Sony's a7 IV gets anti-forgery crypto signature technology (for commercial purposes)

@Antisthenes
Nice reply, however Sony was never a serious player in most of the technologies you suggested. Developers are not Cryptographers. Look at WEP, they can take any well developed crypto and misuse it. As for Mobile, just look at https://owasp.org/www-project-mobile-top-10/ to see that your phone has bus sized security holes, including crypto.
Sorry, I'm a sceptic.

Well, you'd have to show that there are "bus-sized security holes" in the cryptographic algorithms and / or hardware implementations of modern security chips, including but not limited to:

- EMC credit cards (VISA/MC/AMEX etc.)
- public transit fare cards, be they based on NFC or Felica
- mobile phone SIM cards
- the secure element embedded in smartphone NFC chips, like those produced by NXP
- secure processor enclaves found e.g. in iPhones or in ARM TrustZone chips

If there were "bus-sized security holes" that effectively compromise the security of any of the chips in the above applications, I'm pretty sure we'd have heard about it by now, especially as there must be, by now, /hundreds of millions/ of such chips in use all over the world, including in e-wallet / mobile payment applications, and they are obviously an interesting target for security researchers.

So, there is no need for "an expensive ASIC and possibly a FPGA", as the chips mentioned above are all, in essence, TPMs.

RSA's SecurID tokens, as they are based on a shared secret that is conceptually akin to a symmetric key, do not actually use the RSA asymmetric (private/public) key system.

So, basically, that article describes a state-level, sophisticated attack mounted by people who determined that there was no practical way e.g. to extract the secret key of a SecurID token to clone it, and thus choose an "easier" path — i.e. mounting a sophisticated attack against RSA Data Security's corportate network to steal the master database containing the secret SecurID seed numbers.

The compromise of a shared secrets database (the SecurID token seeds) used in a product sold by RSA Data Security Inc. doesn't mean that the completely different RSA public key algorithm, which doesn't rely on shared secrets, has been compromised or weakened in any way.

../..

Injecting bogus picture data to be signed by another camera might indeed be as simple as photoshopping the picture to remove / add some important items, then printing that picture and taking a picture of the print, wth a plausible shutter speed and ISO sensitivity (assuming the digital signature also signs some parts of the EXIF data like shutter speed and ISO)

Anyway, there's an implicit trust that if a potentially doctored picture is important enough that creating a fake signature is worth it — e.g. in a lawsuit, — then, there's also some chance that the original, undoctored picture might be found somewhere.

That's similar to the hope that people become aware, when the stakes are high enough, that there is more than one web site on the internet claiming e.g. to be www.citibank.com

I have AirTags hidden in a small grip extension. Assuming the thief doesn’t detach and separate the extension I have at least some hope of tracking them down if they get stolen.

Sony is on the steering committee of C2PA. The C2PA specifications allow for extending the security of the signed image into a post-processing workflow. The first step in that process is creating and signing the original image, which the camera handles.

Suggest you do a bit more research to see how this may actually work before writing it off as useless.

My concern is that this post capture/editing workflow will be supported by multiple vendors and also that the code will be cracked allowing anyone to make it appear that an image is authentic.

Use case: Police or forensic people take an image of something and want to present it in court. This gives an extra validation that the image has not been manipulated.

Yes it's a small use group, but in that cases it is extremely valuable.

I deal with art reproduction. While I wouldn't necessarily want this most of the time (as I shoot RAW and then apply color profiles and such which will negate and signing) I could see insurance or auction houses wanted trusted images as proof of no damage before shipping out.

Since the press release seems to focus on business solutions. I wouldn't be surprised if some company went to Sony and said "we have this use case and will either pay for you to develop the tools we need" or "if you add this feature we'll buy a non negligible number of cameras."

The code will eventually be cracked but it may be 10 years and by the the camera will be obsolete and the pictures taken will have run their course.

@macantin

Point a) There is nothing in the press release to suggest it's using a propriety cryptographic method. It would be idiotic to use a less secure method than is the current state of the art and/or to reinvent the wheel. No one would buy into it if they did. They won't be using something based on obsolete tech but will probably AES instead (we don't know exactly what they are doing as there is not enough info).

b) it's using the cameras processor so presumably it could be updated but even if it can't be, then so what? Sony will bring out an A7 V (or whatever) and for those who see this functionality as crucial and need the update will buy one. It's not as if AES is likely to compromised any time soon (if ever) so it's not a valid assumption to make that frequent updates will be required anyway. The Sony A7 IV like all digital cameras has a model life of a few years so will most likely be replaced long before an update is needed.

