I feel like you need to be read in, at least a little bit, into what's going on in cryptography research before you take statements like this at face value, because a lot of pretty serious people work on lightweight cryptography constructions and primitives.
(I'm not dissing Green here; I'm saying, I don't think he meant that statement to be proxied to a generalist audience as an effective summation of lightweight cryptography).
There’s a very serious point here. Cryptographers are and always have been deeply concerned with performance. Some of the most skilled low level optimization people I know work on cryptography. It is only relatively recently that computer hardware has gotten powerful enough that cryptography isn’t a serious bottleneck for production systems. In all the recent new crypto standards, a big decision in the whittling down of the finalists was performance.
If someone is telling you that we need a new, faster standard for cryptography, and the selling point is “faster,” you’d better wonder why that wasn’t the standard already in use. If there is not some novel, brand new algorithm that is being employed, the answer is because it is insecure. Or at least that it doesn’t meet the level of security for general use, which took a cryptographer is approximately the same thing.
Performance is an evolving target. Meta reported they spend ~0.05% (1 out of every 2000 CPU cycles) on X25519 key exchange within the last year, which is quite significant. If that can be brought down, that's worthwhile. And ongoing research and deployment of PQC will make key exchange even more expensive.
> If there is not some novel, brand new algorithm that is being employed, the answer is because it is insecure.
Lol that is just not true at all. A major point of discussion when NIST announced the SHA3 finalist being Keccak back in ~2012(?) was that BLAKE1 at the time offered significantly better software performance, which was considered an important practical reality, and was faster than SHA-2 at a higher (but insignificantly so) security margin; their own report admitted as much. The BLAKE1 family is still considered secure today, its internal HAIFA design is very close to existing known designs like Merkle–Damgård, it isn't some radically new thing.
So why did they pick Keccak? Because they figured that SHA-2 was plenty good and already deployed widely, so "SHA-2 but a little faster" was not as compelling as a standard that complimented it in hardware; they also liked Keccak's unique sponge design that was new and novel at the time and allowed AEAD, domain separation, etc. They admit ultimately any finalist including BLAKE would have been a good pick. You can go read all of this yourself. The Keccak team even has new work on more performant sponge-inspired designs, such as their work on Farfalle and deck functions.
The reality is that some standards are chosen for a bunch of reasons and performance is only one of them, though very important. But there's plenty of room for non-standard things, too.
> Meta reported they spend ~0.05% (1 out of every 2000 CPU cycles) on X25519 key exchange within the last year, which is quite significant.
That is not even remotely significant. Facebook spends 25% (1 out of every 4) of my CPU cycles on tracking. Pretty much anything else they optimize (are they still using python and php?) Will have a bigger impact.
> A major point of discussion when NIST announced the SHA3 finalist being Keccak back in ~2012(?) was that BLAKE1 at the time offered significantly better software performance
IIRC, Keccak had a smaller chip area than Blake. Hardware performance is more important than software performance if the algorithm is likely to be implemented in hardware, which is a good assumption for a NIST standard. Of course, SHA3 hasn't taken off yet but that's more to do with how good SHA2 is.
> BLAKE1 family is still considered secure today, its internal HAIFA design is very close to existing known designs like Merkle–Damgård, it isn't some radically new thing.
Given that the purpose of the competition was to replace SHA2 if/when it is weakened, choosing a similar construction would not have been the right choice.
> Hardware performance is more important than software performance if the algorithm is likely to be implemented in hardware
I don't think that's necessarily a given at all, but I grant that's mostly a matter of opinion I guess.
> Given that the purpose of the competition was to replace SHA2 if/when it is weakened
I think the dirty secret hiding there is that I see very few actual expectations SHA2 will ever be broken. Assuming it can be and picking a different secure construction, of course, is a good idea. But even the designers of BLAKE have admitted such and so did NIST.
An additional argument for Keccak was that even if its performance in software implementations was mediocre, it allows very fast and cheap hardware implementations, so from this POV it was definitely better than the alternatives.
> computer hardware has gotten powerful enough that cryptography isn’t a serious bottleneck for production systems ... someone is telling you that we need a new, faster standard for cryptography, and the selling point is “faster"
If the selling point is “faster,” you’d better wonder why that wasn’t the standard already in use.
Because “fast enough” is fast enough, unless it isn’t.
My desktop CPU has AES in hardware, so it’s fast enough to just run AES.
My phone’s ARM CPU doesn’t have AES in hardware, so it’s not fast enough. ChaCha20 is fast enough, though, and especially with the SIMD support on most ARM processors.
All this paper is saying is that ChaCha20 is not fast enough for some devices, and so folks had to put in intellectual effort to make a new thing that is.
But even further: everyone’s definition for “fast enough” is different. Cycles per byte matters more if you encrypt a lot of bytes.
Only extremely old ARM CPUs (i.e. 32-bit CPUs from more than a decade ago) do not have AES hardware. All 64-bit ARM CPUs have it, like also at least the HW for SHA-1 and SHA-256. The more recent ARM CPUs have HW support for more cryptographic algorithms than the majority of the desktop CPUs.
"Lightweight" cryptography is not intended for something as powerful as a smartphone, but only for microcontrollers that are embedded in small appliances, e.g. sensors that transmit wirelessly the acquired data.
Not all 64-bit ARM CPUs have AES support since it's part of the optional Crypto extension. The BCM2837 in the Raspberry Pi 4 is an ARMv8 CPU that lacks it.
You are right that the cryptographic instructions are optional and they have always been so in ARM CPUs, in order to enable the making of CPUs that can be embedded in products that will be exported even to destinations where cryptography is forbidden.
However the poster above was talking about smartphone Arm-based CPUs. I doubt that there has ever existed a 64-bit ARM-based CPU for smartphones that did not implement the cryptography extension. Even the CPUs having only Cortex-A53 cores that I am aware of, made by some Chinese companies for extremely cheap mobile phones, had this extension.
If the selling point is “faster,” you’d better wonder why that wasn’t the standard already in use.
Because the field of cryptography advances? You could have made the same argument about Salsa/ChaCha but those are great ciphers. And now we have Ascon which has the same security level but I guess is even faster.
They're faster than AES on their target platforms. It really feels like people are just trying to run with this out-of-context Matthew Green quote as if it was an axiom.
Rijndael (now AES) wasn't even the strongest finalist in the 2001 AES evaluation. It partially won on dint of being faster on contemporary x86 processors than Serpent or Twofish. Nowadays, it's faster on x86-64 processors because there's dedicated silicon for running it. Modern small platforms don't have this silicon and have different performance characteristics to consider. Also, without that dedicated silicon, implementations tend to be vulnerable to side-channel attacks that were unknown at the time.
