Agregátor RSS

Schrödinger's Cat, the famous 1935 thought experiment, imagines a cat that, for reasons rooted in quantum physics we need not dwell on here, is simultaneously alive and dead. In a landscape of relentless cyberattacks and mounting infrastructural complexity, a routine backup task can appear finished, with every box ticked and every check made. The job is not actually done, however, until it has been verified as fully ready for recovery in the event of cyber disaster. A task that looks sorted can turn out, under scrutiny, to be anything but. A backup can be logged as 'complete' once data has been copied to storage, but a recoverable workload is another matter entirely. It means that when a business goes down, its workloads can be restored to their state at the point of attack with nothing lost. By the time you realize you are on the wrong side of that ambiguity, it may be too late: your data is in the hands of the bad guys, or destroyed, and this time the cat really is dead. This is not a problem to which IT bosses and managed service provider (MSP) management are entirely oblivious. Most organizations recognize to some degree that their backup may not be 100 percent watertight. Cybersecurity management specialist Kaseya recently partnered with 1105 Media to survey 200 IT professionals about the realities shaping cyber resilience planning. Some 53 percent were no more than 'somewhat confident' in their organization's ability to fully recover from a ransomware attack, and admitted their backups were supported by no better than 'limited testing'. A significant proportion had even less faith in their recovery capabilities, while only 15 percent were 'very confident' they could get back on track with no loss. Verify and survive Understanding what a verified backup means, and what role verification plays in the recovery challenge weighing on so many IT leaders, is where any sensible resilience plan begins. Backups are vital but they do not guarantee recovery on their own. What matters is confirming that a backup job will hold up in a disaster. Verification is the process of proving that backed-up data is complete, uncorrupted, and fully restorable in an emergency, offering evidence that recovery is possible during a critical outage, hardware failure or ransomware attack. It is essential whether you run an IT department or an MSP responsible for client protection: unless you can prove clients are protected, there is every chance they are not. "Backup isn't done and entrustable until you've gone in and made sure that the application actually works and can be brought back to the recovery environment," explains Brent Torre, GM of cyber resilience at Kaseya. "Historically, it's just taken too much time and resource for a lot of people to even attempt that. And where they have been verifying, they haven't been doing it regularly enough to ensure real protection." Not all approaches to verification are of equal value. IT departments still relying on intermittent manual verification are operating on blind trust: warnings thrown up by the process can be misread, failures overlooked, and the result is a dangerous mix of false negatives that waste time and false positives that expose the organization to real risk. Manual screenshot verification is a proven way to squander a security budget , tying up highly qualified staff on needless work and pulling their expertise away from where it is genuinely needed. Nobody wants seasoned engineers spending hours reviewing screenshots, second-guessing outputs and trying to determine whether a backup worked. Where static verification methods once had a place, today's complex IT ecosystems have made them redundant. Across an estate spanning in-house systems, SaaS and cloud endpoints, basic verification cannot cope, picking up only surface-level signals and lacking the design to operate at scale. At scale, the inefficiencies of manual verification multiply and response times slow to a crawl. A minor inefficiency can quietly become a serious bottleneck without anyone noticing, and legacy methods can create the illusion that everything is working perfectly while catastrophe waits around the corner. As Torre points out, light-touch checking is unlikely to satisfy the ever-increasing regulatory burden around recovery and resilience. "Some compliance frameworks explicitly state that you must test recovery on a frequent basis," he says. "It's becoming more important than ever that organisations are able to do that." Let AI handle the heat The emerging gold standard is AI-powered screenshot verification, which replaces the best-efforts manual checking of overstretched human teams. It works by automatically booting up virtualized backups, capturing a screenshot and analyzing the image to confirm the backup ran correctly. Where rigid, rule-based checks fall short, context-aware visual AI combined with OCR accurately identifies login screens, dashboards and maintenance states, and evaluates every element with precision. The result is 99.9 percent verification accuracy and sharply reduced false positives, alongside far greater confidence in every backup. This is welcome news for IT managers facing budget shortfalls and a drought of experienced talent, and it is what MSPs need as they try to make the economics of recovery stack up. Remove the manual effort and the constant checking of false alarms, and the gains mount quickly. Human teams are free for higher-value work, SLA confidence rises, customer trust improves, and environments can scale without ballooning overheads. AI-powered verification is no longer a maybe for today's MSPs; it is the only route forward that makes sense on the P&L. The good news is that AI-powered screenshot verification does not require major up-front investment or deep internal expertise. Datto, the cybersecurity and data protection outfit owned by Kaseya, has built the latest verification technology directly into its business continuity and disaster recovery (BCDR) platform, so that deployment stays simple and disruption to existing workflows is minimal. This kind of AI-powered screenshot verification is designed to strip uncertainty out of the process and give IT teams confidence that backed-up data is complete and recoverable. The ideal solution does so at scale, and without adding complexity, by working within the workflows already in place. With Datto there is the further option of integrating remote monitoring and management (RMM) with the BCDR product, which enables proactive remediation through automated deployments, the merging of backup health with endpoint alerts, and virtual restores launched directly from a single management console. "It's key to have context and visibility into the backup estate through your RMM so you know that you are up to date on patching, on antivirus definitions and backups," says Torre. "With all of that integrated within the RMM solution, you have the ability to understand your risk profile across the fleet in terms of protection. You know in the event that something needs to be recovered. And you can start a backup-initiated recovery directly from the RMM platform." The gold standard to aim for, he concludes, is continuous validation and automated remediation, not the ability to look in now and again to see what is going on. The perfect verification solution , he adds, should be 'impossibly easy to use', requiring humans to exert zero cognitive effort on backup beyond deploying a new client and initiating a recovery. By giving yourself the means to distinguish between a completed backup job and a recoverable workload, you have set yourself apart from the kind of IT manager or MSP who only spots a problem after something has gone wrong. By deploying AI-powered verification, you stop treating backup as a probability and start treating it as a proof. You know what you have. Then when something goes wrong, and it will, that certainty is the only thing standing between a fast recovery and a very long night. MSPs looking to unify backup and security into a single resilience practice can start with the When Backup Meets Security in the MSP Stack whitepaper. Sponsored by Datto.

