Screen-Accurate Lightsaber As a Practical Effect

The lightsaber was one of the coolest and most iconic visual effects from the original Star Wars, and people have been trying to get that particular piece of movie magic off the silver screen for about 40 years now. [HeroTech] seems to have cracked the code with their “Impossible Lightsaber”— it’s fully retractable, fully lit, and able to hit things (lightly), all while fitting into a replica prop handle.

The secret is… well, there’s more than one secret, here. The blade itself is made out of a “magician’s cane”, which is a coil of plastic that can spring outwards on demand for magic tricks. Hidden inside of it is a strip of LED lights. Of course one strip of LEDs would not be omnidirectional, and the magician’s cane is pretty floppy, but both of those problems are solved by the same idea: “I’ll try spinning. That’s a good trick.”

The spin-stabilized blade holds up to being waved around much better, and apparently the gyroscopic forces it induces are actually lore-accurate. (Who knew?) Of course fitting a motor to spin the “blade”, and another to winch it back in, along with the circuitry and batteries to drive them was no mean feat. It’s impressive they fit it all inside the replica handle; even more impressive that they fit a speaker so this prop even makes the iconic sound effects. We always wanted to see a stage production of William Shakespeare’s Star Wars, and this gives us verily, a new hope.

[HeroTech] isn’t done yet– while seemingly impossible, this lightsaber isn’t perfect, as it’s not rugged enough for full dueling. It’s also not easy to put together, and apparently can’t handle the delicate attentions of airline baggage handlers. So ruggedization and a bit of design-for-assembly are on the table for the next version. Sadly the project is not open source; they are releasing the build files to subscribers only. Given how much work must have gone into iterating to get to this point, that is disappointing, but understandable. Everybody needs to make a living, after all.

If this project seems familiar, it’s because we featured a much-bulkier previous iteration last year.

You may prefer your lightsabers to match the movie version in effects instead of visuals; if that’s the case, check out this saber that uses HHO to cut through a steel door.

youtube.com/embed/0QArAMtBGx8?…

hackaday.com/2025/11/07/screen…

Medium Format, 3 GigaPixel Camera Puts It All On the Line (Sensor)

It’s a bit of a truism that bigger sensors lead to better pictures when it comes to photography. Of course everyone who isn’t a photographer knows that moar megapixles is moar better. So, when [Gigawipf], aka [Yannick Richter] wanted to make a camera, he knew he had to go big or go home. So big he went: a medium format camera with a whopping 3.2 gigapixel resolution.

Now, getting a hold of a sensor like that is not easy, and [Yannick] didn’t even try. The hack starts by tearing down a couple of recent-model Kodak scanners from eBay to get at those sweet CCD line sensors. Yes, this is that classic hack: the scanner camera. Then it’s off to the oscilloscope and the datasheet for some serious reverse-engineering to figure out how to talk to these things. Protocol analysis starts about 4 minutes in of the embedded video, and is worth watching even if you have no interest in photography.

As for what the line sensor will be talking to, why, it’s nothing other than a Rasberry Pi 5, interfacing through a custom PCB that also holds the stepper driver. Remember this is a line sensor camera: the sensor needs to be scanned across the image plane inside the camera, line by line, just as it is in the scanner. He’s using off-the-shelf linear rails to do that job. Technically we suppose you could use a mirror to optically scan the image across a fixed sensor, but scanner cameras have traditionally done it this way and [Yannick] is keeping with tradition. Why not? It works.

Since these images are going to be huge an SD card in the Pi doesn’t cut it, so this is perhaps the only camera out there with an NVMe SSD. The raw data would be 19 GB per image, and though he’s post-processing on the fly to PNG they’re still big pictures. There probably aren’t too many cameras sporting 8″ touchscreens out there, either, but since the back of the thing is so large, why not? There’s still a CSI camera inside, too, but in this case it’s being used as a digital viewfinder. (Most of us would have made that the camera.) The scanner cam is, of course, far too slow to generate its own previews. The preview camera actually goes onto the same 3D-printed mount as the line sensor, putting it onto the same focal plane as the sensor. Yes, the real-time previews are used to focus the camera.

