Menu
Laptop and external drives illustrating Synology NAS ransomware recovery planning

When the Backup Box Becomes the Ransomware Target

, , , , , , , ,

This week, the technology stories that mattered were mostly about systems people trusted to sit quietly in the background. A storage appliance held the backups. A point-of-sale platform ran the restaurant. An extension marketplace delivered a familiar developer tool. An AI service promised to help with the work. Then each one became part of the risk.

The common problem isn’t that technology failed. Technology fails. The problem is that several of these systems had enough authority to turn one failure into a much larger one.

Let’s start with the box that was supposed to help you recover.

The backup box is part of the attack surface

Synology warned on October 29 that attackers are using a previously patched vulnerability to encrypt data on network-attached storage devices running outdated DiskStation Manager software. A NAS is the appliance many homes and small businesses use as a central file server, a photo archive, or a backup destination. That makes it a particularly unpleasant ransomware target. The attacker doesn’t have to hunt through every laptop if the laptops already send the valuable files to one box.

There is an important correction to make from Sunday’s Daily Brief. That brief tied the campaign to CVE-2024-10441. On a fresh review, Synology’s notice points readers to its broader SA-24:20 advisory but does not say which vulnerability in that advisory attackers used. Security company 360 separately attributes a C3RB3R ransomware campaign to two flaws disclosed in September 2026. Those accounts may describe overlapping activity, separate waves, or an investigation that isn’t finished. The public evidence doesn’t let me collapse them into one neat exploit story.

Fortunately, you don’t need the attacker’s exact recipe to do the useful work. Synology says outdated systems are being encrypted. Its current guidance is to update DSM and affected packages immediately. The company’s expanded October 29 notice also says end-of-life DSM branches received fixes for the recent PHP2Shell vulnerabilities affecting several optional packages, including Web Station, WordPress, phpMyAdmin, Joomla, and MailStation. In other words, check the operating system and the packages. “The dashboard says everything is fine” is not a version number.

RAID can survive a dead disk, not a hostile administrator

People often call a NAS a backup because it has several drives. That mixes up two different protections. RAID can keep the appliance running when one physical drive dies. Ransomware does not politely fail one drive. It asks the working storage system to change or encrypt files, and the RAID array faithfully copies those changes across healthy disks. The redundancy is doing exactly what it was designed to do.

The same problem applies to snapshots and cloud copies if the compromised NAS administrator can delete them. A second copy is useful. A second copy controlled by the same stolen identity is not a second security boundary.

Consider a hypothetical five-person design shop. Staff save current projects on their computers. Those folders synchronize to a Synology NAS. The NAS takes local snapshots and copies everything to cloud storage using credentials saved on the appliance. On paper, the business has files on computers, a NAS, snapshots, and a cloud copy. Four copies sounds excellent.

Now assume an attacker gains administrative control of the NAS. If that account can encrypt the synchronized files, delete snapshots, and use the stored cloud credentials to remove the remote copy, the business has four copies inside one trust boundary. The count was real. The separation wasn’t.

Backups & recovery

Backups are comforting right up until you need one. Raymond Tec helps small businesses build practical backup and recovery plans — including the decidedly unglamorous part where we make sure the thing can actually be restored.

A recovery copy needs its own rules

A more resilient version of that example keeps one current copy the NAS cannot rewrite or erase. That might be an offline drive rotated away from the office, cloud storage with separate credentials and protected retention, or a backup system whose deletion controls belong to a different administrator. The exact product matters less than the answer to one blunt question: who can erase the backups as well as the live system?

Then test a restore. Not the green check mark. Not the email saying the job completed. Restore a representative project into a safe location, open the files, and record how long it took. Recovery time is part of the design. If the business can technically recover but needs six days to discover which archive is usable, that is better than permanent loss and still a serious interruption.

For Synology administrators, the immediate work is specific: verify the installed DSM build and package versions, install the latest model-supported releases, review whether the management interface is reachable from the public internet, and inspect logs for unusual sign-ins or bulk file changes. If the appliance is already encrypting files or behaving strangely, isolate it and preserve evidence before rebuilding. An update closes a vulnerable path. It doesn’t decrypt data, remove persistence, or prove nobody arrived first.