Thought this was common knowledge. When TLS1.3 was standardized, it explicitly left out all NULL and weak (such as RC4) ciphers. It also left out all weaker RSA/static-DH key exchange methods, such that easy decryption of recorded communication became impossible. To that the enterprises who would like to snoop on their employees and the secret services who would like to snoop on everyone reacted negatively and tried to introduce their usual backdoors such as NULL ciphers again:
https://www.rfc-editor.org/rfc/rfc9150.html was the one reintroducing NULL ciphers into TLS1.3. RFC9150 is written by Cisco and ODVA who previously made a fortune with TLS interception/decryption/MitM gear, selling to enterprises as well as (most probably, Cisco has been a long-time bedmate of the US gov) spying governments.
The RFC weakly claims "IoT" as the intended audience due to cipher overhead, however, that is extremely hard to believe. They still do SHA256 for integrity, they still do all the very complicated and expensive TLS dance, but then skip encryption and break half the protocol on the way (since stuff like TLS1.3 RTT needs confidentiality). So why do all the expensive TLS dance at all when you can just slap a cheaper HMAC on each message and be done? The only sensible reason is that you want to have something in TLS to downgrade to.
How exactly do NULL ciphers accomplish enterprise monitoring goals? The point of the TLS 1.3 handshake improvements was to eliminate simple escrowed key passive monitoring. You could have the old PKZip cipher defined as a TLS 1.3 ciphersuite; that doesn't mean a middlebox can get anybody to use it. Can you explain how this would get any enterprise any access it doesn't already have?
The presence of an insecure ciphersuite in the TLS standard does not in fact imply the ability of a middlebox to force that ciphersuite; that's kind of the whole point of the TLS protocol. So, I ask again.
Your first set of links seems to be about central key logging for monitoring connection contents? If there's stuff about null encryption in there I missed it. And even if there is, it all sounds like explicit device configuration, not something you can trigger with a downgrade attack.
Yes, my first link is about that. It illustrates and explains the push to weaken TLS1.3 that has later been accomplished by the re-inclusion of NULL ciphers.
And all the earlier weaker ciphers were explicit device configuration as well. You could configure your webserver or client not to use them. But the problem is that there are easy accidental misconfigurations like "cipher-suite: ALL", well-intended misconfigurations like "we wan't to claim IoT support in marketing, so we need to enable IoT-'ciphers' by default!" and the sneaky underhanded versions of the aforementioned accidents. Proper design would actually just not create a product that can be mishandled, and early TLS1.3 had that property (at least with regards to cipher selection). Now it's back to "hope your config is sane" and "hope your vendor didn't screw up". Which is exactly what malicious people need to hide their intent and get in their decryption backdoors.
The first link is weakening in a way that is as far from a downgrade attack as you can possibly get. And on top of that TLS 1.3 has pretty good downgrade prevention as far as I know.
> well-intended misconfigurations like "we wan't to claim IoT support in marketing, so we need to enable IoT-'ciphers' by default!" and the sneaky underhanded versions of the aforementioned accidents
Maybe... This still feels like a thing that's only going to show up on local networks and you don't need attacks for local monitoring. Removing encryption across the Internet requires very special circumstances and also lets too many people in.
Security standards can move extremely slowly when the security of the incumbent algorithm hasn’t been sufficiently compromised, despite better (faster, smaller) alternatives.
> It is only relatively recently that computer hardware has gotten powerful enough that cryptography isn’t a serious bottleneck for production systems.
Or we've had enough "spare" transistors and die space to devote some to crypto, hashing, and checksumming instructions. I remember the splash Sun made when they announced on-die crypto hardware in 2007 (as well as on-die 10 GigE):
This thread fails to mention that a cipher has to be somewhat hard to compute or someone with a lot of resources can't just brute force it. Therefore you also want an implementation of a given cipher to be as efficient as possible, so that no future improvement lowers the security of your cipher.
"Lightweight" cryptography is not intended for smartphones, personal computers and similarly powerful devices.
It is intended only for microcontrollers embedded in various systems, e.g. the microcontrollers from a car or from a robot that automate various low-level functions (not the general system control), or from various sensors or appliances.
It is expected that the data exchanged by such microcontrollers is valuable only if it can be deciphered in real time.
If an attacker would be able to decipher the recorded encrypted data by brute force after a month, or even after a week, it is expected that the data will be useless. Otherwise, standard cryptography must be used.
The point of these primitives is not to trade security for ease of computation. The point is to find alternatives that are just as strong as AES-128 but with less computation. The trade-off is in how long and hard people have tried to break it.
That's one of the tradeoffs but the most subtextual of them; the bigger tradeoff is that lightweight constructions are optimized for constrained environments, and outperform only on platforms that don't already do a good job with things like AES. That's the answer for why we wouldn't "just call it cryptography": things that perform well (1) on constrained platforms (2) for the kinds of messages exchanged by constrained platforms will probably tend to get their asses kicked by AES on non-constrained platforms.
When you say "get their asses kicked", you mean in terms of performance right? Both sets of cryptography are secure under the same set of assumptions, it's just that one is more performant on limited instruction sets and the other is more performant on full featured instruction sets?
I'm saying that AES isn't viable a decent-sized chunk of existing embedded hardware, and that the constructions that are viable on those platforms both (1) fall below the security thresholds of front-line mainstream constructions like AES-GCM or Chapoly and (2) would in fact be slower that AES or Chapoly on modern workstation, server, and phone platforms.
Hardware capabilities vary widely; there isn't one optimal algorithm that fits (or, in the case of MCUs, is even viable) on every platform.
What's worse, efforts to shoehorn front-line mainstream constructions onto MCUs often result in insecure implementations, because, especially without hardware support (like carryless multiplication instructions), it's very difficult to get viable performance without introducing side channels.
> When you say "get their asses kicked", you mean in terms of performance right?
Depends on what you mean by "performance". It could be latency: high frequency traders (HFTs) could probably be happy if their order data is protected for "only" an hour if it means dropping latency from (e.g.) 42 nanoseconds down to 24.
An hour ago for some trading platforms is stale as a decade ago.
This is why eSTREAM’s decision to leave MCU-class software essentially unaddressed puzzles me. What do I use on an 8-bit or 16-bit controller with no wide multiplies and no barrel shifter? Before this (which seems fine if not exactly tiny at first glance), there was... Gimli (with the same caveats)? Speck?
