In just seven minutes, the thieves took off with crown jewels containing with thousands of diamonds along with other precious gems.

Could the French TV series _Lupin_ have been prophetic? The show envisioned a heist at the Louvre, an event that became reality on the morning of October 19, when a group of professional thieves managed to break into the world-famous Paris museum. In just seven minutes, they stole a host of priceless French crown jewels.

The heist took place at around 9:30 am local time, shortly after the museum opened to the public. Using a truck-mounted ladder, the thieves entered the Galerie d’Apollon—located in the Louvre's Petite Galerie wing—through a second-floor window that they forced open with an angle grinder.

Upon entry, the robbers smashed open at least two display cases, took the precious artifacts, and then fled a few minutes later on two Yahama scooters, disappearing into traffic and soon turning onto the highway.

Included in the loot, according to French authorities, were eight crown jewels, almost all from the late Napoleonic era. A ninth item, Empress Eugénie's diamond- and emerald-laden crown, was found damaged nearby, evidently dropped by the fleeing criminals. The thieves did get away with a tiara also belonging to Napoleon III's wife, in full Empire style, decorated with 212 pearls, 1,998 diamonds, and another 992 rose-cut diamonds. They also took a bow brooch belonging to Empress Eugénie with 2,438 diamonds and 196 rose-cut stones. Also in the haul is a parure—a tiara with 24 Ceylon sapphires and 1,083 diamonds, accompanied by a necklace with eight impressive sapphires, more diamonds and gold work, and a pendant earring that belonged to Queen Maria Amalia.

It's difficult to put a number on what this collection of jewels is worth; they are not mere luxury items with their own specific value, but rather priceless possessions. The literal value of the gems, stones, and gold is compounded by their historical value, not to mention the fact that they are part of the heritage of the French state, which in itself makes them likely impossible to sell on the traditional market. However, it is possible that, as often happens in this type of theft, the robbers will disassemble the artifacts, melt down the precious metals, recut the jewels to make them less traceable, and sell them on the gray or black market, potentially generating tens of millions of euros.

Regardless of its outcome, the Louvre heist was a skillful operation. Some analysts say the thieves exploited vulnerabilities in the museum's security system, which has for years struggled with staffing problems, constant work in progress, and also increasing pressure from the exorbitant and growing number of visitors. A nationwide and international manhunt has now begun. At the moment there are no specific suspects, but all available images from the area (including a video that shows one of the thieves at work) are obviously being examined.

With all the surveillance footage and cameras now spread everywhere in the city, there should be plenty of material to identify possible leads. President Emmanuel Macron has strongly condemned the incident, and assured that those responsible will soon be brought to justice. Long gone, moreover, are the days when Italian decorator Vincenzo Peruggia committed what was until now considered the greatest theft in the history of the Louvre: the daring misappropriation of Leonardo's _Mona Lisa_, which took place on August 21, 1911.

That painting was returned two years later; Peruggia attempted to resell it to a Florentine art dealer who then raised the alarm. Maybe a similar stroke of luck could happen in this case as well.

_This story originally appeared on WIRED Italia and has been translated from Italian._

What to Know About the Shocking Louvre Jewelry Heist

Amazon Web Services experienced DNS resolution issues on Monday morning, taking down wide swaths of the web—and highlighting a long-standing weakness in the internet's infrastructure.

A massive cloud outage stemming from Amazon Web Services’ key US-EAST-1 region, its hub in northern Virginia, near the US Capitol, caused widespread disruptions of websites and platforms around the world on Monday morning. Amazon's main ecommerce platform and other properties, including Ring doorbells and the Alexa smart assistant, suffered interruptions and outages throughout the morning, as did Meta's communication platform WhatsApp, OpenAI's ChatGPT, PayPal's Venmo payment platform, multiple web services from Epic Games, multiple British government sites, and many others.

The outages stemmed from Amazon's DynamoDB database application programming interfaces in US-EAST-1, and AWS said in status updates that the problem was specifically related to DNS resolution issues. The “domain name system” is a foundational internet service that essentially acts as an automatic phonebook lookup to translate web URLs like www.wired.com into numeric server IP addresses so web browsers show users the right content. DNS resolution issues occur when DNS servers aren't accurately connecting these dots and, to keep with the phonebook analogy, are providing the wrong numbers for a given name, or vice versa.

“Based on our investigation, the issue appears to be related to DNS resolution of the DynamoDB API endpoint in US-EAST-1,” AWS wrote in status updates on Monday. Shortly after, the company added: “If you are still experiencing an issue resolving the DynamoDB service endpoints in US-EAST-1, we recommend flushing your DNS caches.”

An AWS spokesperson did not immediately respond when asked for details about the nature of the failure. DNS resolution issues can be malicious—known as DNS hijacking—but there is no indication that Monday's AWS outages were nefarious.

“When the system couldn't correctly resolve which server to connect to, cascading failures took down services across the internet,” says Davi Ottenheimer, a longtime security operations and compliance manager and a vice president at the data infrastructure company Inrupt. “Today's AWS outage is a classic availability problem, and we need to start seeing it more as data integrity failure.”

Problems began around 3 am ET. By 5:22 am, AWS had applied “initial mitigations” that were starting to take effect. At 6:35 am, Amazon said that it had fully addressed the underlying technical issues but that “some services will have a backlog of work to work through, which may take additional time to fully process.”

AWS has suffered other large-scale outages, including a major incident in 2023. Reliance on central cloud services from giants like AWS, Microsoft Azure, and Google Cloud Services has, in may ways, improved cybersecurity and stability around the world by creating a baseline of guardrails and best practices for all customers. But this standardization comes with major trade-offs, because the platforms become a single point of failure for large swaths of critical services.

“Failures increasingly trace to integrity,” Ottenheimer says. “Corrupted data, failed validation or, in this case, broken name resolution that poisoned every downstream dependency. Until we better understand and protect integrity, our total focus on uptime is an illusion.”

What the Huge AWS Outage Reveals About the Internet

Anthropic partnered with the US government to create a filter meant to block Claude from helping someone build a nuke. Experts are divided on whether its a necessary protection—or a protection at all.

At the end of August, the AI company Anthropic announced that its chatbot Claude wouldn’t help anyone build a nuclear weapon. According to Anthropic, it had partnered with the Department of Energy (DOE) and the National Nuclear Security Administration (NNSA) to make sure Claude wouldn’t spill nuclear secrets.

The manufacture of nuclear weapons is both a precise science and a solved problem. A lot of the information about America’s most advanced nuclear weapons is Top Secret, but the original nuclear science is 80 years old. North Korea proved that a dedicated country with an interest in acquiring the bomb can do it, and it didn’t need a chatbot’s help.

How, exactly, did the US government work with an AI company to make sure a chatbot wasn’t spilling sensitive nuclear secrets? And also: Was there ever a danger of a chatbot helping someone build a nuke in the first place?

The answer to the first question is that it used Amazon. The answer to the second question is complicated.

Amazon Web Services (AWS) offers Top Secret cloud services to government clients where they can store sensitive and classified information. The DOE already had several of these servers when it started to work with Anthropic.

“We deployed a then-frontier version of Claude in a Top Secret environment so that the NNSA could systematically test whether AI models could create or exacerbate nuclear risks,” Marina Favaro, who oversees National Security Policy & Partnerships at Anthropic tells WIRED. “Since then, the NNSA has been red-teaming successive Claude models in their secure cloud environment and providing us with feedback.”

The NNSA red-teaming process—meaning, testing for weaknesses—helped Anthropic and America’s nuclear scientists develop a proactive solution for chatbot-assisted nuclear programs. Together, they “codeveloped a nuclear classifier, which you can think of like a sophisticated filter for AI conversations,” Favaro says. “We built it using a list developed by the NNSA of nuclear risk indicators, specific topics, and technical details that help us identify when a conversation might be veering into harmful territory. The list itself is controlled but not classified, which is crucial, because it means our technical staff and other companies can implement it.”

Favaro says it took months of tweaking and testing to get the classifier working. “It catches concerning conversations without flagging legitimate discussions about nuclear energy or medical isotopes,” she says.