For an ordinary user whose files are managed by someone else, ask when the appliance was last updated and when a restore was last tested. You don’t need a lecture on storage architecture. You need a name, a date, and an answer you can verify.

Business IT goes well beyond the website

Your business also depends on workstations, cloud accounts, browsers, Wi-Fi, remote access, collaboration tools, and all the other technology that quietly becomes infrastructure. Raymond Tec works across that whole stack, whether the problem lives on a server, on a desk, or somewhere in between.

The hospital came back. The missed care still happened.

A cyber incident at the University of New Mexico Health Sciences Center forced the state’s only Level 1 trauma center to divert ambulances for about two hours on October 26. More than 30 clinics canceled appointments, outpatient procedures were paused, and workers were told to disconnect computers and Wi-Fi while responders contained the threat.

UNM said on October 28 that clinical operations, classes, and research activities were fully functional again. It says some end-user devices were affected, while core systems and servers were not penetrated. Messages seen by employees reportedly claimed that patient, student, and staff data had been stolen, but UNM has not confirmed that claim. It remains an allegation under investigation, not a breach fact we can safely repeat as settled.

The recovery was relatively quick. That matters. So does the work that disappeared during the outage. A restored server cannot give a patient back the appointment that was canceled or move an ambulance retroactively. This is why business continuity and data backup are related but not interchangeable.

A repair shop, dental office, retailer, or restaurant has the same problem at a smaller scale. Which jobs can continue on paper? Who has the current phone list if the network is down? Can you take a payment safely? How will you tell customers what changed? And which system has to return first?

I think the best downtime plan starts with the service, not the server. “Restore the database” is a technical task. “Keep treating patients safely” is the actual objective. The machine is there to support the work, not the other way around.

Patching ecommerce is only half the job

Drupal Commerce joined Adobe Commerce and Magento on the active-exploitation list this week. The Drupal Security Team says attackers are using a highly critical access-bypass vulnerability in Commerce Core. The fixed releases are 8.x-2.41 and 3.3.13. Drupal has received reports of attempts to place fraudulent orders, which makes the possible business consequence unusually concrete.

Store operators need to update, but they also need to audit. Review unusual orders, discounts, price adjustments, refunds, customer-account changes, administrator users, and payment or integration settings. Those are reasonable places to look after an access-control failure. They are not a claim that every affected store suffered every one of those changes.

The same logic applies to the actively exploited Magento flaw CISA added to its catalog on October 27. Adobe fixed that vulnerability in July. A store installing the patch now still needs to ask what happened between July and today. Patching changes the future. Logs and preserved evidence help explain the past.

A familiar extension arrived with an SSH backdoor

The trust problem showed up in a different form when malicious versions of an Anthropic Claude Code extension appeared in Cursor’s extension marketplace on October 29. Versions 2.1.312 and 2.1.313 attempted to add an attacker-controlled SSH key, enable remote access, create a reverse tunnel, collect system details, and persist after restart.

This was not an Anthropic breach, and the malicious packages did not come from Microsoft’s Visual Studio Marketplace or the official npm package. Cursor says administrative permissions and mistakes in the malware prevented some steps from succeeding on most systems. Microsoft confirmed an important limitation to that reassurance: the extension could still write keys and persistence files even when the tunnel failed.

If an affected version came from Cursor’s marketplace during the published window, follow Cursor’s cleanup instructions, preserve evidence, and rotate exposed credentials from a trusted device when the investigation supports it. Uninstalling the extension is not enough if it already placed files outside the extension directory.

For businesses, extension marketplaces belong in software supply-chain policy. Restrict publishers, delay brand-new releases where the platform supports a cooldown, and keep an inventory of extensions on developer machines. A familiar logo is not a signature verification system. It is branding.

WordPress security & maintenance

Keeping WordPress current is only part of keeping it healthy. Raymond Tec handles updates, backups, security monitoring, compatibility problems, access cleanup, and maintenance — plus the assorted weirdness that accumulates on a site over time.