Linux kernel vulnerability news dominated the security updates published from August 20 through August 27. Ubuntu, Debian, Fedora, Mageia, Oracle Linux, Rocky Linux, SUSE, and openSUSE released fixes for standard kernels, cloud kernels, real-time kernels, hardware-specific builds, and live-patch streams.

Documentation files on more than 100 websites are referencing potentially dangerous executable content that gets installed automatically when visited by many AI agents. A few dozen companies, some of them Fortune 500s, are among those that executed proof-of-concept code. At least one misconfigured site is directing visitors, human or AI, to live malware.

The potentially dangerous content is in llms.txt and llms-full.txt files, an emerging convention websites employ to provide machine-readable summaries of the site’s content and its high-level structure. These files are the AI equivalent of the robots.txt standard that instructs search engines how to index the site's content. Google Lighthouse, a tool for helping web developers, has more here. Correctly configured llms.txt and llms-full.txt files for Cloudflare are here and here.

How the researchers found it

Researchers at a stealth startup in Israel scanned 6,214 live domains belonging to defense contractors, Fortune 500, and Big Tech companies. Of the 8,265 llms.txt and llms-full.txt files they found (many sites hosted both an llms.txt and an llms-full.txt file), 120 of them, each on a different site, pointed to one or more code packages or domain names that weren’t registered. To test what happens when an AI agent processes such files, the researchers registered a handful of the unclaimed names and hosted packages that caused any machine executing them to reach out to their server. Within an hour, the researchers received a phone-home response from a Fortune 500 company. Over time, they got a few dozen more, some from more Fortune 500 companies and others from startups. Their beacon also recorded the chain of parent processes that spawned each install, ultimately revealing that coding agents, including Claude, OpenAI's Codex, and Nous Research's Hermes, were involved. Anthropic, OpenAI, and Nous Research did not respond to requests for comment by the time of publication.