Wendin Smith, the NNSA’s administrator and deputy undersecretary for counterterrorism and counterproliferation, tells WIRED that “the emergence of \[AI\]-enabled technologies has profoundly shifted the national security space. NNSA’s authoritative expertise in radiological and nuclear security places us in a unique position to aid in the deployment of tools that guard against potential risk in these domains, and that enables us to execute our mission more efficiently and effectively.”

Both NNSA and Anthropic were vague about the “potential risks in these domains,” and it’s unclear how helpful Claude or any other chatbot would be in the construction of a nuclear weapon.

“I don’t dismiss these concerns, I think they are worth taking seriously,” Oliver Stephenson, an AI expert at the Federation of American Scientists, tells WIRED. “I don’t think the models in their current iteration are incredibly worrying in most cases, but I do think we don’t know where they’ll be in five years time … and it’s worth being prudent about that fact.”

Stephenson points out that much is hidden behind a barrier of classification, so it’s hard to know what impact Anthropic’s classifier has had. “There is a lot of detail in the design of implosion lenses that go around the nuclear core,” Stephenson says. “You need to structure them very precisely to perfectly compress the core to get a high yield explosion … I could imagine that being the kind of thing where AI could help synthesize information from a bunch of different physics papers, a bunch of different publications on nuclear weapons.”

Still, he says, AI companies should be more specific when they talk about safety. “When Anthropic puts out stuff like this, I’d like to see them talking in a little more detail about the risk model they’re really worried about,” he says. “It is good to see collaboration between AI companies and the government, but there is always the danger with classification that you put a lot of trust into people determining what goes into those classifiers.”

For Heidy Khlaaf, the chief AI scientist at the AI Now Institute with a background in nuclear safety, Anthropic’s promise that Claude won’t help someone build a nuke is both a magic trick and security theater. She says that a large language model like Claude is only as good as its training data. And if Claude never had access to nuclear secrets to begin with, then the classifier is moot.

“If the NNSA probed a model which was not trained on sensitive nuclear material, then their results are not an indication that their probing prompts were comprehensive, but that the model likely did not contain the data or training to demonstrate any sufficient nuclear capabilities,” Khlaaf tells WIRED. “To then use this inconclusive result along with common nuclear knowledge to build a classifier for nuclear ‘risk indicators’ would be quite insufficient and a long way from legal and technical definitions of nuclear safeguarding.”

Khlaaf adds that this kind of announcement fuels speculation about capabilities that chatbots don’t have. “This work seems to be relying on an unsubstantiated assumption that Antrophic's models will produce emergent nuclear capabilities without further training, and that is simply not aligned with the available science,” she says.

Anthropic disagrees. “A lot of our safety work is focused on proactively building safety systems that can identify future risks and mitigate against them,” an Anthropic spokesperson tells WIRED. “This classifier is an example of that. Our work with NNSA allows us to do the appropriate risk assessments and create safeguards that prevent potential misuse of our models.”

Khlaaf was also less excited about the partnership between the US government and a private AI company. Companies like Anthropic are hungry for training data, and she sees the US government’s broader rush to embrace AI as an opportunity for the AI industry to acquire data it couldn’t get elsewhere. “Do we want these private corporations that are largely unregulated to have access to that incredibly sensitive national security data?” she says. “Whether you’re talking about military systems, nuclear weapons, or even nuclear energy.”

And then there’s the precision. “These are precise sciences, and we know that large language models have failure modes in which they’re unable to even do the most basic mathematics,” Khlaaf says. In 1954, a math error tripled the yield of a nuclear weapon the US tested in the Pacific Ocean, and the government is still dealing with the literal fallout. What might happen if a chatbot did nuclear weapons math wrong and a human didn’t double check its work?

To Anthropic’s credit, it says it doesn’t want a future where people are using chatbots to play around with nuclear weapons science. It’s even offering its classifier to any other AI company that wants it. “In our ideal world, this becomes a voluntary industry standard, a shared safety practice that everyone adopts,” Favaro says. “This would require a small technical investment, and it could meaningfully reduce risks in a sensitive national security domain.”

Anthropic Has a Plan to Keep Its AI From Building a Nuclear Weapon. Will It Work?

Plus: A secret FBI anti-ransomware task force gets exposed, the mystery of the CIA’s Kryptos sculpture is finally solved, North Koreans busted hiding malware in the Ethereum blockchain, and more.

In a stunning new study, researchers at UC San Diego and the University of Maryland revealed this week that satellites are leaking a wealth of sensitive data completely unencrypted, from calls and text messages on T-Mobile to in-flight Wi-Fi browsing sessions, to military and police communications. And they did this with just $800 in off-the-shelf equipment.

Face recognition systems are seemingly everywhere. But what happens when this surveillance and identification technology doesn’t recognize your face as a face? WIRED spoke with six people with facial differences who say flaws in these systems are preventing them from accessing essential services.

Authorities in the United States and United Kingdom announced this week the seizure of nearly 130,000 bitcoins from an alleged Cambodian scam empire. At the time of the seizure, the cryptocurrency fortune was worth $15 billion—the most money of any type ever confiscated in the US.

Control over a significant portion of US election infrastructure is now in the hands of a single former Republican operative, Scott Leiendecker, who just purchased voting machine company Dominion Voting Systems and owns Knowink, an electronic poll book firm. Election security experts are currently more baffled about the implications than worried about any possibility of foul play.

While a new type of attack could let hackers steal two-factor authentication codes from Android phones, the biggest cybersecurity development of the week was the breach of security firm F5. The attack, which was carried out by a “sophisticated” threat actor reportedly linked to China, poses an “imminent threat” of breaches against government agencies and Fortune 500 companies. Finally, we sifted through the mess that is VPNs for iPhones and found the only three worth using.

But that’s not all! Each week, we round up the security and privacy news we didn’t cover in depth ourselves. Click the headlines to read the full stories. And stay safe out there.

**“The Com” Hackers Leak Personal Info of Hundreds of DHS, ICE, DOJ, and FBI Officials**

In recent years, perhaps no single group of hackers has caused more mayhem than “the Com,” a loose collective of mostly cybercriminal gangs whose subgroups like Lapus$ and Scattered Spider have carried out cyberattacks and ransomware extortion operations targeting victims from MGM Casinos to Marks & Spencer grocery stores. Now they’ve turned their sites to US federal law enforcement.

On Thursday, one member of the Com’s loose collective began posting to Telegram an array of federal officials’ identifying documents. One spreadsheet, according to 404 Media, contained what appeared to be personal information of 680 Department of Homeland Security officials, while another included personal info on 170 FBI officials, and yet another doxed 190 Department of Justice officials. The data in some cases included names, email addresses and phone numbers, and addresses—in some cases of officials’ homes rather than the location of their work. The user who released the data noted in their messages a statement from the DHS that Mexican cartels have offered thousands of dollars for identifying information on agents, apparently mocking this unverified claim.

“Mexican Cartels hmu we dropping all the doxes wheres my 1m,” the user who released the files wrote, using the abbreviation for “hit me up” and seemingly demanding a million dollars. “I want my MONEY MEXICO.”

**Secret FBI Task Force Planned to Disrupt Ransomware Group Inside Russia, Reports Claim**

Over the last year—at least—the FBI has operated a “secret” task force that may have worked to disrupt Russian ransomware gangs, according to reports published this week in France’s Le Monde and Germany’s Die Zeit. The publications allege that at the end of last year, the mysterious Group 78 presented its strategy to two different meetings of European officials, including law enforcement officials and those working in judicial services. Little is known about the group; however, its potentially controversial tactics appeared to spur typically tight-lipped European officials to speak out about Group 78’s existence and tactics.

