Post a news story, entertain me!

Financial Times is an instant paywall, so I couldn't read the article, but if this is what I think it is, it's not done by AI trying to recognize the images. They will hash the photos with an algorithm, probably SHA256, and use those hashes to detect images that are known to be child abuse imagery.

This is the same way most services already detect these images being sent via emails or stored on cloud services without needing to distribute the actual images. The new part here is that Apple will be using it on photos on your devices that you haven't uploaded anywhere.

The chances of a hash being identical to another one from a totally different image are so vanishingly small that it's not worth talking about. So, false positives are... not likely.

Whether or not the boundary crossed by checking images on your devices is okay is another matter.

NSMike wrote:

Financial Times is an instant paywall, so I couldn't read the article, but if this is what I think it is, it's not done by AI trying to recognize the images. They will hash the photos with an algorithm, probably SHA256, and use those hashes to detect images that are known to be child abuse imagery.

This is the same way most services already detect these images being sent via emails or stored on cloud services without needing to distribute the actual images. The new part here is that Apple will be using it on photos on your devices that you haven't uploaded anywhere.

The chances of a hash being identical to another one from a totally different image are so vanishingly small that it's not worth talking about. So, false positives are... not likely.

Whether or not the boundary crossed by checking images on your devices is okay is another matter.

Yes this is the method they are using, hash matching against a database of hashes of known child abuse imagery.

NSMike wrote:

The chances of a hash being identical to another one from a totally different image are so vanishingly small that it's not worth talking about. So, false positives are... not likely.

Whether or not the boundary crossed by checking images on your devices is okay is another matter.

Asking out of ignorance as to the math of hashing - How vanishingly small? Because we're talking about a very large sample set.

Google says ~ three-quarter of a billion iPhones currently in use worldwide. On average each has, what, a thousand photos sitting on it?

So if your false positive isn't at least a few orders of magnitude less than one-in-750 billion, you're gonna be ruining lives. And that's a helluva accuracy rate.

Jonman wrote:
NSMike wrote:

The chances of a hash being identical to another one from a totally different image are so vanishingly small that it's not worth talking about. So, false positives are... not likely.

Whether or not the boundary crossed by checking images on your devices is okay is another matter.

Asking out of ignorance as to the math of hashing - How vanishingly small? Because we're talking about a very large sample set.

Google says ~ three-quarter of a billion iPhones currently in use worldwide. On average each has, what, a thousand photos sitting on it?

So if your false positive isn't at least a few orders of magnitude less than one-in-750 billion, you're gonna be ruining lives. And that's a helluva accuracy rate.

https://stackoverflow.com/a/4014407

The usual answer goes thus: what is the probability that a rogue asteroid crashes on Earth within the next second, obliterating civilization-as-we-know-it, and killing off a few billion people? It can be argued that any unlucky event with a probability lower than that is not actually very important.

If we have a "perfect" hash function with output size n, and we have p messages to hash (individual message length is not important), then probability of collision is about p2/2n+1 (this is an approximation which is valid for "small" p, i.e. substantially smaller than 2n/2). For instance, with SHA-256 (n=256) and one billion messages (p=109) then the probability is about 4.3*10-60.

A mass-murderer space rock happens about once every 30 million years on average. This leads to a probability of such an event occurring in the next second to about 10-15. That's 45 orders of magnitude more probable than the SHA-256 collision. Briefly stated, if you find SHA-256 collisions scary then your priorities are wrong.

But, tt's not the standard hash algorithm. It allows for some amount of variation so criminals can't easily evade the algorithm by making small changes. It has some amount of fuzzy matching, like face and fingerprint matching.

The short answer is we don't know how likely false positives and negatives will be because Apple hasn't released the code for independent investigation. However, researchers have been able to reverse and fool it already.

https://www.schneier.com/blog/archiv...

https://www.schneier.com/blog/archiv...

There was some good discussion about this on the Lawfare podcast, and Risky Business.

Noted - when the ivermectin runs out, shove your bike pump hose up there.

Jonman wrote:

Noted - when the ivermectin runs out, shove your bike pump hose up there.

I've been looking for the perfect reason to not have to wear a mask.

Does this mean that my pants-free jaunts to the grocery store could be putting me at risk for getting COVID up the ol yin yang?

After 100 Years, a Royal Wedding in Russia Evokes Days of the Czars

Among the list of royal attendees were Princess Leia of Belgium, Queen Sofia of Spain, Prince Rudolph and Princess Tilsim of Liechtenstein, and the last czar of Bulgaria, Simeon II.

IMAGE(https://th.bing.com/th/id/R.56240f4ba7bf4655ee8bb60d629e0e7e?rik=zet6%2bbT%2fRhcILg&pid=ImgRaw&r=0&PC=EMMX01)

Weight Loss May Not Actually Make You Healthier, Study Reveals

The researchers analyzed hundreds of studies looking at how weight loss, exercise and longevity fit together — focusing specifically on research that examined health outcomes in people who were considered overweight or obese. (The Centers for Disease Control and Prevention considers a BMI of 25 to 30 as overweight; anything over 30 is obese or severely obese.)

Ultimately, the evidence shows being active trumps weight loss when it comes to improving heart health and reducing overall mortality risk, the researchers concluded.

In fact, people who are considered obese may have a lower risk of premature death than those who are normal weight but not in good shape, according to the study.

IMAGE(https://c.tenor.com/Q9UYdX1PQ64AAAAC/thinner-creepy.gif)

Duh?!

A bit of satire

White House Warns Supply Chain Shortages Could Lead Americans To Discover True Meaning Of Christmas

As reports of worldwide shipping issues continued to mount, the White House warned Wednesday that supply chain shortages could lead Americans to discover the true meaning of Christmas. “Unless these unprecedented scarcities are reversed soon, hundreds of thousands of Americans could be forced to learn that there is more to life than material objects,” said White House press secretary Jen Psaki, cautioning that delays in shipping of clothing, toys, and other common gifts had the potential to make this Christmas the most communal and brotherly of any on record.