I’m guessing that the chip area of hardware AES is utterly inconsequential compared to all the peripherals you get on a modern micro, but the manufacturers are going to keep charging specialized-applications money for that until we’re all on 32-bit ARMs with multipliers and ChaCha becomes viable.
I am in now way a crypto expert, just someone who enjoys casually reading about it. But in my mind even a security for speed trade off can be worthwhile. Otherwise the alternative is often: no crypto at all because it won't run on that tiny MCU and we don't have the budget for a hardware accelerator.
Even the tiniest MCU can typically perform more than one cryptographic operation per second. If your MCU has any cycles to spare at all it usually has enough cycles for cryptography.
If you don't have any cycles to spare, you can upgrade to an MCU that does have cycles to spare for less than $0.50 in small batches, and an even smaller price delta for larger batches.
Any device that doesn't use cryptography isn't using it because the manager has specifically de-prioritized it. If you can't afford the $0.50 per device, you probably can't afford the dev that knows his way around cryptography either.
> Even the tiniest MCU can typically perform more than one cryptographic operation per second. If your MCU has any cycles to spare at all it usually has enough cycles for cryptography.
Well, no. If you can do 1 AES block per second, that's a throughput of a blazing fast 16 bytes per second.
I know that's a pathological example, but I do understand your point - a typical workload on an MCU won't have to do much more than encrypt a few kilobytes per second for sending some telemetry back to a server. In that case, sure: ChaCha20-Poly1305 and your job is done.
However, what about streaming megabytes per second, such as an uncompressed video stream? In that case, lightweight crypto may start to make sense.
1 operation per second would refer to cryptographic signatures. If you are doing Chacha, the speeds are more like 1 mbps. AES is probably closer to 400 kbps.
An uncompressed video stream at 240p, 24 frames per second is 60 mbps, not really something an IoT device can handle. And if the video is compressed, decompression is going to be significantly more expensive than AES - adding encryption is not a meaningful computational overhead.
>Even the tiniest MCU can typically perform more than one cryptographic operation per second. If your MCU has any cycles to spare at all it usually has enough cycles for cryptography.
>1 operation per second would refer to cryptographic signatures. If you are doing Chacha, the speeds are more like 1 mbps. AES is probably closer to 400 kbps.
It sounds to me like you, sir or madame, have not worked with truly tiny MCUs. :-)
But yes, there are inexpensive MCUs where you can do quite a bit of crypto in software at decent speeds.
Why would you compare an uncompressed video stream to anything in this discussion? Especially at such a small frame size in modern video usage.
Modern encrypted streaming uses pre-existing compressed video where the packets are encrypted on the way to you by the streaming server. It would have to uniquely encrypt the data being sent to every single user hitting that server. So it's not just a one and done type of thing. It is every bit of data for every user. So that scales to a lot of CPU on the server side to do the encryption. Yes, on the receiving side while your device is only dealing with the one single stream, more CPU cycles will be spent decompressing the video compared to decrypting. But again, that's only have of the encrypt/decrypt cycle
A VPN’s encryption is different than a streaming platform’s encryption. The streamer’s encryption is their form of rights management. So the device/app decompressing the video very much is the point of decryption. If not, the rights management is very broken. If the small edge gateway is somehow decrypting the video stream, you’d have a very rogue device that lots of people would be curious to learn more
>If you don't have any cycles to spare, you can upgrade to an MCU that does have cycles to spare for less than $0.50 in small batches, and an even smaller price delta for larger batches.
The monetary cost is most likely not the problem. Tacking on significant additional work is bound to consume more power and generate heat. Tiny devices often have thermal and power limits to consider.
The problem is it isn't significant work. You can perform ChaCha8 on pen and paper in ~30 minutes. It's literally 768 single cycle 32 bit integer operations.
768 single cycle instructions to do WHAT? To encode a few bytes or something? You can't convince me that the cost of cryptography is never prohibitive. Boldly asserting that your idea of "significant work" is universal for all applications isn't going to convince me, nor are weird anecdotes about tangentially related things.
I am not up to speed on these new algorithms. I still remember there was a light weight cryptography algorithm a few years ago championed by the NSA that had a subtle (possibly deliberate) flaw in it.
When dealing with cryptography it is always necessary to remember cryptography is developed and operates in an adversarial environment.
Speck? To my knowledge there aren't any serious flaws despite a lot of public cryptanalysis. I think what sank Speck was that it came out a few years after after the Dual_EC_DRBG fiasco and nobody was ready to trust an NSA developed cipher yet - which is fair enough. The NSA burned their credibility for decades with Dual_EC_DRBG.
I mean, yeah, but also Simon and Speck aren't as good as the new generation of low-footprint designs like Ascon and Xoodyak. We know more about how to do these things now than we did 15 years ago.
In what ways is it better? Security margin or something? I thought Speck has held up pretty well to cryptanalysis (unlike you I'm not in the security field so maybe I'm wrong).
I quite liked the remarkable simplicity of Speck. Performance was better than Ascon in my limited testing. It seems like it should be smaller on-die or in bytes of code, and with possibly lower power consumption. And round key generation was possible to compute on-the-fly (reusing the round code!) for truly tiny processors.
RFID tags might be rather in the area of $0.03. Other devices might have some hard memory constraints. For other targets it is rather about making hacking a bit more difficult rather that protecting sensitive data against malicious international actors.
> Devices that can benefit from its approach include RFID tags, implanted medical devices, and toll-registration transponders attached to car windshields
Eh, the problem with AES is side channel attacks in software implementations. They aren't necessarily obvious, especially if you're deploying on an oddball CPU architecture.
This standard targets hardware without AES accelerators like microcontrollers. Now, realistically, ChaCha20-Poly1305 is probably a good fit for most of those use cases, but it's not a NIST standard so from the NIST perspective it doesn't apply. Ascon is also a fair bit faster than ChaCha20 on resource constrained hardware (1.5x-3x based on some benchmarks I found online).
Its slowness in software and quickness in hardware have almost nothing to do with it being sponge-based, but are caused by the Boolean functions executed by the Keccak algorithm, which are easy to implement in hardware, but need many instructions on most older CPUs (but much less instructions on Armv9 CPUs or AMD/Intel CPUs with AVX-512).
The sponge construction is not inherently slower than the Merkle–Damgård construction. One could reuse the functions iterated by SHA-512 or by SHA-256 and reorganize them to be used in a sponge-based algorithm, obtaining similar speeds with the standard algorithms.
That is not done because for the sponge construction it is better to design a mixing function with a single wider input instead of a mixing function with 2 narrower inputs, like for Merkle–Damgård. Therefore it is better to design the function from the beginning for being used inside a sponge construction, instead of trying to adapt functions intended for other purposes.