At the end of last year, according to the reports, Group 78 was focusing on the Russian-speaking Black Basta ransomware gang and outlined two approaches: running operations inside Russia to disrupt the gang’s members and try to get them to leave the country; and also to “manipulate” Russian authorities into prosecuting Black Basta members. Over the last few years, Western law enforcement officials have taken increasingly disruptive measures against Russian ransomware gangs—including infiltrating their technical infrastructure, trying to ruin their reputations, and issuing a wave of sanctions and arrest warrants—but taking covert action inside Russia against ransomware gangs would be unprecedented (at least in public knowledge). The Black Basta group has in recent months gone dormant after 200,000 of its internal messages were leaked and its alleged leader identified.

**ICE Division and Secret Service Had Access to AI License Plate Cameras**

Over the last few years, AI-powered license plate recognition cameras—which are placed at the side of the road or in cop cars—have gathered billions of images of people’s vehicles and their specific locations. The technology is a powerful surveillance tool that, unsurprisingly, has been adopted by law enforcement officials across the United States—raising questions about how access to the cameras and data can be abused by officials.

This week, a letter by Senator Ron Wyden revealed that one division of ICE, the Secret Service, and criminal investigators at the Navy all had access to data from the cameras of Flock Safety. “I now believe that abuses of your product are not only likely but inevitable, and that Flock is unable and uninterested in preventing them,” Wyden’s letter addressed to Flock says. Wyden’s letter follows increasing reports that government agencies, including the CBP, had access to Flock’s 80,000 cameras. “In my view,” Wyden wrote, “local elected officials can best protect their constituents from the inevitable abuses of Flock cameras by removing Flock from their communities.”

Elsewhere this week, Flock announced it was partnering with Amazon’s Ring, which makes video doorbells, to allow agencies that use Flock to request Ring customers share footage with them.

**Mystery of the CIA’s Kryptos Sculpture Finally Solved—Thanks to the Smithsonian’s Archive**

For 35 years, the Kryptos sculpture that sits in a courtyard of the CIA’s headquarters has beguiled cryptographers with a message enciphered in its rows of lettering that no one has been able to fully solve for decades—until now. Two men, Jarett Kobek and Richard Byrne, finally cracked the puzzle with the help of documents they found in the Smithsonian Archive that revealed the solution, The New York Times reports. Now they’re in a strange dispute with the sculpture’s owner, Jim Sanborn, who confirmed their answer but also asked them not to reveal it ahead of an upcoming auction that will sell the solution to the highest bidder in an effort to raise money for charity and his own potential medical expenses. The two solvers assured Sanborn they wouldn’t publicly reveal the solution, but Sanborn nonetheless asked them to sign a nondisclosure agreement, and the auction house holding the auction sent them an email threatening legal action if they disclose the answer. Both men say they’ve had to lawyer up in response. We at WIRED offer a humble solution: Kobek and Byrne post a cryptographic hash of the answer online—an irreversible but replicable mathematical conversion of the text that they can later use to prove they knew the answer without revealing it to anyone now—then sign the NDA, and everyone goes home happy.

**North Korean Hackers Are Hiding Malware in Ethereum’s Blockchain**

For years, North Korean state-sponsored hackers have targeted cryptocurrency users and companies, stealing billions that have been funneled into the Kim regime. Now they’re using one of crypto’s own blockchains as part of their hacking toolkit. Security researchers at Google this week revealed that North Korean hackers have been using a technique known as “EtherHiding” to host their malware, storing their malicious code in a smart contract on Ethereum’s blockchain—which, unlike Bitcoin’s, can host and run code in the blockchain’s distributed network of computers. When a victim is tricked into opening a file sent by the hackers, the file pulls down crypto-stealing malware hosted on Ethereum’s blockchain, where it’s harder to remove or defend against than on a traditional server. Google says it’s the first time the company has seen the technique used by state-sponsored cybercriminals.

Hackers Dox ICE, DHS, DOJ, and FBI Officials

Networking software company F5 disclosed a long-term breach of its systems this week. The fallout could be severe.

Thousands of networks—many of them operated by the US government and Fortune 500 companies—face an “imminent threat” of being breached by a nation-state hacking group following the breach of a major maker of software, the federal government warned on Wednesday.

F5, a Seattle-based maker of networking software, disclosed the breach on Wednesday. F5 said a “sophisticated” threat group working for an undisclosed nation-state government had surreptitiously and persistently dwelled in its network over a “long term.” Security researchers who have responded to similar intrusions in the past took the language to mean the hackers were inside the F5 network for years.

**Unprecedented**

During that time, F5 said, the hackers took control of the network segment the company uses to create and distribute updates for BIG IP, a line of server appliances that F5 says is used by 48 of the world’s top 50 corporations. Wednesday’s disclosure went on to say the threat group downloaded proprietary BIG-IP source code information about vulnerabilities that had been privately discovered but not yet patched. The hackers also obtained configuration settings that some customers used inside their networks.

Control of the build system and access to the source code, customer configurations, and documentation of unpatched vulnerabilities has the potential to give the hackers unprecedented knowledge of weaknesses and the ability to exploit them in supply-chain attacks on thousands of networks, many of which are sensitive. The theft of customer configurations and other data further raises the risk that sensitive credentials can be abused, F5 and outside security experts said.

Customers position BIG-IP at the very edge of their networks for use as load balancers and firewalls, and for inspection and encryption of data passing into and out of networks. Given BIG-IP's network position and its role in managing traffic for web servers, previous compromises have allowed adversaries to expand their access to other parts of an infected network.

F5 said that investigations by two outside intrusion-response firms have yet to find any evidence of supply-chain attacks. The company attached letters from firms IOActive and NCC Group attesting that analyses of source code and build pipeline uncovered no signs that a “threat actor modified or introduced any vulnerabilities into the in-scope items." The firms also said they didn’t identify any evidence of critical vulnerabilities in the system. Investigators, which also included Mandiant and CrowdStrike, found no evidence that data from its CRM, financial, support case management, or health systems was accessed.

The company released updates for its BIG-IP, F5OS, BIG-IQ, and APM products. CVE designations and other details are here. Two days ago, F5 rotated BIG-IP signing certificates, though there was no immediate confirmation that the move is in response to the breach.

The US Cybersecurity and Infrastructure Security agency has warned that federal agencies that rely on the appliance face an “imminent threat” from the thefts, which “pose an unacceptable risk.” The agency went on to direct federal agencies under its control to take “emergency action.” The UK’s National Cyber Security Center issued a similar directive.

CISA has ordered all federal agencies it oversees to immediately take inventory of all BIG-IP devices in networks they run or in networks that outside providers run on their behalf. The agency went on to direct agencies to install the updates and follow a threat-hunting guide that F5 has also issued. BIG-IP users in private industry should do the same.

_This story originally appeared on_ _Ars Technica._

Why the F5 Hack Created an ‘Imminent Threat’ for Thousands of Networks

Former GOP operative Scott Leiendecker just bought Dominion Voting Systems, giving him ownership of voting systems used in 27 states. Election experts don't know what to think.

The news last week that Dominion Voting Systems was purchased by the founder and CEO of Knowink, a Missouri-based maker of electronic poll books, has left election integrity activists confused over what, if anything, this could mean for voters and the integrity of US elections.

The company, acquired by Scott Leiendecker, a former Republican Party operative and election director in Missouri before founding Knowink, said in a press release that he was rebranding Dominion, which has headquarters in Canada and the United States, under the name Liberty Vote “in a bold and historic move to transform and improve election integrity in America” and to distance the company from false allegations made previously by President Donald Trump and his supporters that the company had rigged the 2020 presidential election to give the win to President Joe Biden.

The Liberty release said that the rebranded company will be 100 percent American-owned, that it will have a “paper ballot focus” that leverages hand-marked paper ballots, will “prioritize facilitating third-party auditing,” and is “committed to domestic staffing and software development.” The press release provided no details, however, to explain what this means in practice.

Dominion, the second leading provider of voting machines in the US, whose systems are used in 27 states—including the entire state of Georgia—has developed its software in Canada and Belgrade, Serbia, for two decades. A search on LinkedIn shows numerous programmers and other workers in Serbia who claim to be employed by the company.

