbolt 4 ports, which allows the laptop to support up to two external displays. Apple claims the M5 MacBook Air will run for up 18 hours on a single charge.

The M5 MacBook Air is available in 13-inch and 15-inch models, with the latter starting at $1,299. It’s available in sky blue, midnight, starlight and silver. Pre-orders start on March 4 at 9:15AM ET. The laptops will be available in stores in 33 countries and regions on March 11.

Advertisement

Source link

Continue Reading

Tech

Building A Hackerspace Entry System

Published

on

A hackerspace is a place that generally needs to be accessed by a wide group of people, often at weird and unusual hours. Handing around keys and making sure everything is properly locked up can be messy, too. To make it easy for hackers to get in to [Peter]’s local hackerspace, a simple electronic system was whipped up to grant access.

The combined use of QR code & PIN adds a layer of security.

The basic components of the system are a keypad, a QR code and barcode scanner, a stepper motor, an Arduino Nano, and a Raspberry Pi. The keypad is read by an Arduino Nano, which is also responsible for talking to a stepper motor driver to actuate the lock cylinder.

The system works on the basis of two-factor authentication. Regular users authenticate to enter by presenting a QR code or barcode, and entering a matching PIN number. The system can also be set up for PIN-only entry on a temporary basis.

For example, if the hackerspace is running an event, a simple four-digit pin can allow relatively free access for the duration without compromising long-term security. Actual authentication is handled by the Raspberry Pi, which takes in the scanned barcode and/or PIN, hashes it, and checks it against a backend database which determines if the credentials are valid for entry.

While it’s not technically necessary for a project like this — in fact, you could argue it’s preposterously overkill — we have to take particular note of the machined aluminum enclosure for the keypad. Mere mortals could just run it off on their 3D printers, but if you’ve got access to a CNC router and a suitably chunky piece of aluminum, why not show off a bit?

Advertisement

It’s a nifty system that has served the hackerspace well over some time. We’ve featured some neat access control systems before, too. If you’ve got your own solution to this common problem, don’t hesitate to notify the tipsline!

Source link

Advertisement
Continue Reading

Trending

Copyright © 2025