Nothing stops us from qualifying "cryptography" with an adjective to refer to a specific part of it, while still referring to full-fledged cryptography. See the Wikipedia article for examples.
Those qualifiers in front of cryptography are very different from lightweight. Lightweight doesn’t convey a good meaning in general and that is the main issue.
I haven’t looked at the lightweight crypto proposals though but if NIST is proposing it I am optimistic. If the proposals from various cryptographers around the world made through all the process it’s pretty good.
This is now besides the original point, which was that in cryptography's case, "lightweight" would work like "alternative" works for "alternative medicine".
However if we shift the focus to marketing, lightweight works nothing like that in that regard. They use it to imply that the product is fast, lean, to the point, contains nothing unnecessary, has fewer, but solid options, energetic, aspiring. Implying that the other products are heavier, bulkier, cumbersome, bloated, slow, overcumbered, overly complex.
It's already used in for example LDAP, which I don't think anyone scoffs at (well, with regards to the lightweight qualifier), and lightweight software has its own wiki page, implying that the usage of the term wrt/ software is widespread enough to be notable. They write that this means "a computer program that is designed to have a small memory footprint, low CPU usage, overall a low usage of system resources".
All this to write that lightweight conveys a good meaning indeed. About the only widespread negative usage I can think of is when people call others "lightweights", implying that they performed below expectations.
I hear you, but they could have chosen performant crypto or low power crypto or some variant like that. For many lightweight reminds them of low effort ciphers. I see your point though
For me if it comes from NIST, I trust it (even with their past shenanigans) because I know the process they use and also they only produce these standards to be used for federal governent software. So they will do a good job.
Algorithms tend to fall pretty squarely in either the “prevent your sibling from reading your diary” or the “prevent the NSA and Mossad from reading your Internet traffic” camps.
Computers get faster every year, so a cipher with a very narrow safety margin will tend to become completely broken rapidly.
Some things must be encrypted well enough so that even if NSA records them now, even 10 years or 20 years later they will not be able to decipher them.
Other things must be encrypted only well enough so that nobody will be able to decipher them close to real time. If the adversaries decipher them by brute force after a week, the data will become useless by that time.
Lightweight cryptography is for this latter use case.
The hard part is that anything designed to be breakable in a week is only ~150x away in strength from being broken in an hour. That means you need to be incredibly confident about how strong your algorithm is. It's much easier to eat a little bit of cost such that you think it's invulnerable for thousands of years because that way you don't need to worry about a factor of 2 here and there.
Right. Strength in a shadow contest like cryptography can be, best case, estimated to sixteen-bit orders of magnitude (+-65000x). Just because you can't break it doesn't mean somebody else secretly knows a game changing way to break it. So you keep padding with huge exponential hedges such that if they scan shave a dozen bits off the strength of the scheme, it's still secure under finite resources.
They chose Ascon which is a good set of same sponge-based cryptographic functions if you don't have hardware acceleration for AES or the CPU resources for chacha20 which is the intention of the standard. The security is 128-bits (comparable to AES-128).
This is pretty cool. But IOT tends to fail hard on key agreement. And nothing here solves that. This seems to pretty much require a pre installed key, otherwise the overhead of securely installing a key would probably nullify the advantage of this encryption.
My point was that AES and SHA are not the reason IOT cryptography is so often broken or missing.
Instead its getting the keys onto the system in a halfway secure manner that is the blocking issue.
Hence I'd be a lot more enthusiastic about NIST guidance on these points.
A pairing system as seen in e.g. Zigbee or BLE seems pretty good to me. Not everyone cares to implement it well and there's still no standard for web-based devices, but it's here and it works.
I'd like to see more devices able to pair with NFC, but even that's standardised for Bluetooth, just underused.
Ah, I see. That's indeed an interesting point. At any rate, IOT cryptography can use this standard as a building point, so it is a step in the right direction anyway.
The world that the algorithm targets is exactly where is this doable. MCUs typically have a protected OTP area that makes it a good place to inject keys right in the factory.
Are these for garage doors and doorbells? Those devices could definitely use more security (it's not hard to stuff a proper TLS stack in a microcontroller but manufacturers balk at even putting something as cheap as a ESP32 in their BOM).
Also for currently-constrained platforms for which hardware updates are infeasible, software updates aren't, and cryptographic security is currently compromised due to lack of availability of appropriate constructions.
Wikipedia says Ascon has 320 bits of state and uses 5 bit s-boxes. That’s tiny compared to sha-256 or Blake2. One would think a pre image attack would be much more tractable at that scale.
This says very little about the strength of the cipher.
The initial state of ChaCha20 also has only at most 320 unknown bits (512 bits - 128 constant bits - 64 bits of a nonce). Actually you normally also know the counter, so there are only 256 unknown bits.
Of course the actual strength of the cipher cannot exceed the size of the state, but the design strength must be much lower for this cipher. It competes with AES-128, which is designed for an 128-bit strength.
320 bits of state is more than enough for a cipher that must have an 128-bit strength, or even for a cipher designed for a 256-bit strength, like AES-256 or ChaCha20.
Sponges divide state into rate (r) and capacity (c). They "absorb" incoming bytes, perform permutation (moving bits around without losing any of them) on the whole state, and "squeeze" out bytes from the rate, while capacity remains hidden.
For the secure hash function, the capacity should be at least twice the target, that is for 128-bit security you need 256 bits of capacity. ASCON hash uses 256 bit capacity and 320-256 = 64 bit rate, so to get a 32-byte hash of a 8-byte string (without padding), you'll need to do at least 4 permutations.
If you can design a secure permutation that permutes 257 bits, you can make a secure, but impractical hash function from it by setting the rate to 1 bit.
For the duplex mode that's used for authenticated encryption, capacity can be lower, because it's keyed -- it's 192 bits in ASCON.
This assumes the permutation of the 320-byte state itself is secure, of course.
If these primitives are less resource intensive than what we use today with the same level of security, then why don't we just use these everywhere? If they are not as secure, then why would be use these anywhere?
Your front door isn't a hardened steel vault door... I'm assuming you still lock it.
This is a bad analogy, even it it's apt. Only in that an encryption scheme can be technically weaker than another, but still plenty secure. As long as it does the job and hasn't been broken.
For example, part of the spec is simply hashing and packet signing... not necessarily making the packets secret, but authenticating the origin/source. It isn't necessarily about creating the most secure vault in the world, but securing what might otherwise be completely insecure channels of communications/attack. It's also not about a $1000 smart phone, but a micro controller that's a fraction of a dollar on devices that have a total BoM that is very low cost or small.