@mick232 I'm familiar with symmetric key encryption (still from the old PGP days), but that doesn't necessarily solve the problem either. The question still remains how the secret key is being put into the camera. Will it be generated on Sony's server and installed with the firmware? Bad idea, because then you can extract it from the firmware file. Will it be generated in-camera with an algorithm? If the camera's memory isn't encrypted, then the key can still be extracted and used to manipulate images.

Discussions of trusted computing of the past decades show that fingerprinting is not trivial to implement if it should be 100% robust and trustworthy. In the case of camera image fingerprinting, the solution is either good or useless. There won't be a "just good enough".

So far, we haven't had a single piece of prosumer hardware where such technologies had robust implementations. If only one hacker succeeds to crack the Sony A7iv's fingerprinting, the whole feature will be worthless.

Sony pretty much only uses open standards nowadays, but other camera companies do the same because it's the only way forward. Also seems to be quite "open" regarding 3rd party: look at the Playstation 5, you can add any 3rd party SSD in its extension bay as long as it's fast enough, whereas Microsoft only allows their own.

I used Sony decades before. I had memory sticks, proprietary bus systems on my mc2700, useless cameracables on my trv 30e and in each and every aspects they tried to use their own Protocoll or hardware. Don't you think that I have reason to react like that? Leaving alone the other far more horrific behaviors on computer systems and audio codecs/formats.

They acted like Microsoft until they faced the consequences... Now they sometimes comply to standards but only if it fits to their profit strategy. If you Sony fanboys need more informations use Google.

Or the Sonysearchengine?

Memory Stick was on the market two full years before SD (SD was created by Matsushita to irritate Sony after the Memory Stick launch), and was licenced and used by multiple companies before SD was even on the market.
What you call useless camera cable on your DCR-TRV30e is actually FireWire (first created by Apple, standardized as IEEE 1394).

About codecs, they use MPEG4, H.264, H265 standards. They created ATRAC years before MP3 was even an idea.

On the Alpha side, they use standard memory cards (SD and CFExpress), standard connectors (HDMI, micro-USB, USB-C, Ethernet), even the standard hot-shoe (don't forget that they previously used iISO for backwards-compatibility with Minolta). Even their communication protocol has been published for everyone to use!

On the PlayStation-side, PS2 used standard PATA hard-drives, PS3 used SATA hard-drives and mini-USB, PS4 used SATA and micro-USB, PS5 uses M.2 NVMe SSD and USB-C.

Please educate yourself about electronics history.

If you think Sony is using lots of proprietary stuff you obviously have not seen Nintendo.

@illogict
Hey Mr wannabeintelligent, you have a big mouth. You know exactly what I mean. I did not mention betamax or other useless cr4p because I didn't use it by myself. The same goes to mini disc or other "standards". I mentioned just the hardware *I* used and after one of my Mc 2700 parts broke it was not replaceable with other components but with Sony ones. That you did not refer to that is clear, because you have no glue of the bussystem of it. I know that my Sony has firewire but it came with a cable like a dragon with 9 heads on it. Very useful. Useful as the DVD format war back then. Keep on tr0lling f4nboy.

You know that Beta was also created before VHS, and that VHS was designed by Matsushita after Sony shared details about it?
Floppy disc, half of CD, part of DVD, and the whole Blu-ray was also created by them: aren't they standard enough for you? Same for S/PDIF and a whole bunch of other stuff.
There was no DVD format war, as actually the Matsushita-created SuperDisc (hint: ever wondered why the SD-card logo looked like a disc?) and the Sony/Philips MultiMedia-CD formats were merged to avoid a format war.

I actually don't care about Sony: I use their products, as I use products of competitors if they better fit my needs. But credit is due where credit is deserved. I have been an engineer in consumer electronics (for Sony competitors) for quite some time, and worked on some IEEE and European standards projects, so am quite knowleadgeable about standards.

Since the camera signs a digest of the image and metadata, the private key has to at some point be inside the camera. Most likely that means storing the key inside the camera's non-volatile internal memory, preferably encrypted with a passcode the user presents at startup via camera menus. An alternate and perhaps preferable mechanism would be to store the encrypted key on a smartphone, using the phone's existing authentication mechanisms to decrypt it for sharing with the camera over a secure channel over Bluetooth.

Third option is to use a hardware key stored inside the camera's chipset, ideally a key that is unique per individual camera.