In many ways, this is the nicest scanner camera we’ve ever featured, but that’s perhaps to be expected: there have been a lot of innovations to facilitate this build since scanner cams were common. Even the 3D printed and aluminum case is professional looking. Of course a big sensor needs a big lens, and after deciding projector lenses weren’t going to cut it, [Yannick] sprung for Pantax 6×7 system lenses, which are made for medium format cameras like this one. Well, not exactly like this one– these lenses were first made for film cameras in the 60s. Still, they offer a huge image, high-quality optics, and manual focus and aperture controls in a format that was easy to 3D-print a mount for.

Is it the most practical camera? Maybe not. Is it an impressive hack? Yes. We’ve always had a soft-spot for scanner cameras, and a in a recent double-ccd camera hack, we were lamenting in the comments that nobody was doing it anymore. So we’re very grateful to [Manawyrm] for sending in the tip.

youtube.com/embed/KSvjJGbFCws?…

hackaday.com/2025/11/07/medium…

2025 Component Abuse Challenge: Pushing a 555 to the Limit

The humble 555 timer has its origins back in the early 1970s as the NE555, a bipolar integrated circuit. Over the years it has spawned a range of derivatives, including dual versions, and ones using CMOS technology. Have these enhancements improved the performance of the chip significantly? [MagicWolfi] has been pushing the envelope in an effort to see just how fast an astable 555 can be.

The Microchip MIC1555 may be the newest of the bunch, a 5-pin CMOS SOT-23 which has lost the frequency control and discharge pins of the original. It’s scarcely less versatile though, and it’s a fine candidate for an oscillator to push. We see it at a range of values for the capacitor and resistor in an astable configuration, each of which is tested across the supply voltage range. It’s rated as having a maximum frequency of 5 MHz, but with a zero Ohm resistor and only the parasitic capacitance of an open circuit, it reaches the giddy heights of 9.75 MHz. If we’re honest we find this surprising, but on reflection the chip would never be a first choice for super-fast operation.

We like it that someone’s managed to tie in the 555 to the contest, and given that it still has a few days to run at the time of writing, we’re hoping some of you might be inspired to enter one of your own.

hackaday.com/2025/11/07/2025-c…

Hackaday Podcast Episode 344: Board with Lasers, Op-Amp Torture, and Farewell Supercon 9

Hackaday Editors Tom Nardi and Al Williams spent the weekend at Supercon and had to catch up on all the great hacks. Listen in as they talk about their favorites. Plus, stick around to the end to hear about some of the highlights from their time in Pasadena.

If you’re still thinking about entering the Component Abuse Contest, you’re just about out of time. Need some inspiration? Tom and Al talk about a few choice entries, and discuss how pushing parts out of their comfort zone can come in handy. Do you make your own PCBs? With vias? If you have a good enough laser, you could. Or maybe you’d rather have a $10 Linux server? Just manage your expectations. The guys both admit they aren’t mechanical geniuses and, unlike [4St4r], aren’t very good at guessing sounds either. They round up with some 3D printing projects and a collection of quick hacks.

Check out the links below if you want to follow along, and as always, tell us what you think about this episode in the comments!

html5-player.libsyn.com/embed/…

Download in DRM-free MP3 no PDP-1, 3D Printer, or lasers needed to listen.

Where to Follow Hackaday Podcast

Places to follow Hackaday podcasts:

• iTunes
• Spotify
• Stitcher
• RSS
• YouTube
• Check out our Libsyn landing page

Episode 344 Show Notes:

News:

• Component Abuse Challenge

What’s that Sound?

• Congratulate [4St4r] for guessing last week’s sound. Want a Hackaday Podcast T-shirt? Try your luck next week.