"Lightweight" here doesn't stop it from proper cryptography. In this case, the qualifier doesn't take away from the original meaning, just focuses it on a subset. Like "public-key cryptography".
> If lightweight cryptography was a good idea, we’d just call it “cryptography.”
https://x.com/matthew_d_green/status/1948476801824485606
I feel like you need to be read in, at least a little bit, into what's going on in cryptography research before you take statements like this at face value, because a lot of pretty serious people work on lightweight cryptography constructions and primitives.
(I'm not dissing Green here; I'm saying, I don't think he meant that statement to be proxied to a generalist audience as an effective summation of lightweight cryptography).
There’s a very serious point here. Cryptographers are and always have been deeply concerned with performance. Some of the most skilled low level optimization people I know work on cryptography. It is only relatively recently that computer hardware has gotten powerful enough that cryptography isn’t a serious bottleneck for production systems. In all the recent new crypto standards, a big decision in the whittling down of the finalists was performance.
If someone is telling you that we need a new, faster standard for cryptography, and the selling point is “faster,” you’d better wonder why that wasn’t the standard already in use. If there is not some novel, brand new algorithm that is being employed, the answer is because it is insecure. Or at least that it doesn’t meet the level of security for general use, which took a cryptographer is approximately the same thing.
Performance is an evolving target. Meta reported they spend ~0.05% (1 out of every 2000 CPU cycles) on X25519 key exchange within the last year, which is quite significant. If that can be brought down, that's worthwhile. And ongoing research and deployment of PQC will make key exchange even more expensive.
> If there is not some novel, brand new algorithm that is being employed, the answer is because it is insecure.
Lol that is just not true at all. A major point of discussion when NIST announced the SHA3 finalist being Keccak back in ~2012(?) was that BLAKE1 at the time offered significantly better software performance, which was considered an important practical reality, and was faster than SHA-2 at a higher (but insignificantly so) security margin; their own report admitted as much. The BLAKE1 family is still considered secure today, its internal HAIFA design is very close to existing known designs like Merkle–Damgård, it isn't some radically new thing.
So why did they pick Keccak? Because they figured that SHA-2 was plenty good and already deployed widely, so "SHA-2 but a little faster" was not as compelling as a standard that complimented it in hardware; they also liked Keccak's unique sponge design that was new and novel at the time and allowed AEAD, domain separation, etc. They admit ultimately any finalist including BLAKE would have been a good pick. You can go read all of this yourself. The Keccak team even has new work on more performant sponge-inspired designs, such as their work on Farfalle and deck functions.
The reality is that some standards are chosen for a bunch of reasons and performance is only one of them, though very important. But there's plenty of room for non-standard things, too.
> Meta reported they spend ~0.05% (1 out of every 2000 CPU cycles) on X25519 key exchange within the last year, which is quite significant.
That is not even remotely significant. Facebook spends 25% (1 out of every 4) of my CPU cycles on tracking. Pretty much anything else they optimize (are they still using python and php?) Will have a bigger impact.
Or just reduce the quality of served photos and videos by 0.05%, probably nobody would notice
> Facebook spends 25% (1 out of every 4) of my CPU cycles on tracking.
That's their core business.
This is significant for Facebook, not for you. I solved this problem for myself by not using Facebook.
> A major point of discussion when NIST announced the SHA3 finalist being Keccak back in ~2012(?) was that BLAKE1 at the time offered significantly better software performance
IIRC, Keccak had a smaller chip area than Blake. Hardware performance is more important than software performance if the algorithm is likely to be implemented in hardware, which is a good assumption for a NIST standard. Of course, SHA3 hasn't taken off yet but that's more to do with how good SHA2 is.
> BLAKE1 family is still considered secure today, its internal HAIFA design is very close to existing known designs like Merkle–Damgård, it isn't some radically new thing.
Given that the purpose of the competition was to replace SHA2 if/when it is weakened, choosing a similar construction would not have been the right choice.
> Hardware performance is more important than software performance if the algorithm is likely to be implemented in hardware
I don't think that's necessarily a given at all, but I grant that's mostly a matter of opinion I guess.
> Given that the purpose of the competition was to replace SHA2 if/when it is weakened
I think the dirty secret hiding there is that I see very few actual expectations SHA2 will ever be broken. Assuming it can be and picking a different secure construction, of course, is a good idea. But even the designers of BLAKE have admitted such and so did NIST.
An additional argument for Keccak was that even if its performance in software implementations was mediocre, it allows very fast and cheap hardware implementations, so from this POV it was definitely better than the alternatives.
> computer hardware has gotten powerful enough that cryptography isn’t a serious bottleneck for production systems ... someone is telling you that we need a new, faster standard for cryptography, and the selling point is “faster"
Google needed faster than standard AES cryptography for File-based Encryption on Android Go (low cost) devices: https://tosc.iacr.org/index.php/ToSC/article/view/7360 / https://security.googleblog.com/2019/02/introducing-adiantum...
This doesn't appear to use LWC constructions though, mostly ChaCha20.
My desktop CPU has AES in hardware, so it’s fast enough to just run AES.
My phone’s ARM CPU doesn’t have AES in hardware, so it’s not fast enough. ChaCha20 is fast enough, though, and especially with the SIMD support on most ARM processors.
All this paper is saying is that ChaCha20 is not fast enough for some devices, and so folks had to put in intellectual effort to make a new thing that is.
But even further: everyone’s definition for “fast enough” is different. Cycles per byte matters more if you encrypt a lot of bytes.
Only extremely old ARM CPUs (i.e. 32-bit CPUs from more than a decade ago) do not have AES hardware. All 64-bit ARM CPUs have it, like also at least the HW for SHA-1 and SHA-256. The more recent ARM CPUs have HW support for more cryptographic algorithms than the majority of the desktop CPUs.
"Lightweight" cryptography is not intended for something as powerful as a smartphone, but only for microcontrollers that are embedded in small appliances, e.g. sensors that transmit wirelessly the acquired data.
> Only extremely old ARM CPUs (i.e. 32-bit CPUs from more than a decade ago) do not have AES hardware.
I remember when Sun announced the UltraSPARC T2 in 2007 which had on-die hardware for (3)DES, AES, SHA, RSA, etc:
* https://en.wikipedia.org/wiki/UltraSPARC_T2
(It also had two 10 GigE modules right on the chip.)