Right, but the goal here is not to protect the camera but to sign images taken with it. Probably it would be better to have a separate system to prevent unauthorized use of the camera.

You are right, the article even mentions the creator explicitly. In that case, one would probably use secondary keys per user, protected by the user with a passphrase or e.g. his phone. The secondary keys would be signed by the built-in hardware key.

@Arcimboldo
It's not

@Panther Fan Exactly, so owning an equivalent RAW doesn‘t mean content hasn‘t been tampered with.

Wow - why on earth are so many people commenting on this article who haven't understood what this is about at all. @Name That Camera: "Anti-Forgery" = "you cannot alter the image without it being detected". This is NOT about protecting the image against being stolen.

Anything can be altered in a digital world.
Especially when its screen related....

It is you who don't seem to understand how easy it is to bypass such solution. Or have you never made a screencopy of something...

You could even photograph the 'original' and bypass it.
The hashtag will never be the same due to different settings or a little different cropping.

There is no difference in todays solutions.
No crypto technology needed.
As long as you have the RAW you can always proof the original is yours.

This is not about protecting against theft. It's about proving the authenticity of an image I think. For journalists for example. So, the original photographer can proof that the image hasn't been altered.

This technology seems to be about forgery protection - ie : prevent alterations from being passed off as the original creator's work. It does not seem to be about copy/theft prevention or just preventing alterations (at least I don't see how it can help with that). I think that is what Arcimboldo was trying to say.

Exactly, Alfred, this is not at all about copy protection or anything like it. Hashing content with a crypto-hash is a standard technique to protect that content from being altered unnoticed - you recompute the hash with the content and if the hash has changed the content has, too. This technology will proof that an image was made with a certain camera and has not been modified afterwards. For it to work, the camera must also be tamper-proof: It must not be possible for anyone to modify the camera in such a way that altered or faked content can be injected into the system and the hash computed as if the content had been created with the camera.

I'm really wondering about the tamper-proof part: The only way known to me to achieve this is on-chip, something like Apple's so-called 'Secure Enclave'. If the A7IV is the first camera to get this this would suggest that only future cameras will also get it - that or the BIONZ XR has it built in already and every camera featuring it can be upgraded by firmware.

@Name That Camera "Anything can be altered in a digital world. " true but it's also true to say if you use cryptographic technology you can always detect IF it has been altered and that is what this tech will allow.

"You could even photograph the 'original' and bypass it." Won't work. That would be forgery. You know the original images "hash-value". A new one will have a different hash-value so you can say it's forged.

"The hashtag will never be the same due to different settings or a little different cropping." That's the whole point! If it's altered the hash-value is different so you know its not the absolute original.

"There is no difference in todays solutions.
No crypto technology needed.
As long as you have the RAW you can always proof the original is yours."

Wrong on all counts. Someone could steal your RAW file and edit it. You both have the "original" RAW file. With this technology the stolen, edited RAW would have a different hash value.

@Name That Camera,

Your argument is as off-topic as a counterfeiter using a crayon to draw an extra zero on a $10 and expecting its value to go up.

For the last time: this is not about image theft. It's about verification. You can screenshot a photo, but the result will not have this authentication. That's the whole point of this technology. You (we all) just need to better understand authentication.

It sounds like you're trying to make this a discussion about modification and trust itself, which is a very important topic, and many of us have a lot to say about it.

At best, viewers are currently at risk of disbelieving all images they see, because AI technology (Dall E etc) is getting so good. However, at worst, viewers are at risk of just naively believing fake images, and/or not caring about image theft, etc.

Either way, photographers face a near-total devaluation of their work, so this is a serious topic. And this technology is a huge step in the right direction.

I fully understand - It is YOU who don't (want to) understand that no digital image is safe from theft in any way.

It can be signed, you can proof its yours.
But you already could proof its yours without the technology by owning the RAW file.

The technology doesn't stop anybody from stealing your images if they want to. The 'verification' is a farce, a joke in every way. It is people like who who feel 'safer' with technology that doesn't protect you from anything except for giving you false hopes.

All you do is fall for the marketing of Sony.
But then again, what else is new.

@Name That Camera you don't understand is this technology is not about image theft but image authentication. If I steal your RAW image it doesn't matter how you "signed" it using existing tech I have an exact copy, including the signature. If I alter the image it's still signed by you so I can pass it off as your original. This Sony technology would prevent me doing that.

Digital signatures, which is what this is, are a well understood and common use of cryptography.