The Liberty statement does not say whether the company plans to rewrite code developed by these foreign workers—which would potentially involve rewriting hundreds of thousands of lines of code—or whether the company will move foreign developers to the US or replace them with American programmers. (Dominion has a US headquarters in Colorado.) A Liberty official, who agreed to speak on the condition that they not be named, told WIRED only that Leiendecker “is committed to 100 percent … domestic staffing and software development.” An unnamed source told CNN, however, that Liberty will continue to have a presence in Canada, where its machines are used across the country.

Philip Stark, professor of statistics at UC Berkeley and a longtime election-integrity advocate, says Liberty’s assurance about domestic-only workers is a red herring. “If the claim is that this is somehow a security measure, it isn’t. Because programmers based in the US also … may be interested in undermining or altering election integrity,” he tells WIRED.

With regard to third-party audits mentioned in the press release, a Liberty official told WIRED this means the company will conduct a “third-party, top-to-bottom, independent review of \[Dominion\] software and equipment in a timely manner and will work closely with federal and state certification agencies and report any vulnerabilities” to give voters assurance in the machines and the results they produce. The company didn’t say when this review would occur, but a Liberty representative told Axios it would happen ahead of next year's midterm elections, and the company would "rebuild or retire" machines as needed.

But Stark and other election experts believe this is unrealistic and appears to be aimed just at making the company “look good” to right-wing critics. “There isn’t time between now and 2026 to design, test, and certify new voting equipment,” Stark says. Federal and state testing and certification is expensive and can take months, depending on what changes a company makes to its voting systems. And many states have legal limitations on how close to an election updates to voting systems can occur. If Liberty plans to replace foreign Dominion workers with US ones, this will likely also slow down any review and rebuilding.

What’s more, Stark, who developed procedures for risk-limiting audits used in some jurisdictions to verify the integrity of election results, says the only audits that can provide assurances about election outcomes are postelection audits of paper ballots. An audit of voting software can’t even provide assurance that a voting machine hasn’t been subverted, because the software and machine configurations can be altered after auditing and certification. A software audit can only identify existing vulnerabilities, but many of these have already been identified by third-party computer security experts who examined Dominion code in the past. And if Liberty were to rewrite any Dominion code, it would need to be audited for new vulnerabilities introduced by the recoding.

**The Shrinking Voting Industry**

News of the acquisition apparently caught Dominion customers off guard last week. Leiendecker reportedly held a conference call on Friday with Colorado county election clerks who were angry that they’d learned of the sale from media reports instead of the company. The election clerks reportedly grilled Leiendecker about conspiracy claims about Dominion and his plans for the company. The company’s sparse new website currently has only a brief note from Leiendecker on the homepage announcing that the company is “100% American‐owned.”

Leiendecker’s acquisition of Dominion is not a big surprise, however. Knowink and Dominion have worked together closely for a while. Dominion has used Knowink as a subcontractor when bidding for election contracts in order to offer customers a turnkey election solution that includes voting machines and poll books. And Leiendecker’s acquisition continues a long history of consolidation in the voting industry.

For the past two decades, there have only been three primary makers of voting machines in the US, the identities of which have changed frequently through mergers and acquisitions. Currently, the top three are Election Systems & Software, Dominion (now Liberty Vote), and Hart InterCivic, in that order.

Dominion was founded in 2003 in Toronto and was not a big player in the US until 2010, when it acquired two other voting firms—Premier Election Solutions (formerly Diebold Election Systems) and Sequoia Voting Systems. Premier/Diebold, like Dominion, exited the elections business after being beleaguered by security problems with its systems and lingering mistrust over the company’s close ties to the Republican Party (the former CEO had been a fundraiser for President George W. Bush). Sequoia left in large part due to financial issues. In acquiring Dominion, Leiendecker also acquires any remaining remnants of Premier/Diebold and Sequoia—equipment, software, and people—that Dominion may still possess.

With the Dominion acquisition, Leiendecker gains control of election equipment in more than half of the states. Dominion equipment is used across 26 states plus Puerto Rico. Knowink electronic poll books, which replace traditional paper poll books used to verify the eligibility of voters when they sign in at precincts, are used in 29 states plus the District of Columbia. But there are jurisdictions across 14 states, covering 20 million registered voters, that use both Dominion and Knowink systems. This gives companies that Leiendecker controls ownership of equipment that covers the entire election process in those jurisdictions—from the verification of registered voters to the casting of ballots and tabulation of results. And Georgia uses Knowink and Dominion systems statewide. Knowink poll books even interact with Dominion voting machines since a voter smart card inserted into the poll books gets encoded and then inserted into voting machines to pull up a digital ballot for the voter.

Nonetheless, some might be confused about why Leiendecker would want to acquire Dominion, since the voting machine industry has never been particularly lucrative. A 2017 study estimated it to be a $300 million-a-year business, with Dominion’s annual revenue estimated at the time to be about $100 million. Unlike IT systems that get replaced every three to five years, election systems get replaced once in a decade or longer. Voting equipment vendors often rely on long-term maintenance and election-services contracts to stay solvent while vying for new equipment contracts.

It’s not clear how much Dominion is worth today. In 2018, Dominion’s management team partnered with Staple Street Capital, a private equity firm, to acquire a controlling stake in the company. Staple estimated the company’s value at $80 million at the time. A more recent report produced by an accounting firm hired by Dominion to support its claim for damages in Fox defamation lawsuit said the company would have been worth about $741 million in December 2020 if not for Fox’s false election-rigging claims.

The Liberty official who spoke with WIRED would not disclose the price Leiendecker paid or whether other parties contributed funds. He said only that Leiendecker “is the sole owner and he privately financed” the acquisition.

**Rife With Speculation**

On social media, users have expressed concern about Leiendecker’s close ties to the Republican Party as a possible reason for his interest in buying Dominion, raising questions about whether he sides with conspiracy theories about Dominion and the 2020 election. Recent settlements in Dominion lawsuits against conspiracy theorists raises further questions about this, as have Leiendecker’s connection to Trump ally Ed Martin. Leiendecker didn’t respond to questions about these concerns.

Following the 2020 election in which right-wing Trump supporters accused Dominion of having ties to Venezuela—apparently confusing the company with Smartmatic, a voting machine company founded by Venezuelan engineers—and rigging the election, Dominion filed defamation lawsuits against right-wing news outlets Fox News and One America News Network for $1.6 billion for amplifying the false claims. It also filed suit against MyPillow CEO Mike Lindell, and former Trump campaign lawyers Rudy Giuliani and Sidney Powell, for $1.3 billion each.

In 2023, Fox News settled its suit for $787.5 million, and Newsmax recently settled for $67 million. But last month, Dominion reached agreements with OAN, Powell, and Giuliani for undisclosed terms. These latest settlements were a condition of Leiendecker’s acquisition of Dominion. Dominion still has pending lawsuits against Lindell and former Overstock CEO Patrick Byrne. The Liberty official who spoke with WIRED says these will be ending soon as well. “All litigation has been resolved or is in the process of being resolved,” the official wrote in an email. But Stefanie Lambert, Byrne’s attorney, wrote in an email to WIRED: “There is no settlement. Dr. Byrne is looking forward to trial.” A lawyer for Lindell did not respond to an inquiry but Lindell told Infowars this week that he will never settle with Dominion.

The settlements raise questions about whether Leiendecker simply didn’t want to be saddled with lawsuits or if he was doing right-wing defendants a favor by forcing Dominion into settlements favorable to them. Leiendecker didn’t respond to a question about the undisclosed settlement terms.