I think I’m just getting old. I have a collection of KaiOS phones with Cortex-A7s but now they’re over a decade old :(
Newer ones have the Qualcomm 215, which, yes, is 64-bit 4x A53
From that perspective, LWC is only useful on old (existing?) microcontrollers: the Cortex-A320 that came out this year is 64-bit.
Hardware cycles take time, though, and it will be some time before everything is 64-bit!
Not all 64-bit ARM CPUs have AES support since it's part of the optional Crypto extension. The BCM2837 in the Raspberry Pi 4 is an ARMv8 CPU that lacks it.
You are right that the cryptographic instructions are optional and they have always been so in ARM CPUs, in order to enable the making of CPUs that can be embedded in products that will be exported even to destinations where cryptography is forbidden.
However the poster above was talking about smartphone Arm-based CPUs. I doubt that there has ever existed a 64-bit ARM-based CPU for smartphones that did not implement the cryptography extension. Even the CPUs having only Cortex-A53 cores that I am aware of, made by some Chinese companies for extremely cheap mobile phones, had this extension.
If the selling point is “faster,” you’d better wonder why that wasn’t the standard already in use.
Because the field of cryptography advances? You could have made the same argument about Salsa/ChaCha but those are great ciphers. And now we have Ascon which has the same security level but I guess is even faster.
If these were faster than AES and as strong as AES, they would be replacing AES, not only being used for "lightweight devices unable to use AES"
They're faster than AES on their target platforms. It really feels like people are just trying to run with this out-of-context Matthew Green quote as if it was an axiom.
Rijndael (now AES) wasn't even the strongest finalist in the 2001 AES evaluation. It partially won on dint of being faster on contemporary x86 processors than Serpent or Twofish. Nowadays, it's faster on x86-64 processors because there's dedicated silicon for running it. Modern small platforms don't have this silicon and have different performance characteristics to consider. Also, without that dedicated silicon, implementations tend to be vulnerable to side-channel attacks that were unknown at the time.
> If these were faster than AES and as strong as AES […]
Not everything needs to be as strong as AES, just "strong enough" for the purpose.
Heck, the IETF has published TLS cipher suites with zero encryption, "TLS 1.3 Authentication and Integrity-Only Cipher Suites":
* https://datatracker.ietf.org/doc/html/rfc9150
Lightweight cryptography could be a step between the above zero and the 'heavyweight' ciphers like AES.
NULL ciphers in TLS are intended to enable downgrade attacks. Nothing else.
Same thing with weaker ciphers. They are a target to downgrade to, if an attacker wishes to break into your connection.
> NULL ciphers in TLS are intended to enable downgrade attacks. Nothing else.
Intended... Do any experts think that? Can you cite a couple? Or direct evidence of course.
Unless I'm missing a joke.
Thought this was common knowledge. When TLS1.3 was standardized, it explicitly left out all NULL and weak (such as RC4) ciphers. It also left out all weaker RSA/static-DH key exchange methods, such that easy decryption of recorded communication became impossible. To that the enterprises who would like to snoop on their employees and the secret services who would like to snoop on everyone reacted negatively and tried to introduce their usual backdoors such as NULL ciphers again:
https://www.nist.gov/news-events/news/2024/01/new-nccoe-guid... with associated HN discussion https://news.ycombinator.com/item?id=39849754
https://www.rfc-editor.org/rfc/rfc9150.html was the one reintroducing NULL ciphers into TLS1.3. RFC9150 is written by Cisco and ODVA who previously made a fortune with TLS interception/decryption/MitM gear, selling to enterprises as well as (most probably, Cisco has been a long-time bedmate of the US gov) spying governments. The RFC weakly claims "IoT" as the intended audience due to cipher overhead, however, that is extremely hard to believe. They still do SHA256 for integrity, they still do all the very complicated and expensive TLS dance, but then skip encryption and break half the protocol on the way (since stuff like TLS1.3 RTT needs confidentiality). So why do all the expensive TLS dance at all when you can just slap a cheaper HMAC on each message and be done? The only sensible reason is that you want to have something in TLS to downgrade to.
How exactly do NULL ciphers accomplish enterprise monitoring goals? The point of the TLS 1.3 handshake improvements was to eliminate simple escrowed key passive monitoring. You could have the old PKZip cipher defined as a TLS 1.3 ciphersuite; that doesn't mean a middlebox can get anybody to use it. Can you explain how this would get any enterprise any access it doesn't already have?
> How exactly do NULL ciphers accomplish enterprise monitoring goals?
I don't understand how this isn't obvious. Unencrypted means it is monitorable.
The presence of an insecure ciphersuite in the TLS standard does not in fact imply the ability of a middlebox to force that ciphersuite; that's kind of the whole point of the TLS protocol. So, I ask again.
Your first set of links seems to be about central key logging for monitoring connection contents? If there's stuff about null encryption in there I missed it. And even if there is, it all sounds like explicit device configuration, not something you can trigger with a downgrade attack.
Yes, my first link is about that. It illustrates and explains the push to weaken TLS1.3 that has later been accomplished by the re-inclusion of NULL ciphers.
And all the earlier weaker ciphers were explicit device configuration as well. You could configure your webserver or client not to use them. But the problem is that there are easy accidental misconfigurations like "cipher-suite: ALL", well-intended misconfigurations like "we wan't to claim IoT support in marketing, so we need to enable IoT-'ciphers' by default!" and the sneaky underhanded versions of the aforementioned accidents. Proper design would actually just not create a product that can be mishandled, and early TLS1.3 had that property (at least with regards to cipher selection). Now it's back to "hope your config is sane" and "hope your vendor didn't screw up". Which is exactly what malicious people need to hide their intent and get in their decryption backdoors.
The first link is weakening in a way that is as far from a downgrade attack as you can possibly get. And on top of that TLS 1.3 has pretty good downgrade prevention as far as I know.
> well-intended misconfigurations like "we wan't to claim IoT support in marketing, so we need to enable IoT-'ciphers' by default!" and the sneaky underhanded versions of the aforementioned accidents
Maybe... This still feels like a thing that's only going to show up on local networks and you don't need attacks for local monitoring. Removing encryption across the Internet requires very special circumstances and also lets too many people in.
Most modern processors have hardware support for AES, that's why it's fast. ChaCha is significantly faster when run on the CPU
Security standards can move extremely slowly when the security of the incumbent algorithm hasn’t been sufficiently compromised, despite better (faster, smaller) alternatives.
Tech Politics comes into it.
I mean, they are faster and as strong and are gradually replacing it?
Sponges generally? Maybe? LWC constructions not so much?
I thought the “they” being referenced were chacha/salsa.