Interesting Hacks of the Week:

• Is This The Last PCB You’ll Ever Buy?
• 2025 Component Abuse Challenge: Weigh With A TL074
• Repurposing Dodgy Android TV Boxes As Linux Boxes
• Reproduced And Recovered: The First Chinese Keyboard-based MingKwai Typewriter
• Print In Place Pump Pushes Limits Of Printing
• 3D Printering: Liquid-Filled Filament Was Not On Our Bingo Card

Quick Hacks:

• Tom’s Picks:
  
  • CardFlix: NFC Cards For Kid-Friendly Streaming Magic
  • Print-and-Clamp: Rubber Band PCB Stand Slides Into Duty
  • SolidWorks Certification… With FreeCAD?

  
  

• Al’s Picks:
  
  • Multitasking On The Humble Z80 CPU
  • Simple Device Can Freeze Wi-Fi Camera Feeds
  • 2025 Component Abuse Challenge: An Input Is Now An Output

  
  

Supercon 9:

• Crafting the Final Frontier Keynote
• Announcing The 2025 Hackaday Superconference Communicator Badge

hackaday.com/2025/11/07/hackad…

An LED Projector as a Lighting Effect

If you had an array of high power addressable LEDs, how would you project them onto a wall? Perhaps you’d use a Fresnel lens, or maybe an individual lens on the top of each. [Joo] faced this problem when making a lighting effect using just such an array, and the solution they came up with used both.

The problem facing a would-be LED array projector is that should the lens be too good, it will project the individual points of light from the LEDs themselves, when a more diffuse point is required. Thus the Fresnel required the aid of a separate array of lenses, resin printed in one in clear plastic. From this we get some useful tips on how to do this for best lens quality, and while the result is not quite optically perfect, it’s certainly good enough for the job in hand.

The linked Printables page comes with all you need to make the parts, and you too can have your own projected LED effect. Now we want one, too! Perhaps we really need our own Wrencher signal instead.

hackaday.com/2025/11/07/an-led…

This Week in Security: Bogus Ransom, WordPress Plugins, and KASLR

There’s another ransomware story this week, but this one comes with a special twist. If you’ve followed this column for long, you’re aware that ransomware has evolved beyond just encrypting files. Perhaps we owe a tiny bit of gratitude to ransomware gangs for convincing everyone that backups are important. The downside to companies getting their backups in order is that these criminals are turning to other means to extort payment from victims. Namely, exfiltrating files and releasing them to the public if the victim doesn’t pay up. And this is the situation in which the Akira ransomware actors claim to have Apache’s OpenOffice project.

There’s just one catch. Akira is threatening to release 23 GB of stolen documents, which include employee information — and the Apache Software Foundation says those documents don’t exist. OpenOffice hasn’t received a demand and can’t find any evidence of a breach. It seems likely that Akira has hit some company, but not part of the Apache Software Foundation. Possibly someone that heavily uses OpenOffice, or even provides some level of support for that application. There is one more wrinkle here.

Since Apache OpenOffice is an open source software project, none of our contributors are paid employees for the project or the foundation…

First off, there are plenty of open source projects that have employee contributors, and it’s quite odd to imply otherwise. But second, for something as important as an office suite, this is a rather startling statement: there are no paid employees working on the OpenOffice code base.

NPM Typosquat Sophistication

There’s another NPM typosquatting campaign, which is barely news at this point. This one is newsworthy because these malicious packages use multiple layers of obfuscation, and lived on NPM for over four months. They use a clever bit of social engineering during package installation, in the form of a fake CAPTCHA prompt. The idea is that it makes the user less suspicious of the package, and also gives a legitimate reason for network access. But in reality, requiring user interaction defeats any automated analysis efforts.

The first layer of obfuscation consists of an eval() call with a bunch of decoder functions and an ugly encoded string. The result from that set of functions is URL-encoded and needed decoding, followed by an XOR with a key value. And finally, the executable function that finally emerges uses switch/case statements and hard-to-read values. It’s just a web to work through.

The payload behavior is boring in comparison, looking for any credentials on the system and uploading them to a remote server. It also checks for interesting browser cookies and passwords in the password manager, and any authentication tokens it can find.