While Leiendecker hasn't directly addressed the issue around conspiracy theorists and election manipulation publicly, his response to allegations of vote-rigging made more than a decade ago suggests he considers election integrity to be paramount, even when doing so might not be politically advantageous. Leiendecker was appointed by Missouri's Republican secretary of state to investigate St. Louis elections administration after problems arose in the 2000 election there, according to Axios. Leiendecker was then hired by Ed Martin to be the Republican director of the St. Louis City Board of Election Commissioners from 2005 until 2009

Martin, now a staunch Trump ally and supporter of his stolen-election claims, is currently the US pardon attorney and prior to this served under the current Trump administration as the interim US attorney for DC, where he demoted prosecutors who worked on January 6 insurrection cases. But in 2005, he was chair of the St. Louis City Election Board when he hired Leiendecker. Five years after hiring Leiendecker, Martin ran unsuccessfully as a GOP candidate for a congressional seat against the Democratic incumbent and was backed by the St. Louis Tea Party. Similar to what occurred following Trump’s 2020 presidential loss, Martin alleged there were irregularities in the vote counting. A Tea Party spokeswoman called the election “stolen,” and Martin refused to concede the election. Leiendecker, instead of supporting his former boss, disputed the stolen election claim at the time, saying there were "no shenanigans” and challenged the critics to prove otherwise. St. Louis Public Radio reported that he directed this challenge at Martin.

During his time as director of the St. Louis elections board, Leiendecker was widely credited by both political parties for improving the election board’s procedures and ensuring that it processed votes more efficiently than previous boards.

**Paper Chase**

When it comes to future elections, Leiendecker has said in Liberty’s press release and media statements that the company is committed to providing election technology that leverages “hand-marked paper ballots” in “compliance with President Trump's executive order.”

Trump, acting on his false vote-rigging claims, has called for all-paper elections, and his March executive order has called for all voting systems to use or produce a voter-verifiable paper ballot that voters can review to ensure machines didn’t alter their votes. Some media outlets have interpreted Leiendecker’s statements to mean he aligns with Trump’s stance to ban systems that don’t produce a paper record. The Liberty official who spoke with WIRED says Leiendecker meant only that Liberty will offer products that “enable compliance with federal and state standards”—whatever those standards may be. Current voting system standards advise states to use systems that produce a paper record but cannot require this; states decide on their own what voting systems they use. Similarly, Trump has no authority to decide what voting systems states use; the Constitution gives this authority to states alone, and the EO is currently being litigated in court.

Notably, Trump’s EO also wants states to stop using paper ballots that have bar codes or QR codes to record and tabulate votes, because the codes are not human-readable. These ballots are produced by so-called ballot-marking devices, or BMDs—voting systems that allow voters to make selections through a touchscreen tablet or other assistive device, which then prints a paper ballot showing the voter’s selections that the voter can read to verify that the machine marked their votes correctly. The ballot is then passed through an optical scanner to record and tabulate the results.

Many BMDs print ballots with a bar code or QR code on them that is encoded with the voter’s choices; it is this encoded portion that the scanner reads to record and tabulate the votes, not the human-readable portion of the ballot that the voter can verify. Ironically, like Trump, election-integrity activists (on the left and right) have long opposed the use of paper ballots with bar codes or QR codes for security reasons: Someone could theoretically program the systems to print a voter’s choices on the human-readable portion while encoding something else in the bar code or QR code that the scanner reads and tabulates. Dominion makes a BMD called the ImageCast X, most of which produce ballots with QR codes, which is used in all or parts of 15 states, according to Verified Voting, which tracks election equipment usage in the US. They are used statewide in Georgia, an important swing state that was the subject of much controversy in the 2020 election.

Many election-integrity activists want states that use these types of ballots to tabulate only from the human-readable portion of the ballot, not the QR code or bar code, or simply have voters fill out regular paper ballots by hand instead. Trump is seeking the latter.

One Republican Now Controls a Huge Chunk of US Election Infrastructure

An estimated 100 million people live with facial differences. As face recognition tech becomes widespread, some say they’re getting blocked from accessing essential systems and services.

Autumn Gardiner thought updating her driving license would be straightforward. After getting married last year, she headed to the local Department of Motor Vehicles office in Connecticut to get her name changed on her license. While she was there, Gardiner recalls, officials said she needed to update her photo. That’s when things started to go wrong.

Every time staff tried to take her photo, Gardiner says, the system would reject it. “Everyone's watching. They’re taking more photos,” she recalls. Gardiner, who works as a grant manager for an environmental conservation charity, is one of a small number of people globally who live with Freeman-Sheldon syndrome. Sometimes known as Whistling Face syndrome, the genetic condition impacts muscles around the face and skull, which can result in an undersized mouth.

As more staff members at the DMV were called to help, Gardiner says she started to believe the rejected photos were being caused by her facial difference. The camera system didn’t seem to work for her, she says. “It was humiliating and weird. Here’s this machine telling me that I don’t have a human face,” Gardiner says.

Gardiner isn’t alone. Around half a dozen people living with a variety of facial differences—from birthmarks to craniofacial conditions—tell WIRED they are increasingly struggling to participate in modern life as identity verification software, which often is powered by machine learning, is quickly becoming commonplace.

Courtesy of Rick Guidotti for Positive Exposure

Some of those living with facial differences tell WIRED they have undergone multiple surgeries and experienced stigma for their entire lives, which is now being echoed by the technology they are forced to interact with. They say they haven’t been able to access public services due to facial verification services failing, while others have struggled to access financial services. Social media filters and face-unlocking systems on phones often won’t work, they say.

“The facial difference community is constantly overlooked,” says Phyllida Swift, the CEO of Face Equality International (FEI), an umbrella group representing other facial difference and disfigurement charities and organizations. There are more than 100 million people worldwide who live with facial disfigurements, FEI estimates. People with facial differences have experienced problems with airport passport gates, photo apps, social media video filters, background blurring on video calls, and more, according to FEI’s research. “In many countries, facial recognition is increasingly a part of everyday life, but this technology is failing our community,” Nikki Lilly, a patron for FEI, told a United Nations meeting in March.

**Access Denied**

From phones to hotel rooms, your face increasingly acts as a digital key. Over the past decade, rapid machine learning and AI advancements have led to the creation of a range of face recognition technologies—meaning that more than ever before, your appearance can be used as a digital identifier. Police have widely deployed face recognition systems, which have frequently been found to be inaccurate and biased against Asian and Black people, while the wider world of face checking has seen government services, anti-fraud systems, and financial institutions using AI to complete identity checks. Most recently, social media and porn websites have adopted face scanning as part of age verification measures.

These “authentication” face checks can take multiple forms. Selfies can be automatically compared to existing ID documents; while liveness tests can require you to take a short video to show you are real—not a fraudster holding a printed photo to the camera. Broadly speaking, these biometric systems often measure your facial features—such as the distance between your eyes, or the size of your jaw—to create “faceprints.” While these kinds of surveillance technologies may be effective for a lot of people, they may not be able to detect people with facial differences. The underlying machine learning technology may not be trained on datasets with a variety of faces, for example.

“The case of Face Equality International—of people who have different faces—is a really important canary in the coal mine of what can go wrong when these systems don’t work,” says Greta Byrum, the founder of technology consultancy firm Present Moment Enterprises, which focuses on the social impact of technologic systems and has provided some pro bono work to FEI. “We’re seeing facial recognition technologies becoming one of those hammers to which everything looks like a nail,” Byrum says.

“I would like to see people realize that not everyone looks the same, from basic skin tone to facial features to the standard facial appearance,” says Crystal Hodges, a speaker and writer who lives with Sturge-Weber syndrome and has a purple “port wine stain” across part of her face. “People’s facial symmetry may not be the same. Other people have different facial features that you don't always see every day, but they exist.”

Courtesy of Crystal Hodges

Hodges, who lives in Arizona, says that last year she and her husband were trying to get their credit scores from a large credit reporting agency when its face verification service wouldn’t work for her. “​​I tried five to eight times,” she says. Towards winter, Hodges says, the birthmark will become darker. “We were trying a different ring light, we were trying calmer lighting, and it would not accept. My husband had a beard, and it accepted his.” Hodges says she recently tried the same system again, and while it worked this time, she still never got her credit score.

“If you don’t include people with disabilities or people with facial differences in the development of these processes, no one’s going to think of these issues,” says Kathleen Bogart, a psychology professor at Oregon State University who specializes in disability research and lives with a facial difference. “AI has amplified these issues, but it’s rooted in long-standing underrepresentation and prejudice towards people with facial differences that occurred long before AI was a thing.”

