Sufficiently Advanced Technology And Justice

Imagine that you’re serving on a jury, and you’re given an image taken from a surveillance camera. It looks pretty much like the suspect, but the image has been “enhanced” by an AI from the original. Do you convict? How does this weigh out on the scales of reasonable doubt? Should you demand to see the original?

AI-enhanced, upscaled, or otherwise modified images are tremendously realistic. But what they’re showing you isn’t reality. When we wrote about this last week, [Denis Shiryaev], one of the authors of one of the methods we highlighted, weighed in the comments to point out that these modifications aren’t “restorations” of the original. While they might add incredibly fine detail, for instance, they don’t recreate or restore reality. The neural net creates its own reality, out of millions and millions of faces that it’s learned.

And for the purposes of identification, that’s exactly the problem: the facial features of millions of other people have been used to increase the resolution. Can you identify the person in the pixelized image? Can you identify that same person in the resulting up-sampling? If the question put before the jury was “is the defendant a former president of the USA?” you’d answer the question differently depending on which image you were presented. And you’d have a misleading level of confidence in your ability to judge the AI-retouched photo. Clearly, informed skepticism on the part of the jury is required.

Unfortunately, we’ve all seen countless examples of “zoom, enhance” in movies and TV shows being successfully used to nab the perps and nail their convictions. We haven’t seen nearly as much detailed analysis of how adversarial neural networks create faces out of a scant handful of pixels. This, combined with the almost magical resolution of the end product, would certainly sway a jury of normal folks. On the other hand, the popularity of intentionally misleading “deep fakes” might help educate the public to the dangers of believing what they see when AI is involved.

This is just one example, but keeping the public interested in and educated on the deep workings and limitations of the technology that’s running our world is more important than ever before, but some of the material is truly hard. How do we separate the science from the magic?

Remoticon Video: How To Use Machine Learning With Microcontrollers

Going from a microcontroller blinking an LED, to one that blinks the LED using voice commands based on a data set that you trained on a neural net work is a “now draw the rest of the owl” problem. Lucky for us, Shawn Hymel walks us through the entire process during his Tiny ML workshop from the 2020 Hackaday Remoticon. The video has just now been published and can be viewed below.

This is truly an end-to-end Hello World for getting machine learning up and running on a microcontroller. Shawn covers the process of collecting and preparing the audio samples, training the data set, and getting it all onto the microcontroller. At the end of two hours, he’s able to show the STM32 recognizing and responding to two different spoken words. Along the way he pauses to discuss the context of what’s happening in every step, which will help you go back and expand in those areas later to suit your own project needs.

Continue reading “Remoticon Video: How To Use Machine Learning With Microcontrollers”

Hackaday Podcast 096: Diaphragm Engine, DIY Dish Washer, Forgotten Soviet Computers, And A Starlink Teardown

Hackaday editors Elliot Williams and Mike Szczys discuss the latest and greatest in geeky goodness. This week we saw a Soviet time capsule come to light with the discovery of a computer lab from a building abandoned in the 1990’s. A two-cycle compressed air engine shatters our expectations of what is involved in RC aircraft design. There’s a new toolkit for wireless hacking on the scene in the form of a revitalized HackRF PortaPack firmware fork. And what goes into dishwasher design? Find out in this exciting episode.

Take a look at the links below if you want to follow along, and as always, tell us what you think about this episode in the comments!

Direct download (55 MB)

Places to follow Hackaday podcasts:

Continue reading “Hackaday Podcast 096: Diaphragm Engine, DIY Dish Washer, Forgotten Soviet Computers, And A Starlink Teardown”

This Week In Security: IOS Wifi Incantations, Ghosts, And Bad Regex

I hope everyone had a wonderful Thanksgiving last week. My household celebrated by welcoming a 4th member to the family. My daughter was born on Wednesday morning, November 25th. And thus explains what I did last week instead of writing the normal Hackaday column. Never fear, we shall catch up today, and cover the news that’s fit to be noticed.

iOS Zero-click Wifi Attack

[Ian Beer] of Google’s Project Zero brings us the fruit of his lockdown-induced labors, a spectacular iOS attack. The target of this attack is the kernel code that handles AWDL, an Apple WiFi protocol for adhoc mesh networks between devices. The most notable feature that makes use of AWDL is AirDrop, Apple’s device-to-device file sharing system. Because AWDL is a proprietary protocol, the WiFi hardware can’t do any accelerated processing of packets. A few years back, there was an attack against Broadcom firmware that required a second vulnerability to jump from the WiFi chip to the device CPU. Here, because the protocol is all implemented in Apple’s code, no such pivot is necessary.

