The paper [1] is interesting, although they spend too much space writing about the purported benefits. The concept here is to exploit WiFi gear's usages of many different frequencies. Measuring the amplitude and phase shift on each channel provides some info about what each channel bounced off of and diffracted around.

Inevitably, they then dump all the data into a neural net. They trained on about 15 objects, which is a reasonable start. They didn't go on to testing with enough objects to be useful. "The results show that our system can detect over 95% dangerous objects in different types of bags and successfully identify 90% dangerous material types." By "dangerous" they just mean metal or liquid. It's not like they can tell a gun from a soup can.

[1] http://www.winlab.rutgers.edu/~yychen/papers/Towards%20In%20...

> The concept here is to exploit WiFi gear's usages of many different frequencies. Measuring the amplitude and phase shift on each channel provides some info about what each channel bounced off of and diffracted around.

This exists. They're just reimplemented non-cooperative SFCW radar. It's not new.

It seems like they can detect liquid and metal, which is very different from detecting bombs and guns. That's like building a face detection library and saying you can find terrorists. If mostly you just identify people with water bottles or laptops, what problem would this really solve?

Proof of concept. If they get funded, expect this to disappear from public view for a couple years while they build up their object database and train their network. Then it will be quietly sold to the police and military. Then there will be a "whistleblower" in 2028, but you heard it here first.

Pattern recognition of common bomb and gun configurations? Or better, any non-standard configuration.

That might be able to detect something like a hand grenade (but honestly I'm dubious), but finding "non-standard configurations" seems like it requires a full up strong AI. It has to be able to look at a jumble of components in a configuration it has never seen before and recognize that it is a bomb. This is hard problem for humans and hits on several hard aspects of AI at once.

Or just recognize it as not-a-bomb. Easier - e.g. the set of metal things normally in luggage may be small

Bombs aren't necessarily made out of metal.

The fusing wires almost certainly are, unless you're dealing with a nation-state or, possibly, a major corporation.

Unless you just use a fuse, Looney Toons/Shoe Bomber style.

No, but guns are.

guns can be made with 3d printed plastic. https://defdist.org/

This isn't new, or unique to 3d printing. Plastic/polymer firearms have been around for a long time, including a few out there with ceramic barrels and chambers.

Ammunition is easier to detect, generally - almost all of it is metallic-cased and uses a metallic projectile. There have been attempts to create practical baseless ammunition, but they've generally been failures and I'm not aware of any available for purchase commercially.

Ultimately, this sort of thing is interesting from a technical perspective but not really very useful. This is even more the case in the US, as firearms are ubiquitous. In 2016, there were 14.5 million active concealed carry permits in the US, and in 11 states, no license was required to carry concealed.

Given that something like one in twenty or so adults in most parts of the country are probably carrying a firearm at any given time, knowing "there's a gun in the room" isn't very useful.

> Given that something like one in twenty or so adults in most parts of the country are probably carrying a firearm at any given time

I think that’s a gross overestimation. I would guess that most CCW permit holders hardly ever carry in practice (source: used to have one).

There are states where you don’t need a permit to own a pistol at all. In others, you may not technically need one to take your gun to the shooting range (because after all, if you own one, you should practice) but...

If you’re stopped by police in a context where you don’t need a permit, it’s still always better to have one. If the police aren’t aware of the law or the particular permit exception you fall under, the best likely outcome is you get put in a cage while they figure it out. Having a permit helps avoid that.

If the NRA was clever they would make a much bigger deal about how many guns are already out in circulation in America, and how there's literally nothing that can ever be done to remove them all. The vanguard of gun violence prevention's focus on keeping guns out of the hands of America's growing population of violent mentally ill individuals isn't realistic. It's going to be much easier to identify and remove the violent mentally ill people from society than to remove all of the ways in which they can inflict harm.

Yeah I don't think it's any cause for alarm, and I agree that it's mostly interesting from a technical perspective.

> guns can be made with 3d printed plastic.

If you're referring to the Liberator, it uses a metal nail as a firing pin.

But now you're talking about detecting a nail. Which if you decided to ban (somehow?), you'd still be tasked with detecting a tiny object. Most metal detectors are tuned not to even care about normal size belt buckles as far as I know.