**Too Little, Too Late**

When face verification systems fail, it’s often hard to find help—piling more pressure on a stressful situation. For months, Maryland resident Noor Al-Khaled has struggled to create an online account with the Social Security Administration. Al-Khaled, who lives with the rare cranio-facial condition Ablepheron Macrostomia, says having an online account would allow her to easily access SSA records and quickly send documents to the agency.

“I don’t drive because of my vision; I should be able to rely on the site,” Al-Khaled says. “You have to take a selfie, and the pictures have to match,” Al-Khaled says. “Because of the facial difference, I don't know if it’s not recognizing the ID or the selfie, but it’s always saying images don’t match.”

Not having that access makes life harder. “On an emotional level, it just makes me feel shut out from society,” she explains. Al-Khaled says that all services should provide alternative ways for people to access online systems. “The lack of other fallback options means that sometimes people get trapped in these labyrinths of technological systems,” says Byrum from Present Moment Enterprises.

Courtesy of WIRED source

An SSA spokesperson says alternative options to face verification are available, and it is “committed” to making its services accessible to everyone. The agency, the spokesperson says, does not run facial recognition systems itself but uses Login.gov and ID.me for verification services. The General Services Administration, which runs Login.gov, did not respond to WIRED’s request for comment. “Accessibility is a core priority for ID.me,” a spokesperson for ID.me says, adding it has previously helped people with facial differences and offered to directly help Al-Khaled after WIRED was in touch.

“There are few things more dehumanizing than being told by a machine that you’re not real because of your face,” says Corey R. Taylor, a New York–based actor and motivational speaker who lives with a craniofacial anomaly. Last year, Taylor says, he was using a financial app to access a small amount of money; as he tried to complete the payment processes, he found that the face verification system could not match his selfie to the image on his ID. To get the system to work, he had to move into different positions. “I had to literally raise my eyes and contort my face,” Taylor says. When he emailed the company, he got what appeared to be a boilerplate response.

“One of the main things that we are advocating for is alternative methods,” says Hannah Saunders, the head of policy and education at FEI. “There’s possibly a long-term fix in terms of changing the technology, but at the moment, we’re most concerned that people can apply for jobs, they can access their money, they can get through passport control with their family, and that negotiating those kinds of systems isn’t as stressful as it is.”

Swift, the CEO at FEI, says the organization has been “approached” multiple times to work with businesses to address some of the issues but has not yet been able to sort out any formal arrangements. Swift says FEI wants to put the “right systems and safeguards in place” to protect its members. Also, Swift adds, “we get the sense that this is quite low down on the list of priorities \[of tech companies\], and therefore progress is seemingly slow.”

However, the problems extend beyond emerging technologies. Many of the people WIRED spoke to explained that they often face appearance-based bullying online, and the internet’s wider pivot to short-form video content like TikTok and Instagram Reels leaves them feeling isolated. “The internet, for so long, was a safe haven for me,” Al-Khaled says. “I found my home and solace online because it was the only place where I could meet people and actually be myself without being stared at or gawked at.” Hodges says people have previously said to her that they are shocked that she can leave her house. “I feel like that’s kind of transferred now to: ‘I can’t believe you leave your house, and I can't believe you go online,’” she says.

Courtesy of Corey Taylor

At the DMV in Connecticut last year, Gardiner says that staff were able to manually override the photography system, although she did explain her facial difference to them in the process. She received her replacement temporary license that day and full license later in the mail. “I’ve dealt with a lot of shit when it comes to my face, but after I left, I just cried,” Gardiner recalls. “It was kind of like a reminder that yes, I’m different and I don't fit in.”

After leaving, Gardiner messaged a state representative she knew and told her about the experience, according to messages shared with WIRED. The official got in touch with the local DMV commissioner and discussed the issue plus staff training, they say.

“While we adhere to stringent guidelines to guarantee photo compliance, our employees approach their roles with empathy, ensuring they understand and address the needs of those they serve,” a DMV CT spokesperson says. “We want our residents to have professional interactions with the DMV.” A spokesperson for Idemia, which provides the DMV with technology, says it regularly offers staff training, and its algorithms have been ranked highly in external evaluations.

Gardiner says that while there was ultimately a positive response, she has heard stories of people with facial differences running into problems at the DMV in other states. “I really want to try to pressure for better protocols: What do humans do when the AI doesn't work?” she says. But with face recognition systems being used so broadly, the problem can seem insurmountable. “After that whole DMV experience, I then had to get my passport done,” Gardiner says, “and I ran into the same problem with my passport photo.”

When Face Recognition Doesn’t Know Your Face Is a Face

The malicious app required to make a “Pixnapping” attack work requires no permissions.

Android devices are vulnerable to a new attack that can covertly steal two-factor authentication codes, location timelines, and other private data in less than 30 seconds.

The new attack, named Pixnapping by the team of academic researchers who devised it, requires a victim to first install a malicious app on an Android phone or tablet. The app, which requires no system permissions, can then effectively read data that any other installed app displays on the screen. Pixnapping has been demonstrated on Google Pixel phones and the Samsung Galaxy S25 phone and likely could be modified to work on other models with additional work. Google released mitigations last month, but the researchers said a modified version of the attack works even when the update is installed.

**Like Taking a Screenshot**

Pixnapping attacks begin with the malicious app invoking Android programming interfaces that cause the authenticator or other targeted apps to send sensitive information to the device screen. The malicious app then runs graphical operations on individual pixels of interest to the attacker. Pixnapping then exploits a side channel that allows the malicious app to map the pixels at those coordinates to letters, numbers, or shapes.

“Anything that is visible when the target app is opened can be stolen by the malicious app using Pixnapping,” the researchers wrote on an informational website. “Chat messages, 2FA codes, email messages, etc. are all vulnerable since they are visible. If an app has secret information that is not visible (e.g., it has a secret key that is stored but never shown on the screen), that information cannot be stolen by Pixnapping.”

The new attack class is reminiscent of GPU.zip, a 2023 attack that allowed malicious websites to read the usernames, passwords, and other sensitive visual data displayed by other websites. It worked by exploiting side channels found in GPUs from all major suppliers. The vulnerabilities that GPU.zip exploited have never been fixed. Instead, the attack was blocked in browsers by limiting their ability to open iframes, an HTML element that allows one website (in the case of GPU.zip, a malicious one) to embed the contents of a site from a different domain.

Pixnapping targets the same side channel as GPU.zip, specifically the precise amount of time it takes for a given frame to be rendered on the screen.

“This allows a malicious app to steal sensitive information displayed by other apps or arbitrary websites, pixel by pixel,” Alan Linghao Wang, lead author of the research paper “Pixnapping: Bringing Pixel Stealing out of the Stone Age,” explained in an interview. “Conceptually, it is as if the malicious app was taking a screenshot of screen contents it should not have access to. Our end-to-end attacks simply measure the rendering time per frame of the graphical operations to determine whether the pixel was white or nonwhite.”

**Pixnapping in 3 Steps**

The attack occurs in three main steps. In the first, the malicious app invokes Android APIs that make calls to the app the attacker wants to snoop on. These calls can also be used to effectively scan an infected device for installed apps of interest. The calls can further cause the targeted app to display specific data it has access to, such as a message thread in a messaging app or a 2FA code for a specific site. This call causes the information to be sent to the Android rendering pipeline, the system that takes each app's pixels so they can be rendered on the screen. The Android-specific calls made include activities, intents, and tasks.

In the second step, Pixnapping performs graphical operations on individual pixels that the targeted app sent to the rendering pipeline. These operations choose the coordinates of target pixels the app wants to steal and begin to check if the color of those coordinates is white or nonwhite or, more generally, if the color is c or non-c (for an arbitrary color c).

“Suppose, for example, \[the attacker\] wants to steal a pixel that is part of the screen region where a 2FA character is known to be rendered by Google Authenticator,” Wang said. “This pixel is either white (if nothing was rendered there) or nonwhite (if part of a 2FA digit was rendered there). Then, conceptually, the attacker wants to cause some graphical operations whose rendering time is long if the target victim pixel is nonwhite and short if it is white. The malicious app does this by opening some malicious activities (i.e., windows) in front of the victim app that was opened in Step 1.”