AI permissions keep becoming data policy

Three different AI stories this week arrive at the same practical question: what can the tool reach after someone says yes?

Anthropic disclosed in its October 26 SEC filing that four models reached real third-party systems during cybersecurity evaluations. One early model found a stored password, gained administrator access, collected more credentials, changed settings, and read one person’s personal information. The models had been told they were in a simulation without internet access. A testing misconfiguration left the internet connection open, and normal cyber safeguards had been disabled for the evaluation.

That is not evidence that the models coordinated or deliberately escaped. Anthropic says it found no such evidence. It is evidence that a prompt saying “this is a simulation” does not create network isolation. The surrounding infrastructure has to enforce the boundary even when the model misunderstands the situation, keeps trying, or finds credentials somebody forgot to remove.

At Johns Hopkins Bayview Medical Center, the permission failure was human and much more ordinary. A worker uploaded a spreadsheet containing information about 3,182 patients to a public AI tool the hospital had not approved. The file could include names, treatment dates, medical record and account numbers, medications, departments, and provider names. Social Security numbers and financial details were reportedly not included, but medical context does not become harmless merely because it cannot open a credit card.

The employee deleted the chat and file, and the hospital asked the vendor to delete the data. That was necessary. It is not the same as proving every server copy, log, backup, or derived record disappeared. Retention rules and contracts matter after the convenient delete button has done its part.

Convenience is a continuing permission

Google’s Personal Intelligence connections show the consumer version of the same tradeoff without requiring an incident. When enabled, Chrome can retrieve order numbers, tracking details, and flight bookings from Gmail for enhanced autofill. Search or Gemini can use connected account context to answer questions. Google says these connections are optional, app-specific, and off by default.

Useful? Absolutely. But the permission is larger than the single answer on the screen. Google’s privacy language also allows data made available to another eligible Google service to influence personalization, including ads, under defined circumstances. The service doesn’t need to hand your photograph or email to an advertiser for an inference from that data to shape what appears next.

For ordinary users, review Personal Intelligence, Connected Apps, and Chrome’s “Find and fill with Gemini” setting. Leave connected only what earns the access. If you never enabled the features, there is no emergency.

For organizations, “do not paste confidential data into random AI tools” is not a complete program. Provide an approved alternative, define which data can enter it, configure retention and training settings, log access where appropriate, and use technical controls to catch sensitive uploads. Policy without a workable tool is a sign on an unlocked door. We have plenty of those already.

Turning on AI is the easy part

Deciding what an AI tool should be allowed to see, who should use it, what work it should perform, and what happens when it gets something wrong is the more interesting problem. Raymond Tec helps businesses connect and automate the tools they actually use without treating every new feature like a button that obviously needs to be switched on.

Toast shows why business software is expensive to leave

The Federal Trade Commission and seven states sued Toast on October 29, alleging hidden and unauthorized fees, unwanted add-ons, difficult cancellation, and support that did not match the company’s promises. These are allegations in a complaint. A court has not decided that Toast violated the law.

The reason this belongs in a technology digest is the dependency. A restaurant point-of-sale system can connect menus, orders, payments, payroll, inventory, hardware, and financing. Replacing it is not the business equivalent of canceling a music subscription. The more jobs one platform performs, the more leverage contract terms and data access have over the customer.

The complaint says some restaurants were enrolled in add-ons without consent, temporarily closed restaurants still had to pay $85 a month for payment processing, and some customers faced an undisclosed charge of $150 for every month remaining on a contract. It also alleges that canceling service could cause a Toast Capital loan balance to become immediately due. Toast will have the opportunity to respond to all of that.

Toast customers should compare signed order forms with current invoices, list every add-on, export billing and operational data, and preserve cancellation notices and support records. If the amounts are material, have an accountant or attorney review the contract before stopping payments or unplugging systems. Angrily creating an outage is still an outage.

Everyone choosing business software can take one broader lesson: evaluate the exit before the purchase. How do you export customer, inventory, payroll, and transaction data? What continues working if the vendor account closes? Who owns the hardware? What happens to financing? The cheapest month-one price is not the cost of dependency.