I hadn't thought about that but it makes sense. I wonder if there's any type of plastic filament that could do the job.

> I wonder if there's any type of plastic filament that could do the job.

On a practical note, probably not: the designers probably tried really hard to make a fully plastic gun - they took the trouble to make the barrel out of plastic - but didn't. Stands to reason that it's due to necessity.

At a more theoretical level, I think you'd find it exceedingly difficult to trigger a primer using a material softer than the primer itself.

I'm sure it's not impossible - for example, you could probably trigger a primer by shooting it with a plastic bullet - but within the engineering confines of a handgun it seems implausible, particularly since the easy solution, using a nail, is so readily available.

What about ceramic, glass, or heat activated?

some landmines use glass ceramic firing pins, according to wikipedia.

Ya, but they're not very good.

Considering the world impact a couple of box cutters had, I think we can agree that a gun doesn’t need to be all that good.

They would have zero impact now. Cockpit doors are reinforced in planes.

>Or better, any non-standard configuration.

Sure, if your definition of better is generating a ton of false positives as a pretext for patting people down.

Can't wait to see the day police show up at peoples houses because an opaque AI/ML/Statistical-thingy said they were creating bombs, when in fact it was a loaf of bread (or whatever). Minority report gets more and more real.

99% accuracy sounds impressive until you realize this thing will flag 1% of all objects as potentially threatening.

And of course it'll entirely miss the 2,600 rounds of ammo in my apartment.

And the box of tannerite, and the separate containers of powdered iron and aluminum...

Paper (PDF): http://www.winlab.rutgers.edu/~yychen/papers/Towards%20In%20...

My first reading of that headline was that "bombs" means unexploded ordnance (which is a regular problem for construction projects in my home town thanks to WW2 air raids). I'm not sure if this means I'm sheltered or level-headed.

Another take on this paper would read, “Engineeers have created a way to weaponize Wi-Fi signals for invisible and invasive surveillance”.

This turns consumer level tech into a militant authoritarian’s dream. Have they also developed ways to identify when this technology is being used surreptitiously?

This method appears to just measure the Tx and Rx paths on an off-the-shelf modem, so you'd have to physically check the router, or I'm sure this could be handled in firmware. They're measuring phase and amplitude differences from unmodified traffic on 2.4GHz so you can't detect it as an outsider.

My cynical post script to the title, without reading the article: "in a laboratory setting".

Agreed. When this method is applied in, say, Germany, to safely detect WWII ordinance, then it will have proven itself.

Reminds me of Wi-Vi: See Through Walls with Wi-Fi Signals from 2013:

http://people.csail.mit.edu/fadel/wivi/

This technology can prove useful for public safety since there are WiFi almost everywhere these days and all entry points in malls or schools can be covered very easily. There is a chance this can be misused as well.

My mom worked at a bank that installed what they called a "man trap". Essentially there's a large magnetometer between two sets of heavy doors that locks the doors if it detects someone with a firearm (or any large block of magnetic material) has tried to enter the bank.

On any given week, they'd catch about a half dozen police officers, a maintenance person or two who still had their tools on them, and the occasional delivery person.

After a few weeks they disabled the system entirely.

An interesting article about false positives and negatives and their impact in the reliability of a detection system. (I believe it has been linked in HN before) It explains why "95% accurate" is BS.

https://blog.danslimmon.com/2012/11/02/car-alarms-and-smoke-...

I thought every bank branch had that (or a revolving door, also with a magnetometer). I don't recall ever seeing a bank branch without one. And it's not just large blocks of magnetic material, even small things like key chains can set them off.

"Researchers rediscover non-cooperative radar, news at 11"

It's really, really, really hard to do this with cooperative RF sources and way way more bandwidth. This is horseshit.

So we see forcing the use of clear backpacks as an invasion of privacy but don't see tech that scans a backpack as invasive?

I'm curious if I can do this at home, and if so if there are any plans online besides the paper. I'd love to play around with it.

This headline immediately make me think of Arthur C. Clarke's "The Trigger".

"The Trigger" (2000): https://archive.org/details/trigger00clar