And as you’ve likely deduced, there was a vulnerability found. AWDL uses Type-Length-Value (TLV) messages for sending management data. For a security researcher, TLVs are particularly interesting because each data type represents a different code path to attack. One of those data types is a list of MAC addresses, with a maximum of 10. The code that handles it allocates a 60 byte buffer, based on that maximum. The problem is that there isn’t a code path to drop incoming TLVs of that type when they exceed 60 bytes. The remainder is written right past the end of the allocated buffer.

There is more fun to be had, getting to a full exploit, but the details are a bit too much to fully dive in to here. It interesting to note that [Ian] ran into a particular problem: His poking at the target code was triggering unexpected kernel panics. He discovered two separate vulnerabilities, both distinct from the vuln he was trying to exploit.

Finally, this exploit requires the target device to have AWDL enabled, and many won’t. But you can use Bluetooth Low Energy advertisements to trick the target device into believing an Airdrop is coming in from a trusted contact. Once the device enables AWDL to verify the request, the attack can proceed. [Ian] reported his findings to Apple way back in 2019, and this vulnerability was patched in March of 2020.

Via Ars Technica.
Continue reading “This Week In Security: IOS Wifi Incantations, Ghosts, And Bad Regex”

Remoticon Video: Circuit Sculpture Workshop

Circuit Sculpture was one of our most anticipated workshops of Hackaday Remoticon 2020, and now it’s ready for those who missed it to enjoy. A beginning circuit sculptor could hardly ask for more than this workshop, which highlights three different approaches to building firefly circuit sculptures and is led by some of the most prominent people to ever bend brass and components to their will — Jiří Praus, Mohit Bhoite, & Kelly Heaton.

For starters, you’ll learn the different tools and techniques that each of them uses to create their sculptures. For instance, Kelly likes to use water-based clay to hold components in specific orientations while forming the sculpture and soldering it all together. Jiří and Mohit on the other hand tend to use tape. The point is that there is no right or wrong way, but to instead have all of these tips and tricks under your belt as you sculpt. And that’s what this workshop is really about.

Continue reading “Remoticon Video: Circuit Sculpture Workshop”

Crossed Wires Crash Rockets

On November 17th, a Vega rocket lifted off from French Guiana with its payload of two Earth observation satellites. The booster, coincidentally the 17th Vega to fly, performed perfectly: the solid-propellant rocket engines that make up its first three stages burned in succession. But soon after the fourth stage of the Vega ignited its liquid-fueled RD-843 engine, it became clear that something was very wrong. While telemetry showed the engine was operating as expected, the vehicle’s trajectory and acceleration started to deviate from the expected values.

There was no dramatic moment that would have indicated to the casual observer that the booster had failed. But by the time the mission clock had hit twelve minutes, there was no denying that the vehicle wasn’t going to make its intended orbit. While the live stream hosts continued extolling the virtues of the Vega rocket and the scientific payloads it carried, the screens behind them showed that the mission was doomed.

Displays behind the hosts clearly showed Vega wasn’t following the planned trajectory.

Unfortunately, there’s little room for error when it comes to spaceflight. Despite reaching a peak altitude of roughly 250 kilometers (155 miles), the Vega’s Attitude Vernier Upper Module (AVUM) failed to maintain the velocity and heading necessary to achieve orbit. Eventually the AVUM and the two satellites it carried came crashing back down to Earth, reportedly impacting an uninhabited area not far from where the third stage was expected to fall.

Although we’ve gotten a lot better at it, getting to space remains exceptionally difficult. It’s an inescapable reality that rockets will occasionally fail and their payloads will be lost. Yet the fact that Vega has had two failures in as many years is somewhat troubling, especially since the booster has only flown 17 missions so far. A success rate of 88% isn’t terrible, but it’s certainly on the lower end of the spectrum. For comparison, boosters such as the Soyuz, Falcon 9, and Atlas have success rates of 95% or higher.

Further failures could erode customer trust in the relatively new rocket, which has only been flying since 2012 and is facing stiff competition from commercial launch providers. If Vega is to become the European workhorse that operator Arianespace hopes, figuring out what went wrong on this launch and making sure it never happens again is of the utmost importance.

Continue reading “Crossed Wires Crash Rockets”

Remoticon Video: How To Reverse Engineer A PCB

You hold in your hand a circuit board from a product you didn’t make. How does the thing work? What a daunting question, but it’s both solvable and approachable if you know what you’re doing. The good news is that Eric Schlaepfer knows exactly what he’s doing and boiled down the process of reverse engineering printed circuit boards into this excellent workshop. It was presented live during the 2020 Hackaday Remoticon, and the edited video, which you’ll find below, was just published. Slides for the talk have been published on the workshop project page.

Need proof that he has skills that we all want? Last year Eric successfully reverse-engineered the legendary Sound Blaster audio card and produced his own fully-functional drop-in replacement called the Snark Barker. And then re-engineered it to work with the ancient MCA bus architecture. Whoa.

Continue reading “Remoticon Video: How To Reverse Engineer A PCB”