"If there is not some novel, brand new algorithm that is being employed, the answer is because it is insecure."
No, this is not at all true.
> It is only relatively recently that computer hardware has gotten powerful enough that cryptography isn’t a serious bottleneck for production systems.
Or we've had enough "spare" transistors and die space to devote some to crypto, hashing, and checksumming instructions. I remember the splash Sun made when they announced on-die crypto hardware in 2007 (as well as on-die 10 GigE):
* https://en.wikipedia.org/wiki/UltraSPARC_T2
* PDF: https://www.oracle.com/technetwork/server-storage/solaris/do...
This thread fails to mention that a cipher has to be somewhat hard to compute or someone with a lot of resources can't just brute force it. Therefore you also want an implementation of a given cipher to be as efficient as possible, so that no future improvement lowers the security of your cipher.
"Lightweight" cryptography is not intended for smartphones, personal computers and similarly powerful devices.
It is intended only for microcontrollers embedded in various systems, e.g. the microcontrollers from a car or from a robot that automate various low-level functions (not the general system control), or from various sensors or appliances.
It is expected that the data exchanged by such microcontrollers is valuable only if it can be deciphered in real time.
If an attacker would be able to decipher the recorded encrypted data by brute force after a month, or even after a week, it is expected that the data will be useless. Otherwise, standard cryptography must be used.
... without, however, creating the impression that Ascon or Xoodyak could be broken by brute force in a week, a month, or a century.
That feels too reductive.
The point of these primitives is not to trade security for ease of computation. The point is to find alternatives that are just as strong as AES-128 but with less computation. The trade-off is in how long and hard people have tried to break it.
That's one of the tradeoffs but the most subtextual of them; the bigger tradeoff is that lightweight constructions are optimized for constrained environments, and outperform only on platforms that don't already do a good job with things like AES. That's the answer for why we wouldn't "just call it cryptography": things that perform well (1) on constrained platforms (2) for the kinds of messages exchanged by constrained platforms will probably tend to get their asses kicked by AES on non-constrained platforms.
When you say "get their asses kicked", you mean in terms of performance right? Both sets of cryptography are secure under the same set of assumptions, it's just that one is more performant on limited instruction sets and the other is more performant on full featured instruction sets?
I'm saying that AES isn't viable a decent-sized chunk of existing embedded hardware, and that the constructions that are viable on those platforms both (1) fall below the security thresholds of front-line mainstream constructions like AES-GCM or Chapoly and (2) would in fact be slower that AES or Chapoly on modern workstation, server, and phone platforms.
Hardware capabilities vary widely; there isn't one optimal algorithm that fits (or, in the case of MCUs, is even viable) on every platform.
What's worse, efforts to shoehorn front-line mainstream constructions onto MCUs often result in insecure implementations, because, especially without hardware support (like carryless multiplication instructions), it's very difficult to get viable performance without introducing side channels.
> When you say "get their asses kicked", you mean in terms of performance right?
Depends on what you mean by "performance". It could be latency: high frequency traders (HFTs) could probably be happy if their order data is protected for "only" an hour if it means dropping latency from (e.g.) 42 nanoseconds down to 24.
An hour ago for some trading platforms is stale as a decade ago.
This is why eSTREAM’s decision to leave MCU-class software essentially unaddressed puzzles me. What do I use on an 8-bit or 16-bit controller with no wide multiplies and no barrel shifter? Before this (which seems fine if not exactly tiny at first glance), there was... Gimli (with the same caveats)? Speck?
I’m guessing that the chip area of hardware AES is utterly inconsequential compared to all the peripherals you get on a modern micro, but the manufacturers are going to keep charging specialized-applications money for that until we’re all on 32-bit ARMs with multipliers and ChaCha becomes viable.
I am in now way a crypto expert, just someone who enjoys casually reading about it. But in my mind even a security for speed trade off can be worthwhile. Otherwise the alternative is often: no crypto at all because it won't run on that tiny MCU and we don't have the budget for a hardware accelerator.
Even the tiniest MCU can typically perform more than one cryptographic operation per second. If your MCU has any cycles to spare at all it usually has enough cycles for cryptography.
If you don't have any cycles to spare, you can upgrade to an MCU that does have cycles to spare for less than $0.50 in small batches, and an even smaller price delta for larger batches.
Any device that doesn't use cryptography isn't using it because the manager has specifically de-prioritized it. If you can't afford the $0.50 per device, you probably can't afford the dev that knows his way around cryptography either.
> Even the tiniest MCU can typically perform more than one cryptographic operation per second. If your MCU has any cycles to spare at all it usually has enough cycles for cryptography.
Well, no. If you can do 1 AES block per second, that's a throughput of a blazing fast 16 bytes per second.
I know that's a pathological example, but I do understand your point - a typical workload on an MCU won't have to do much more than encrypt a few kilobytes per second for sending some telemetry back to a server. In that case, sure: ChaCha20-Poly1305 and your job is done.
However, what about streaming megabytes per second, such as an uncompressed video stream? In that case, lightweight crypto may start to make sense.
1 operation per second would refer to cryptographic signatures. If you are doing Chacha, the speeds are more like 1 mbps. AES is probably closer to 400 kbps.
An uncompressed video stream at 240p, 24 frames per second is 60 mbps, not really something an IoT device can handle. And if the video is compressed, decompression is going to be significantly more expensive than AES - adding encryption is not a meaningful computational overhead.
>Even the tiniest MCU can typically perform more than one cryptographic operation per second. If your MCU has any cycles to spare at all it usually has enough cycles for cryptography.
>1 operation per second would refer to cryptographic signatures. If you are doing Chacha, the speeds are more like 1 mbps. AES is probably closer to 400 kbps.
It sounds to me like you, sir or madame, have not worked with truly tiny MCUs. :-)
But yes, there are inexpensive MCUs where you can do quite a bit of crypto in software at decent speeds.
Why would you compare an uncompressed video stream to anything in this discussion? Especially at such a small frame size in modern video usage.
Modern encrypted streaming uses pre-existing compressed video where the packets are encrypted on the way to you by the streaming server. It would have to uniquely encrypt the data being sent to every single user hitting that server. So it's not just a one and done type of thing. It is every bit of data for every user. So that scales to a lot of CPU on the server side to do the encryption. Yes, on the receiving side while your device is only dealing with the one single stream, more CPU cycles will be spent decompressing the video compared to decrypting. But again, that's only have of the encrypt/decrypt cycle
If it's compressed you don't need to decompress it first.
on the receiving end.
The device doing the decryption may not be the same device that does the decompression.