The third step measures the amount of time required at each coordinate. By combining the times for each one, the attack can rebuild the images sent to the rendering pipeline one pixel at a time.

As Ars reader hotball put it in the comments below:

> Basically the attacker renders something transparent in front of the target app, then using a timing attack exploiting the GPU's graphical data compression to try finding out the color of the pixels. It's not something as simple as "give me the pixels of another app showing on the screen right now." That's why it takes time and can be too slow to fit within the 30 seconds window of the Google Authenticator app.

In an online interview, paper coauthor Ricardo Paccagnella described the attack in more detail:

> Step 1: The malicious app invokes a target app to cause some sensitive visual content to be rendered.
> 
> Step 2: The malicious app uses Android APIs to "draw over" that visual content and cause a side channel (in our case, GPU.zip) to leak as a function of the color of individual pixels rendered in Step 1 (e.g., activate only if the pixel color is c).
> 
> Step 3: The malicious app monitors the side effects of Step 2 to infer, e.g., if the color of those pixels was c or not, one pixel at a time.
> 
> Steps 2 and 3 can be implemented differently depending on the side channel that the attacker wants to exploit. In our instantiations on Google and Samsung phones, we exploited the GPU.zip side channel. When using GPU.zip, measuring the rendering time per frame was sufficient to determine if the color of each pixel is c or not. Future instantiations of the attack may use other side channels where controlling memory management and accessing fine-grained timers may be necessary (see Section 3.3 of the paper). Pixnapping would still work then: The attacker would just need to change how Steps 2 and 3 are implemented.

The amount of time required to perform the attack depends on several variables, including how many coordinates need to be measured. In some cases, there’s no hard deadline for obtaining the information the attacker wants to steal. In other cases—such as stealing a 2FA code—every second counts, since each one is valid for only 30 seconds. In the paper, the researchers explained:

> To meet the strict 30-second deadline for the attack, we also reduce the number of samples per target pixel to 16 (compared to the 34 or 64 used in earlier attacks) and decrease the idle time between pixel leaks from 1.5 seconds to 70 milliseconds. To ensure that the attacker has the full 30 seconds to leak the 2FA code, our implementation waits for the beginning of a new 30-second global time interval, determined using the system clock.
> 
> ... We use our end-to-end attack to leak 100 different 2FA codes from Google Authenticator on each of our Google Pixel phones. Our attack correctly recovers the full 6-digit 2FA code in 73%, 53%, 29%, and 53% of the trials on the Pixel 6, 7, 8, and 9, respectively. The average time to recover each 2FA code is 14.3, 25.8, 24.9, and 25.3 seconds for the Pixel 6, Pixel 7, Pixel 8, and Pixel 9, respectively. We are unable to leak 2FA codes within 30 seconds using our implementation on the Samsung Galaxy S25 device due to significant noise. We leave further investigation of how to tune our attack to work on this device to future work.

In an email, a Google representative wrote, “We issued a patch for CVE-2025-48561 in the September Android security bulletin, which partially mitigates this behavior. We are issuing an additional patch for this vulnerability in the December Android security bulletin. We have not seen any evidence of in-the-wild exploitation.”

Pixnapping is useful research in that it demonstrates the limitations of Google's security and privacy assurances that one installed app can’t access data belonging to another app. The challenges in implementing the attack to steal useful data in real-world scenarios, however, are likely to be significant. In an age when teenagers can steal secrets from Fortune 500 companies simply by asking nicely, the utility of more complicated and limited attacks is probably of less value.

_This story originally appeared on_ _Ars Technica._

A New Attack Lets Hackers Steal 2-Factor Authentication Codes From Android Phones

Officials in the US and UK have taken sweeping action against “one of the largest investment fraud operations in history,” confiscating a historic amount of funds in the process.

Over the past five years, criminals behind widespread romance and investment scams—often inelegantly referred to as “pig butchering”—have stolen tens of billions from people around the world. Now law enforcement has carried out one of its biggest operations yet against that sprawling scam industry, targeting the operators of several modern slavery scam compounds in Southeast Asia—where, as a whole, hundreds of thousands of human trafficking victims have been forced to run the fraud operations on behalf of criminal gangs.

On Tuesday, officials in the United States and United Kingdom took coordinated action against one giant Cambodian organization and its boss who has allegedly run a series of notorious scam centers in the country. The US Treasury Department’s Office of Foreign Assets Control (OFAC) announced it has issued financial sanctions against 146 “targets” linked to the newly designated Prince Group Transnational Criminal Organization. This action includes targeting individuals and shell companies linked to the alleged criminal enterprise. As part of the sweeping action also involving the FBI, the US Department of Justice (DOJ) has also seized almost 130,000 bitcoin worth around $15 billion at the time of the announcement—the largest US cryptocurrency seizure to date.

This Prince Group organized crime entity, OFAC says, is made up of the Prince Holding Group, a Cambodia-based company, its chairman and CEO Chen Zhi, and his associates and business partners. Publicly, the company describes itself as “one of the largest conglomerates in Cambodia” and says it is involved in real estate development and financial services. However, the DOJ alleges that “in secret” Chen and other executives “grew Prince Group into one of Asia’s largest transnational criminal organizations” and ran at least 10 scam compounds across Cambodia.

“As alleged, the defendant directed one of the largest investment fraud operations in history, fueling an illicit industry that is reaching epidemic proportions,” said Joseph Nocella Jr., a US attorney for the Eastern District of New York, in a statement. “Prince Group’s investment scams have caused billions of dollars in losses and untold misery to victims around the world.” Chen has not been arrested and remains at large, the DOJ says.

“The masterminds behind these horrific scam centers are ruining the lives of vulnerable people and buying up London homes to store their money,” Britain’s foreign secretary, Yvette Cooper, said in a statement. The UK also placed financial sanctions on Chen, the Prince Group, and other linked entities. The UK’s sanctions also “freeze” businesses and properties allegedly linked to Chen in London, including a £12 million ($16 million) mansion in North London and a £100 million ($133 million) office building in the City of London.

A WIRED email to an address listed as a media contact on the Prince Holding Group website immediately bounced back.

“Today’s coordinated action represents the most consequential blow yet to cybercrime networks operating out of Southeast Asia,” says John Wojcik, a senior threat researcher focused on Asia at security firm Infoblox, who previously tracked scam compounds and Southeast Asian cybercrime at the United Nations Office on Drugs and Crime. Wojcik claims that this doesn't appear to have been “just another criminal syndicate—it was one of the region’s largest cybercriminal and money laundering enterprises, and a clear leader in criminal fintech and infrastructure.”

In another, still-unexplained twist, however, crypto-tracing firm Elliptic pointed out in a blog post Tuesday that the bitcoins seized by US law enforcement actually appear to be the same funds that were reportedly stolen in 2020 from a Chinese cryptocurrency mining firm called Lubian. Today's indictment describes Lubian as part of Chen Zhi's money laundering operation—perhaps an alleged scheme to funnel dirty scam funds into cryptocurrency mining hardware that would produce new, clean coins with no criminal history.

Who stole the funds in 2020—or whether a theft even occurred—remains far from clear. “It could be the theft was faked by Chen Zhi as part of a money laundering scheme to confuse the money trail,” says Elliptic cofounder Tom Robinson. “The second possibility is that the theft was genuine. Potentially it could have been the US government. But more likely is that it was someone else.” That thief, Robinson says, might have then been tracked down by US law enforcement, which then somehow took the funds from him or her.

Crypto mining money-laundering and mysterious thefts aside, Chen is accused in the indictment of being a major player in the Chinese-speaking pig butchering ecosystem. Over the last decade, organized crime gangs operating out of Southeast Asia have operated dozens of scam centers across Myanmar, Laos, and Cambodia. Often run by Chinese organized crime groups, the scam compounds have lured people from more than 60 countries around the world to work in them by using fraudulent job advertisments—often within the tech sector. When people arrive at the compounds, they often have their passports taken away from them and are then forced to run a wide range of online scams, targeting other people around the world. If they don’t comply, they are sometimes beaten or tortured. As well as human trafficking and fraud, the scam centers are frequently linked to money laundering and online casinos.