WordPress Plugin Problems

[István Márton] at Wordfence has the story on a pair of WordPress plugins with severe vulnerabilities, effecting a whopping 500,000 sites combined. Up first is AI Engine, with 100,000 installs. This plugin has an unauthenticated URL endpoint that can expose a bearer token, which then allows access to the MCP endpoint, and arbitrary control of users. The good news here is that the plugin is not vulnerable by default, and requires the “No-Auth URL” setting to be configured to be vulnerable.

The other plugin is Post SMTP, with 400,000 installs. It replaces WordPress’s PHP email handling, and one of the features is the ability to view those emails from the logs. The problem was that before 3.6.1, viewing those email logs didn’t require any permissions. At first blush, that may seem like a medium severity problem, but WordPress is often configured to allow for password resets via emailed links, which means instant account takeover. Both issues have been fixed, and releases are available.

React Native CLI and Metro

A combination of the React Native CLI package and the Metro development server exposed React Native developers to a nasty 9.8 CVSS Remote Code Execution (RCE) CVE. The first element of this vulnerability is the fact that when Metro opens ports for hosting development work, it doesn’t bind to localhost, but listens on all interfaces by default.

When a new Reactive Native project is created without using a framework, some boilerplate code is run as part of the initialization. The end result is that /open-url handler is added to the project, and this handler calls open() with an outside string from the URL. It’s not hard to imagine how this can be abused for arbitrary code injection.

KASLR

Let’s talk about address randomization. Specifically, Kernel Address Space Layout Randomization (KASLR). It’s one of the defenses against turning an arbitrary memory write into a working exploit. If an attacker can’t predict where kernel objects will be in memory, twiddling bits is more likely to crash the system than result in code execution. It’s great in theory. The problem is that it doesn’t necessarily exist in reality.

That’s the story from [Seth Jenkins] at Google’s Project Zero, who was looking for ways to crack Pixel phones. It turns out that memory hotplugging is supported by Linux on Android, and that potential hotplug memory needs a lot of room in the linear memory map. So much room that it’s impractical to also randomize that layout. So while we still technically have KASLR protecting the kernel from attacks, there’s a really big gotcha in the form of the linear memory map.

Bits and Bytes

If you want a really deep dive into how BLE works, and how to investigate an existing BLE connection with an SDR, [Clément Ballabriga] from Lexfo has the scoop. It is significantly more complicated than you might expect, particularly since BLE uses frequency hopping, and a wide enough range of frequencies that your SDR almost certainly can’t capture them all at once. That means breaking a tiny part of the signal security, in order to accurately predict the frequency hops.

Cisco’s Unified Contact Center Express (UCCX) has several vulnerabilities that allows an attacker to run code as root. One vulnerability is in handling arbitrary file uploads by the Java Remote Method Invocation system. Another is an authentication bypass that can be exploited by coercing the target system to use a malicious remote server as part of the authentication process. Fixes are available, and so far it doesn’t look like these flaws have been used in the wild.

And finally, there’s the November Android security bulletin, that fixes CVE-2025-48593, a logic error in security updates in apexd.cpp that can lead to escalation of privilege.

I’ve seen this flaw conflated with CVE-2025-38593, a Bluetooth vulnerability recently fixed in the Linux kernel. This is a medium severity race condition in the kernel that can lead to a double-free and a system crash. There doesn’t seem to be a way to turn this into an RCE, as is reflected by its CVSS of 4.7.

hackaday.com/2025/11/07/this-w…

Pi-Powered Camera Turns Heads and Lenses In Equal Measure

Have you ever seen photos of retro movie sets where the cameras seem to be bedazzled with lenses? Of course you can only film via one lens at a time, but mounting multiple lenses on a turret as was done in those days has certain advantages –particularly when working with tiny M12 lenses, like our own [Jenny List] recently did with this three-lens, Pi-zero based camera.

Given that it’s [Jenny], the hardware is truly open source, with not just the Python code to drive the Pi but the OpenSCAD code used to generate the STLs for the turret and the camera body all available via GitHub under a generous CC-BY-SA-4.0 license. Even using a cheap sensor and lenses from AliExpress, [Jenny] gets good results, as you can see from the demo video embedded below. (Jump to 1:20 if you just want to see images from the camera.)