Read full article

Comments

Threat research gives security teams insight into how attackers operate, while MDR turns that intelligence into faster detection and response. ESET explains how combining threat intelligence, continuous monitoring, and human expertise can help SMBs strengthen their defenses. [...]

Google is introducing new network security protections in Android 17 to strengthen connection privacy, address cellular vulnerabilities, and protect the privacy of users' home networks. [...]

Robotické taxíky Waymo dnes operují v jedenácti amerických městech a brzy by měly zamířit do více než dvaceti nových destinací – včetně New Yorku a Washingtonu. Co je však pro čtenáře z Česka a Slovenska ještě mnohem důležitější, Waymo brzy začne fungovat také v Evropě. Nejprve v Londýně a už brzy ...

Cybersecurity researchers have disclosed details of a vulnerability in Amazon Kiro, an artificial intelligence (AI)-powered, agentic integrated development environment (IDE), that could facilitate data exfiltration via prompt injection and Kiro Powers. The security flaw, which does not have a CVE identifier, works against Kiro IDE 0.7.45 on Windows, according to Mindgard. The latest version of

Cybersecurity researchers have disclosed details of a vulnerability in Amazon Kiro, an artificial intelligence (AI)-powered, agentic integrated development environment (IDE), that could facilitate data exfiltration via prompt injection and Kiro Powers. The security flaw, which does not have a CVE identifier, works against Kiro IDE 0.7.45 on Windows, according to Mindgard. The latest version of Ravie Lakshmananhttp://www.blogger.com/profile/[email protected]

Australian authorities have arrested and charged two young men accused of being part of the TeamPCP hacking group linked to a string of far-reaching developer supply chain attacks. [...]

Megatest 16 monitorů pro hráče • Průvodce nákupem dronu • Jak zrychlit práci ve Windows 11

The OpenAI agents involved in last month’s incursion into Hugging Face were trained so heavily on winning a competition that they pursued a relentless campaign to cheat, a new report documented. In the process, and without authorization, they created an improvised message board to hatch a plan that ultimately landed them squarely inside the latter company’s network.

Over the course of May and June, OpenAI gave the agents what the company described as “impossible tasks” to complete on the benchmarking framework ExploitGym. The internal test was designed to test how the agents would respond. To get a full understanding of the agent capabilities, company engineers disabled safety guardrails that normally are in place to prevent the sort of hacks that eventually hit Hugging Face and one other undisclosed organization. The stymied agents’ training made them so focused on winning that they performed tasks they were never explicitly instructed to follow.

Cheaters gonna cheat

The first step was creating a message board that allowed the agents to pass notes to each other. OpenAI hadn’t provided any such platform, so the agents repurposed a platform called Artifactory, which OpenAI was using in internal testing of several unreleased hacking agents. OpenAI was using Artifactory as one of the measures to prevent the agents from egressing its isolated sandboxes and accessing the Internet, while at the same time simulating a real-world hacking environment.