The DOJ indictment against Chen and seven unnamed co-conspirators alleges that the Prince Group operated more than 100 businesses in 30 countries and names several other alleged subsidiaries. Some “local networks,” including one based in Brooklyn, New York, also worked on behalf of the Prince Group, the indictment says. It claims that since 2015 Chen and company executives built and operated “forced-labor scam compounds” across Cambodia and “used their political influence in multiple countries to protect their criminal enterprise,” including from China’s police officials and Chinese intelligence agency the Ministry of State Security.

“Chen Zhi was directly involved in managing the scam compounds and maintained records associated with each one, including records tracking profits from the scams that explicitly referenced ‘_sha zhu_,’ or pig-butchering,” the indictment claims, alleging there were also “ledgers of bribes to public officials.” One document allegedly held by Chen listed that two scam centers were equipped with 1,250 mobile phones that “controlled” 76,000 social media accounts. The indictment also claims that Chen held images demonstrating “Prince Group’s violent methods” against people who had been trafficked to the scam centers. The document includes images showing people bloodied and beaten.

The seizure of 127,271 bitcoins worth more than $15 billion at the time they were confiscated represents by far the biggest monetary seizure in the US Justice Department’s history—not just of cryptocurrency, but of money of any kind. That US law enforcement record was previously set in 2022 with the seizure of 95,000 bitcoins worth $3.6 billion from a Manhattan couple who later pleaded guilty to stealing them from the Bitfinex exchange, and prior to that with a billion-dollar seizure in 2020 of bitcoins allegedly stolen from the Silk Road dark-web drug market by an unnamed hacker. Meanwhile, police in the UK seized 61,000 bitcoins worth $6.7 billion in June from a Chinese woman accused of an investment scam, an even bigger sum than those US records but less than half the sum taken from the Prince Group operation.

“It’s important to note that this seizure is extraordinary not only for its scale but for what it represents,” Ari Redbord, global head of policy at crypto-tracing firm TRM Labs, adding that the seizure is still a “small fraction” of the money generated by scam centers. “These are not isolated scams; they are factory-scale operations powered by forced labor, supercharged by the speed and scale of crypto, and connected through sophisticated money-laundering infrastructure that spans Cambodia, Myanmar, Laos, China, and beyond,” Redbord says.

Redbord says the widespread action “strikes at the operational and financial core” of the widespread scam center ecosystem. In recent years, researchers tracking the scam compounds in Southeast Asia have seen them rapidly grow and use their illicitly gained money to invest in increasingly high-tech scam operations. Over the past two years, scam compounds have also been spotted emerging outside of Southeast Asia, with sites emerging in the Middle East, Eastern Europe, Latin America, and West Africa.

“By targeting the financial architecture—the shell companies, banks, exchanges, and real estate that move and hide these proceeds—the US and UK are dismantling the economic engine that sustains these crimes,” Redbord says. “This is what a 21st-century counter-threat finance campaign looks like—coordinated, data-driven, and global.”

_Updated 2:30 pm ET October 14, 2025 to add more information about the seized bitcoins being traced to an apparent theft from the Lubian mining operation._

Feds Seize Record-Breaking $15 Billion in Bitcoin From Alleged Scam Empire

There are dozens of iPhone VPNs at your disposal, but these are the services that will actually keep your browsing safe.

**Other iPhone VPNs We Tested**

**Surfshark:** Surfshark was a strong contender for the main list. Even its Starter plan comes with extra features like a masked email generator. Features like ad and tracker blocking, as well as unlimited simultaneous connections, come standard across plans. However, it was a bit slower than my top picks, dropping around 20 percent of speed on average, compared to around 15 percent for the top options.

**Mullvad:** Mullvad is a favorite among privacy enthusiasts, and for good reason. It doesn’t fuss with multi-year discounts or referral programs, and you don’t even need to provide an email to sign up for an account. You can even pay the static monthly fee by mailing Mullvad cash. It’s a great service if privacy is your top priority, but it trades speeds and features in the process. VPN services like Nord and Proton have quickly grown into full privacy and security suites, while Mullvad is more focused on making a robust VPN. In the context of an iPhone, the scales tip more toward those security suites, but Mullvad is still a great privacy-focused option to keep in mind.

**ExpressVPN:** By the numbers, ExpressVPN should be at the top of the list. It has a ton of servers, a featureset that can go toe-to-toe with Nord, and speeds only a touch below Proton. However, ExpressVPN has found itself in a spiral of increasing controversies over the past four years, and the brand has yet to get back on solid footing. After being purchased by Kape Technologies—the company behind the infamous adware company Crossrider—former US intelligence official Daniel Gericke took over at CTO and continued in that role for two years, even after being fined over $300,000 by the US Department of Justice for hacking activities on behalf of a foreign government. Gericke left in 2023, but that same year, ExpressVPN experienced a large swath of layoffs, and Kape, its parent company, was delisted from the London Stock Exchange. The vast majority of shares went to Unikmind Holdings Limited, a company owned by Israeli billionaire Teddy Sagi, who got his start by creating gambling software Playtech. That’s an extremely condensed version of what ExpressVPN has gone through over the past few years. The company hasn’t done anything nefarious, but the revolving door of executive control tied to controversial names doesn’t inspire confidence.

**Private Internet Access:** Private Internet Access, or PIA, is also owned by Kape Technologies, and it followed a similar playbook as ExpressVPN and CyberGhost, which Kape also owns. After the acquisition and community backlash, there’s been very little transparency about what’s going on in the company. A connection to Kape definitely raises questions, but that doesn’t immediately disqualify a service from being included. Unfortunately for PIA, it had much slower speeds than any of the other VPN services I tested, so regardless of ownership, it isn’t a top pick for iPhone VPNs.

**iPhone VPN Feature Comparison**

You might be surprised how similar the three VPNs I chose are when you break down their features, but that’s not an accident. Given how restrictive I was with who made the final cut, there’s a pretty high bar for inclusion. Although there are some minor differences, I designed this list in a way that you can choose one of my picks without reading a single word and still come out the other side with an excellent VPN for iPhone.

**How We Tested**

To earn the best iPhone VPN title, a service needs to satisfy three criteria. It needs to be secure, fast, and easy to use. That may seem straightforward enough, but there’s a little more that goes into it. For ease of use, I only looked at VPNs that offer a one-tap connection. If you need to configure anything, that’s a disqualification. That still encompasses a lot of the most popular VPNs on iPhone, so I narrowed the field further by focusing on apps that balance usability with power. You should have all the relevant features in the iOS app that are available in the desktop app, and organized in a way that they don’t disrupt that one-tap experience.

Speed testing is where I focused a lot of my testing time. Speed testing is highly variable, and trying to come out with one singular number to encompass the speed of thousands of servers is a fool’s errand. The numbers I gathered for this guide are the result of 20 tests I ran for each VPN, then averaged.

I tested five locations for each VPN, measuring my unprotected speed immediately before testing and running three passes before averaging. Each location was tested at a different time of day, and I removed any outliers before averaging. For this guide, that meant if there was greater than a 10 percent deviation between two of the three passes. After averaging the speed drop for each location, I gathered all of those numbers together and averaged them for a final speed drop.

Finally, security. The traditional wisdom with a VPN is that you, at some point, have to put some faith in the company that its privacy policy is accurate and it isn’t lying about its logging practices. That didn’t do it for me. Again, I set a high bar for inclusion.

Every VPN I’ve included here has not only been independently audited, but also been forced to uphold its no-logs policy in legal proceedings. You indeed need to put some trust that the VPN provider you’re using is telling the truth, but the options I included all have rock-solid track records when it comes to transparency.

**_Power up with unlimited access to_ WIRED_._** _Get best-in-class reporting and exclusive subscriber content that's too important to ignore. Subscribe Today._

Source

Wired