The lenses are mounted to a 3D printed ring with detents to lock each quickly in place, held in place by a self-tapping screw, proving we at Hackaday practice what we preach. (Or that [Jenny] does, at least when it comes to fasteners.) Swapping lenses becomes a moment’s twist, as opposed to fiddling with tiny lenses hoping you don’t drop one. We imagine the same convenience is what drove turret cameras to be used in the movie industry, once upon a time.

youtube.com/embed/UntZmKj_IJE?…

hackaday.com/2025/11/07/pi-pow…

Artsy and Durable Recycling from a Heat Press

Plastic recycling is something that many of us strive to accomplish, but we often get caught up in the many hurdles along the way. [Brothers Make] are experienced in the world of plastic recycling and graced us with a look into a simple and reliable way to get consistent thin sheets of durable plastic. Using a common T-shirt press and a mixture of plastic scraps, you can get the process down quickly.

Summarizing the process is pretty easy due to its simplicity. You take a T-shirt press, put some Teflon baking sheets on both sides of some plastic scraps, and then press. Repeating this a couple of times with different colored plastic will get you a nice looking sheet of usable sheets for any purpose you could dream of. Thicker pieces can have some life changing applications, or as simple as guitar picks, as shown by [Brothers Make].

Make sure to try out this technique yourself if you have access to a press! Overuse of plastic is a widely known issue, and yet it feels like almost no one attempts to solve it. If you want a different kind of application, try making your own 3D printing filament out of recycled plastic!

youtube.com/embed/xPGunwTUMSE?…

Thanks to [George Graves] for the tip!

hackaday.com/2025/11/07/artsy-…

La Cina supererà gli USA nella corsa all’AI! Il CEO di Nvidia avverte “siamo bloccati dal cinismo”

Il CEO di Nvidia, Jensen Huang, ha avvertito che la Cina sconfiggerà gli Stati Uniti nella corsa all’intelligenza artificiale (IA).

Il FT ha riportato che Huang, al Future of AI Summit ospitato dal Financial Times (FT) quel giorno, ha affermato: “I paesi occidentali, inclusi Stati Uniti e Regno Unito, sono bloccati nel cinismo. Abbiamo bisogno di maggiore ottimismo“. Citando le nuove normative sull’IA in fase di elaborazione negli Stati Uniti, Huang ha avvertito: “Potrebbero esserci 50 nuove normative (in tutti i 50 stati)“.

Ha sostenuto che questo contesto normativo sta minando la competitività occidentale. Le sue dichiarazioni sono arrivate dopo che l’amministrazione Donald Trump ha mantenuto le restrizioni sull’esportazione dei più recenti semiconduttori per l’IA di Nvidia in Cina.

Huang ha aggiunto: “L’elettricità è (praticamente) gratuita in Cina” e che la Cina sta creando un ambiente in cui le aziende locali possono utilizzare chip di IA prodotti internamente come alternativa molto più economica ai chip di IA di Nvidia, espandendo i sussidi energetici per le aziende tecnologiche.

È noto che i chip di IA cinesi sono significativamente meno efficienti dal punto di vista energetico rispetto ai chip Nvidia. Tuttavia, i sussidi energetici del governo cinese hanno contribuito a compensare in parte questa carenza. Infatti, secondo quanto riferito, la Cina ha recentemente aumentato i sussidi energetici per importanti aziende tecnologiche come ByteDance, Alibaba e Tencent.

Il Wall Street Journal (WSJ) ha riportato il 3 che l’amministratore delegato di NVIDIA, vicino al presidente Trump, ha cercato di convincere il presidente Trump a consentire l’esportazione in Cina dell’ultimo semiconduttore per l’intelligenza artificiale di Nvidia, “Blackwell“, poco prima del vertice USA-Cina del 30 del mese scorso, ma i suoi collaboratori hanno bloccato la sua richiesta.

Huang ha sottolineato l’importanza del mercato cinese, affermando: “Circa la metà dei ricercatori mondiali nel campo dell’intelligenza artificiale si trova in Cina“. Ha inoltre sostenuto che le esportazioni verso la Cina devono essere consentite per garantire che l’ecosistema aziendale cinese dell’intelligenza artificiale continui a dipendere dai semiconduttori di Nvidia.