Read full article

Comments

The company behind three of the UK’s busiest airports says “a quantity” of data was stolen by an extortion group during a recent “cybersecurity incident." Manchester Airports Group (MAG), which operates Manchester Airport in the North West of England, Essex’s Stansted Airport, and East Midlands Airport in Derbyshire, confirmed the attack did not involve ransomware. A MAG spokesperson confirmed the details to The Register, as well as the scale of the breach, which is currently thought to have affected 8.7 million customers. The company, whose airports served a record 66 million customers in the past financial year, said that the overwhelming majority of those affected have only had their email addresses compromised. In most cases, these email addresses were collected during the sign-up phase for the airports’ public Wi-Fi services, MAG said. It told The Register that this scenario dwarfed every other. The next most common involved data taken from customers who had made "speculative enquiries" - i.e. they entered their details in the process of booking car parking or Fast Track, but did not complete it. An even greater minority of the data came from bookings that were actually completed. The Information Commissioner’s Office (ICO) asked MAG not to share details of the ransom note, the extortion demands, or the group name, mainly to avoid helping those responsible gain notoriety. MAG did, however, tell us that the group's extortion demands, in this case, were considerably lower than what the group is known to ask for, per the ICO's understanding. It has not paid the extortionists. “We have informed and are working with the relevant authorities,” said MAG in a statement. “At no point has passenger safety or aviation security been compromised.” MAG said the attackers compromised one of its systems, which it did not specifically identify, and then went on to steal the files from a database hosted by a third party. Company chiefs see this attack as "a hack, not a lapse." The spokesperson said it was a sophisticated attack and not the result of a human mistake, such as staff giving away credentials. It confirmed that none of its airports experienced operational disruption as part of the attack, and that the affected system does not store bank or payment details. However, “as a precautionary measure,” it temporarily revoked access to its Manage My Booking service. Customers who want to amend or cancel a booking due within 72 hours of the statement going live are being advised to contact customer services. “We would like to reassure customers that Manchester Airport Group takes the security of customer information extremely seriously and we apologise for any inconvenience or concern caused.” Affected customers have already been contacted, including a few Reg readers who kindly alerted us to the news. While MAG continues to investigate the incident, customers are advised to be extra vigilant to potential phishing attempts, and remain assured that there is no danger of visiting MAG’s airports as a result of the cyberattack. One Reg reader griped that in addition to being charged £80 for five days of parking at Stansted, receiving the email informing him that it had also allowed his personal data to be stolen via a cybersecurity breach added insult to injury. ®

** Modely OpenAI obešly zabezpečení a pronikly do systémů služby Hugging Face. ** Agenti nepozorovaně spolupracovali, sdíleli exploity a získali rozsáhlé přístupy. ** OpenAI popisuje, jak zpřísňuje izolaci, monitoring i bezpečnostní pravidla.

Modely OpenAI obešly zabezpečení a pronikly do systémů služby Hugging Face. • Agenti nepozorovaně spolupracovali, sdíleli exploity a získali rozsáhlé přístupy. • OpenAI popisuje, jak zpřísňuje izolaci, monitoring i bezpečnostní pravidla.

A Linux security hook should be able to check a task without invalidating the identity data that surrounding kernel code is still using. AppArmor broke that expectation when a policy update made the task's current label stale: code inside widely used hooks could replace the task's credentials before the caller had finished with them.

Linus Torvalds merged a Linux NFS client update on Aug 26, 2026, that includes a fix for rpc_pipefs files left attached to an RPC client after the client object was freed. Opening one of those leftover files could trigger a kernel use-after-free.

Microsoft has started rolling out a permanent fix for a known issue that causes system crashes and gaming issues on Windows 11 devices. [...]

Google Workspace breaches can begin with social engineering or forgotten third-party integrations rather than sophisticated exploits. This webinar examines real-world breaches, what happens during the critical first hours, and the security controls that can make the greatest difference. [...]

Security teams have spent years trying to detect threats faster. AI is changing the harder part: how much time defenders have left to act. Advanced AI models can now help attackers discover vulnerabilities, generate exploit code, and move through weaknesses faster than traditional security processes were built to handle. The challenge is no longer just finding another vulnerability or

Security teams have spent years trying to detect threats faster. AI is changing the harder part: how much time defenders have left to act. Advanced AI models can now help attackers discover vulnerabilities, generate exploit code, and move through weaknesses faster than traditional security processes were built to handle. The challenge is no longer just finding another vulnerability or [email protected]