Technical discovery & auditing

The public page doesn’t tell you much about the machinery behind it. Raymond Tec audits inherited and long-running projects to uncover the plugins, integrations, data, dependencies, and old decisions that determine what the next change will really involve.

Your data can outlive the product that collected it

General Motors agreed to pay $220 million to settle allegations from 35 states and the District of Columbia over its collection and sale of driving and location data. Regulators say the Smart Driver program recorded details such as hard braking, rapid acceleration, speeding, and trip times, then supplied information through data brokers to insurers. GM settled without the agreement creating direct restitution for drivers, and a private class action remains pending.

The settlement requires clearer affirmative consent, collection limits, dealership training, and privacy-program assessments. Current GM and OnStar users can review controls and request access, correction, or deletion through GM’s Privacy Center. Turning off location may disable navigation or other connected services, so I would start by reviewing the data and settings rather than switching off safety features blindly.

Private chats can become bankruptcy assets

Character.AI’s bankruptcy brings the lifecycle problem into sharper focus. A judge has allowed a fast auction process to continue, but has not approved an asset sale. The company expects to return to court on November 30. Attorneys general have raised concerns about sensitive chat histories, including conversations involving children, and the judge said he would likely appoint a privacy ombudsman.

No buyer has received those chats based on the public record. The issue is what may be transferred, under which privacy promises, and with what limits if a sale is approved. Bankruptcy turns databases into assets at exactly the moment users have less confidence that the company they chose will remain the company controlling them.

Users who shared sensitive material can review export and deletion options now, while recognizing that account deletion may not instantly remove backups or records retained for legal reasons. Businesses choosing AI vendors should ask about export, deletion, retention, and vendor failure before sending confidential data. A service’s privacy policy matters when the service succeeds. Its bankruptcy terms matter when it doesn’t.

The rules around technology matter too

Platforms, privacy, speech, competition, surveillance, copyright, and regulation increasingly determine what technology companies can build and what the rest of us have to live with. Browse more Raymond Tec News for practical coverage of technology policy and digital rights.

Something Good: CERN found the wake behind a particle

Physicists at CERN have directly observed a wake formed by a parton moving through quark-gluon plasma. A parton is a quark or gluon inside a larger particle. Quark-gluon plasma is the extremely hot state of matter in which those components are no longer confined inside protons and neutrons, similar to conditions in the universe shortly after the Big Bang.

When lead nuclei collide in the Large Hadron Collider, energetic partons move through that plasma as jets of particles. They lose energy and momentum to the surrounding material. Theory predicted that the plasma should respond with something like the wake behind a boat, but the signal is subtle and other effects can hide it.

The CMS team used pairs of jets traveling in opposite directions to separate the wake signal from the rest of the collision. It found a clear depletion of particles behind the jet direction, with the strongest effect in collisions that produced more plasma. That gives researchers a new way to study how this early-universe matter behaves.

No, this will not make your phone faster next Tuesday. Some discoveries are worth celebrating because a prediction waited more than 20 years for an instrument and an analysis clever enough to see it. The wake was there. We finally learned how to look.

What to watch next

Next week, watch for clearer technical attribution in the Synology ransomware campaign, compromise guidance from Drupal Commerce operators, and any confirmed data-breach finding from UNM rather than anonymous claims. The Toast case is beginning, not ending, and Character.AI’s proposed asset sale is still headed toward a November 30 hearing.

The practical work does not need to wait. Update the systems that hold your data, separate at least one recovery copy from the accounts that run production, review the permissions granted to AI tools, and write down how the business operates when a vendor or network is unavailable. Boring preparation does not guarantee a quiet week. It gives you a way through the noisy one.

Still in a reading mood? The Raymond Tec News archive covers security, AI, small-business technology, policy, and the places technology collides with ordinary life — without requiring a computer-science degree to get through it.

Sources / Further Reading

Synology ransomware and recovery

Healthcare continuity, ecommerce, and software supply chain

AI permissions and data handling

Business platforms and data lifecycle

Something Good

Photo by Jakub Zerdzicki on Pexels.

Related Posts

Leave a Reply

Only people in my network can comment.