Alcuni ipotizzano che l’amministrazione Trump possa limitare l’uso di semiconduttori per l’intelligenza artificiale all’avanguardia, come Blackwell, solo per uso interno.

Secondo quanto riferito, alcuni funzionari chiave dell’amministrazione Trump sono preoccupati che l’esportazione di Blackwell in Cina possa rafforzare le capacità di intelligenza artificiale della Cina e indebolire la sicurezza degli Stati Uniti. In un’intervista alla CBS andata in onda il 2, alla domanda se avrebbe permesso a Nvidia di vendere semiconduttori all’avanguardia alla Cina, il presidente Trump ha risposto: “No, non lo faremo. Non lo faremo. Non permetteremo a nessun altro di avere semiconduttori all’avanguardia“

L'articolo La Cina supererà gli USA nella corsa all’AI! Il CEO di Nvidia avverte “siamo bloccati dal cinismo” proviene da Red Hot Cyber.

Tesla vuole dare a Musk 1000 miliardi di dollari per un piano da fantascienza. Più del PIL della Svizzera!

Numerosi apprezzamenti stanno arrivando per Elon Musk, che continua a guidare con destrezza le molteplici attività del suo impero, tra cui Tesla, SpaceX, xAI e Starlink, mantenendo salda la sua posizione al comando.

Gli estimatori celebrano una strategia che, secondo loro, supera i limiti dell’industria automobilistica. Dietro tale plauso si cela un successo epocale: un piano di remunerazione senza precedenti per Musk, qualora Tesla raggiunga i propri traguardi. La situazione è di grande portata, il contesto è enorme e le conseguenze sono ben più ampie delle semplici auto elettriche.

La traiettoria di Musk e Tesla si trasforma in una saga di profitti non appena si entra in contatto con questo nuovo piano. Il consiglio di amministrazione di Tesla ha proposto di assegnare a Musk 423,7 milioni di azioni vincolate, quasi il 12% del capitale attuale, se verranno raggiunti determinati obiettivi di performance.

Affinché Musk riceva l’intero importo, la capitalizzazione di mercato di Tesla deve aumentare a circa 8,5 trilioni di dollari (rispetto ai circa 1,4 trilioni di dollari attuali). Inoltre, gli obiettivi includono 20 milioni di veicoli venduti, 1 milione di robotaxi in servizio, 1 milione di robot umanoidi “Optimus” e un EBITDA cumulativo potenzialmente pari a 400 miliardi di dollari.

Sebbene le vendite di auto elettriche rimangano al centro dell’attenzione di Tesla, Musk invia un segnale chiaro: il futuro si gioca nell’intelligenza artificiale, nella robotica e nei veicoli autonomi.

In questo contesto, la compensazione XXL diventa più una scommessa su un futuro dirompente che un enorme bonus.

L’evento si è trasformato in una festa. All’assemblea generale, il 75% dei voti si è espresso a favore del piano. Questa esplosione di approvazione è stata accompagnata da una standing ovation in sala. Tuttavia, dietro questo sostegno, alcuni importanti investitori rimangono cauti. Alcuni fondi istituzionali – come il fondo sovrano norvegese – hanno votato contro.

Sebbene Musk e Tesla godano di un ampio sostegno da parte degli investitori al dettaglio, gli aspetti di governance sollevano interrogativi: il potere di voto di Musk potrebbe aumentare da circa il 13% a oltre il 25% se tutti gli obiettivi saranno raggiunti.

L’azienda ha fatto una scelta strategica: mantenere quella che considera la “risorsa principale”. Come ha commentato l’analista Dan Ives :

Con questo piano di compensazione ora approvato, che mantiene la risorsa più importante di Tesla – Musk – al vertice per i prossimi anni, continuiamo a credere che la valutazione legata all’intelligenza artificiale stia iniziando a sbloccarsi. A nostro avviso, il passaggio a una valutazione basata sull’intelligenza artificiale per Tesla è ora avviato per i prossimi 6-9 mesi, con l’implementazione di FSD, la penetrazione delle tecnologie autonome nella base installata di Tesla e l’accelerazione dei progetti Cybercab e Optimus negli Stati Uniti.