In the UK if you sent me a payment from your bank to mine the payment sent is not encrypted. Tacked on the end is a Message Authentication Code (MAC), a digital signature.

If I intercepted that payment and changed the amount from 10 to 100 the payment would be rejected as the receiving bank would recalculate the MAC based on the message contents and it would not be the same as the MAC included in the message.

The Sony tech works the same. If you think this kind of authentication is useless you have no clue.

And you don't want to understand that it won't stop anything.. and therefore being obnoxious.

As it should. Only the original unaltered image should work. You can always deliver an edited image along with the original for comparison and proof.

A hash is not a hashtag.

https://blog.elcomsoft.com/2011/04/nikon-image-authentication-system-compromised/

Both Nikon's and Canon's systems have been compromised, rendering the feature completely useless and robbing everyone that relied on it of the evidence they thought they had.

Hopefully the Sony implementation is not as naive as those by Nikon and Canon.

Re. Nikon and Canon: You don't need to have cryptography as your core competency to call a library.

Arcimboldo: you are making a beginner's mistake. Calling a library will not help you if you can't keep the private key secure. That's how the early systems got compromised.

See my other comment above: I assumed that Sony's BIONZ XR sports something like Apple's Secure Enclave. So, yes, you're right, for that part you would need advanced expertise - but that kind of thing is available from other sources, say, ARM, so it's again about using a kind of library.

No matter whether it's hardware or software, stitching together bits and pieces without understanding the whole system will not lead to a secure final result. It is not as simple as using one or more pre-existing components, especially if your are building a specialized system like here, which has not yet been built (in a secure manner) by a camera vendor before.

@Dave, It's using public key encryption.

That's my big question, because in Nikon's case, it apparently failed so badly that they gave up after only 1-2 years.

I don't think getting enough entropy for key generation is the problem here. Read the sensor at high ISO and there is plenty of noise. Or let the user take an image, throw it away, but take the image data to feed the PRNG.

Storing the key is a bigger concern. As is e.g. getting the correct time and date without built-in GPS.

Yea, because every photographer in the world should be generating 10,000 resource-hungry blockchain transactions a year.

I can see it now: we were on course to limiting climate change to 2°C, until the latest TikTok challenge hit the blockchain, and that's how we lost half of Africa and Central America and the rising oceans swallowed Australia.

Yeah, with something like Cardano or Algorand. I guess the other comment and people that liked it haven't heard of "proof of stake" tokens. Anyways, from my limited knowledge of how it all works... There would be a built-in cost per image. But for news and other photo jobs needing pure unmodified pictures I think it would be worth the cost. The camera would also need an internet connection to generate the contract on the chain for each image (put a cellular modem in or tether). Each contract would hold some type of signature of the image file rather than the image itself (CRC, etc). Though I think it would be smarter to have a signature of the raw image and a portion of the exif data rather than the entire file so that the contract link could be inserted into the image once the camera gets confirmation. You'd likely load up your camera with ADA (into a hardware wallet in the camera) that would be used to "pay" for each image contract. IMO it would be super cool tech!

SigmaChrome: to be fair, it is not the blockchain that is eco-unfriendly. It's systems like BitCoin that need lots of energey due to the "proof of work" approach.

SigmaChrome I don't know why do you comment on something you don't understand. Blockchain, as the name suggest, is simply data structure that has one certified origin and anyone can check that further appended data are valid, come from specific owner/node.

They beauty of open source blockchain is that the data is distributed over global network with many nodes and any changes to the structure or protocal ned to have approval of majority of nodes. In privately held blockchain this doesn't apply as one company holds all the nodes and design decisions, like a national bank control the whole banking system in a country.

Heavy energy cost only come from the need to have price for adding new block to chain that is used as store of value, chain of coins if you will. You don't need to use blockchain to contain wealth, therefore you don't need to secure the network by presenting increasingly difficult math problems to solve. List of transactions for credit card isn't energy intensive.

This is authentication and there would be no reason whatsoever to employ blockchain. The energy cost of signing the picture would be FAR less than creating the jpeg.

Blockchain by itself is not resource intensive. It's the use of artificially-difficult hash calculations during block creation that use up the most power, followed by the peer-peer trustless consensus mechanisms to validate blocks. Those are unique to cryptocurrencies like Bitcoin and Ether. You can use a centralized trust model for blockchain instead, for instance with certificate authorities like those used for SSL on websites.

Mainstream deepfake video creation is right around the corner too. That's going to be very scary once it matures.