A questo bivio, Tesla oscilla tra l’adulazione dei suoi fan e la cautela degli osservatori.

Inoltre, la realtà interna di Tesla getta ombre: la sua quota di mercato negli Stati Uniti è scesa al livello più basso dal 2017.

Tesla punta molto su questo mix di veicoli elettrici, intelligenza artificiale, robotica, prendendo spunto dalla blockchain e dalle dinamiche del Web3 (etica, trasparenza, disruption): un interessante gioco di specchi per gli investitori curiosi.

L'articolo Tesla vuole dare a Musk 1000 miliardi di dollari per un piano da fantascienza. Più del PIL della Svizzera! proviene da Red Hot Cyber.

Gli hacker russi di Sandworm colpiscono l’industria ucraina del grano con malware wiper

Gli hacker russi Sandworm utilizzano malware wiper contro l’industria cerealicola ucraina.

L’industria cerealicola ucraina è diventata l’ultimo obiettivo della famigerata unità di hacker russi Sandworm, sostenuta dallo Stato, nel quadro dei continui sforzi di Mosca per minare l’economia bellica del Paese.

Secondo una nuova ricerca della società slovacca di sicurezza informatica ESET, tra giugno e settembre il gruppo legato al Cremlino ha utilizzato diversi tipi di malware per cancellare i dati delle organizzazioni ucraine nei settori cerealicolo, energetico, logistico e governativo. Sebbene gli attacchi wiper abbiano colpito frequentemente le infrastrutture ucraine dall’invasione russa, l’industria agricola, fonte fondamentale delle entrate da esportazione del Paese, è stata raramente presa di mira direttamente.

Sandworm, che le agenzie di intelligence occidentali collegano al servizio di intelligence militare russo (GRU), è responsabile di alcuni dei cyberattacchi più dannosi nella storia dell’Ucraina, tra cui il blackout della rete elettrica del 2015, l’epidemia di malware NotPetya del 2017 e l’hacking dello scorso anno del principale fornitore di telecomunicazioni Kyivstar.

ESET ha affermato che le recenti operazioni hanno incluso due wiper, Zerolot e Sting, distribuiti ad aprile contro un’università ucraina, seguiti da ulteriori ondate contro aziende del settore cerealicolo ed energetico. Il malware wiper è progettato per cancellare definitivamente i dati e interrompere le operazioni.

L’azienda ha anche collegato gli attacchi a un altro gruppo di hacker, noto come UAC-0099, che avrebbe effettuato le intrusioni iniziali prima di passare l’accesso a Sandworm. Secondo il team di risposta alle emergenze informatiche dell’Ucraina (CERT-UA), UAC-0099 è attivo almeno dal 2022 e ha preso di mira istituzioni governative e di difesa ucraine in campagne di spionaggio.

“Questi attacchi distruttivi da parte di Sandworm ci ricordano che i wiper rimangono uno strumento frequente degli attori di minaccia allineati con la Russia in Ucraina”, ha affermato ESET.

Sebbene alcuni rapporti suggerissero un passaggio ad attività di spionaggio da parte di tali gruppi alla fine del 2024, i ricercatori hanno affermato che Sandworm ha continuato a condurre regolarmente attacchi wiper contro entità ucraine dall’inizio del 2025.

Le autorità ucraine competenti in materia di sicurezza informatica hanno ripetutamente avvertito che gli attori russi, tra cui Sandworm, spesso coordinano tali operazioni con attacchi missilistici e con droni per amplificarne l’impatto.

Al di là dell’Ucraina, ESET ha osservato che gruppi di hacker russi, tra cui RomCom e Gamaredon, continuano a prendere di mira gli Stati membri dell’Unione Europea, concentrandosi spesso su entità legate alle reti di difesa o logistiche dell’Ucraina.

“Anche obiettivi non ucraini presentano spesso alcuni legami evidenti con l’Ucraina e il suo sforzo bellico complessivo”, hanno scritto i ricercatori, “suggerendo fortemente che il conflitto continui a mobilitare la maggior parte dell’attenzione e delle risorse dell’intelligence russa”.

L'articolo Gli hacker russi di Sandworm colpiscono l’industria ucraina del grano con malware wiper proviene da Red Hot Cyber.

Malware adattivi e altamente evoluti utilizzano l’AI per confondere le difese di sicurezza

Sulla base di una recente analisi del Google Threat Intelligence Group (GTIG), è stato identificato un cambiamento avvenuto nell’ultimo anno da parte degli attori delle minacce.

Miglioramenti continui nelle underground per l’abuso delle AI

Gli aggressori non sfruttano più l’intelligenza artificiale (IA) solo per aumentare la produttività negli attacchi, ma stanno implementando nuovi malware basati sull’IA nelle operazioni attive.

Questo segna una nuova fase operativa di abuso dell’IA, che coinvolge strumenti che alterano dinamicamente il comportamento durante l’esecuzione.

Il report del team di analisi sulle minacce di Google, rappresenta un aggiornamento dell’analisi di gennaio 2025, “Adversarial Misuse of Generative AI” e descrive in dettaglio come gli attori delle minacce informatiche e i criminali informatici supportati dai governi stiano integrando e sperimentando l’intelligenza artificiale, durante l’intero ciclo di vita dell’attacco.

PROMPTFLUX: il nuovo malware che utilizza l’API Gemini

L’analisi di GTIG rivela che attori sponsorizzati da stati provenienti da Corea del Nord, Iran e Cina, insieme a criminali motivati ​​finanziariamente, stanno abusando sempre più di Gemini durante l’intero ciclo di vita dell’attacco, dalle esche per il
phishing fino alle configurazioni di comando e controllo.

Un nuovo malware rilevato, induce l’LLM a generare script di evasione autonomi, producendo solo il codice senza testo estraneo e registrando le risposte in un file temporaneo per consentirne il perfezionamento.

Google TIG fa notare che, sebbene esistano funzionalità di auto aggiornamento commentate, questo indica uno sviluppo attivo iniziale. Il malware tenta anche di diffondersi lateralmente su unità rimovibili e condivisioni di rete.

Questo approccio sfrutta il potere generativo dell’intelligenza artificiale non solo per la creazione, ma anche per la sopravvivenza continua del malware stesso, a differenza del malware statico che si basa su firme fisse facilmente rilevabili dai difensori.

L’emergere di PROMPTFLUX è in linea con la maturazione del mercato della criminalità informatica, in cui gli strumenti di intelligenza artificiale inondano i forum underground, offrendo funzionalità che vanno dalla generazione di deepfake allo sfruttamento delle vulnerabilità a prezzi di abbonamento.

Malware sempre più dinamici per confondere le difese attive

Una volta scoperto il malware, Google ha rapidamente disabilitato le chiavi API e i progetti associati, mentre DeepMind migliora i classificatori e le misure di sicurezza del modello di Gemini per bloccare le richieste di utilizzo improprio.

Si può quindi dedurre, che gli attori delle minacce stanno costantemente riducendo le barriere per gli autori alle prime armi.

GTIG mette in guardia dai rischi maggiori, tra cui ransomware adattivi come PROMPTLOCK che creano dinamicamente script Lua per la crittografia.

L’azienda sottolinea il suo impegno verso un’intelligenza artificiale responsabile attraverso principi che danno priorità a robuste barriere di sicurezza, condividendo informazioni tramite framework come Secure AI (SAIF) e strumenti per la gestione delle vulnerabilità tramite red-teaming.

Innovazioni come Big Sleep per la ricerca delle vulnerabilità e CodeMender per l’applicazione automatica delle patch sottolineano gli sforzi per contrastare in modo proattivo le minacce dell’intelligenza artificiale.

L'articolo Malware adattivi e altamente evoluti utilizzano l’AI per confondere le difese di sicurezza proviene da Red Hot Cyber.