Eg a small edge gateway could be doing the VPN, while the end device is decoding the video.
A VPN’s encryption is different than a streaming platform’s encryption. The streamer’s encryption is their form of rights management. So the device/app decompressing the video very much is the point of decryption. If not, the rights management is very broken. If the small edge gateway is somehow decrypting the video stream, you’d have a very rogue device that lots of people would be curious to learn more
CCTV doesn’t need rights management. Nor do advertising billboards.
This whole framing is weird, because you can't spend $0.50 per already-deployed part to upgrade to something that can viably do AES.
>If you don't have any cycles to spare, you can upgrade to an MCU that does have cycles to spare for less than $0.50 in small batches, and an even smaller price delta for larger batches.
The monetary cost is most likely not the problem. Tacking on significant additional work is bound to consume more power and generate heat. Tiny devices often have thermal and power limits to consider.
The problem is it isn't significant work. You can perform ChaCha8 on pen and paper in ~30 minutes. It's literally 768 single cycle 32 bit integer operations.
768 single cycle instructions to do WHAT? To encode a few bytes or something? You can't convince me that the cost of cryptography is never prohibitive. Boldly asserting that your idea of "significant work" is universal for all applications isn't going to convince me, nor are weird anecdotes about tangentially related things.
I am not up to speed on these new algorithms. I still remember there was a light weight cryptography algorithm a few years ago championed by the NSA that had a subtle (possibly deliberate) flaw in it.
When dealing with cryptography it is always necessary to remember cryptography is developed and operates in an adversarial environment.
Speck? To my knowledge there aren't any serious flaws despite a lot of public cryptanalysis. I think what sank Speck was that it came out a few years after after the Dual_EC_DRBG fiasco and nobody was ready to trust an NSA developed cipher yet - which is fair enough. The NSA burned their credibility for decades with Dual_EC_DRBG.
Speck uses less resources to implement and is faster when I have tested it to compared ASCON.
I think the biggest problem is how they went about trying standardize it back in the day.
I mean, yeah, but also Simon and Speck aren't as good as the new generation of low-footprint designs like Ascon and Xoodyak. We know more about how to do these things now than we did 15 years ago.
In what ways is it better? Security margin or something? I thought Speck has held up pretty well to cryptanalysis (unlike you I'm not in the security field so maybe I'm wrong).
I quite liked the remarkable simplicity of Speck. Performance was better than Ascon in my limited testing. It seems like it should be smaller on-die or in bytes of code, and with possibly lower power consumption. And round key generation was possible to compute on-the-fly (reusing the round code!) for truly tiny processors.
Makes sense! Also, how does Speck fare in power analysis side channel attacks vs Ascon? My understanding was that was also one of the NIST criteria.
I am way out of my depth both on power consumption and leakage, but presumable Ascon does better on both counts than Chapoly.
Realy ChaCha seems trivially implementable without leaking anything.
That would be correct if the question was ‘For a $2000 32-core desktop should we switch to lightweight cryptography’.
The question is should we switch from some ridiculously insecure crappy crypto on a $3 part to this better lightweight crypto implementation.
Yah, we probably should, it’s better than what we had right?
RFID tags might be rather in the area of $0.03. Other devices might have some hard memory constraints. For other targets it is rather about making hacking a bit more difficult rather that protecting sensitive data against malicious international actors. > Devices that can benefit from its approach include RFID tags, implanted medical devices, and toll-registration transponders attached to car windshields
Good news. We’ve waited so long that AES is pretty damn lightweight now too.
Eh, the problem with AES is side channel attacks in software implementations. They aren't necessarily obvious, especially if you're deploying on an oddball CPU architecture.
This standard targets hardware without AES accelerators like microcontrollers. Now, realistically, ChaCha20-Poly1305 is probably a good fit for most of those use cases, but it's not a NIST standard so from the NIST perspective it doesn't apply. Ascon is also a fair bit faster than ChaCha20 on resource constrained hardware (1.5x-3x based on some benchmarks I found online).
Knowing how much of a dog SHA-3 is due to its sponge construction basis, it’s superficially surprising to me to see a sponge-based LWC algorithm.
I guess we’ve had quite a few years to improve things.
SHA-3 in fast when implemented in hardware.
Its slowness in software and quickness in hardware have almost nothing to do with it being sponge-based, but are caused by the Boolean functions executed by the Keccak algorithm, which are easy to implement in hardware, but need many instructions on most older CPUs (but much less instructions on Armv9 CPUs or AMD/Intel CPUs with AVX-512).
The sponge construction is not inherently slower than the Merkle–Damgård construction. One could reuse the functions iterated by SHA-512 or by SHA-256 and reorganize them to be used in a sponge-based algorithm, obtaining similar speeds with the standard algorithms.
That is not done because for the sponge construction it is better to design a mixing function with a single wider input instead of a mixing function with 2 narrower inputs, like for Merkle–Damgård. Therefore it is better to design the function from the beginning for being used inside a sponge construction, instead of trying to adapt functions intended for other purposes.
> We encourage the use of this new lightweight cryptography standard wherever resource constraints have hindered the adoption of cryptography
Nothing stops us from qualifying "cryptography" with an adjective to refer to a specific part of it, while still referring to full-fledged cryptography. See the Wikipedia article for examples.
"Elliptic Curve Cryptography"
"Post-quantum cryptography"
"Public-key cryptography"
https://en.wikipedia.org/wiki/Cryptography
Those qualifiers in front of cryptography are very different from lightweight. Lightweight doesn’t convey a good meaning in general and that is the main issue.
I haven’t looked at the lightweight crypto proposals though but if NIST is proposing it I am optimistic. If the proposals from various cryptographers around the world made through all the process it’s pretty good.
This is now besides the original point, which was that in cryptography's case, "lightweight" would work like "alternative" works for "alternative medicine".
However if we shift the focus to marketing, lightweight works nothing like that in that regard. They use it to imply that the product is fast, lean, to the point, contains nothing unnecessary, has fewer, but solid options, energetic, aspiring. Implying that the other products are heavier, bulkier, cumbersome, bloated, slow, overcumbered, overly complex.
It's already used in for example LDAP, which I don't think anyone scoffs at (well, with regards to the lightweight qualifier), and lightweight software has its own wiki page, implying that the usage of the term wrt/ software is widespread enough to be notable. They write that this means "a computer program that is designed to have a small memory footprint, low CPU usage, overall a low usage of system resources".
All this to write that lightweight conveys a good meaning indeed. About the only widespread negative usage I can think of is when people call others "lightweights", implying that they performed below expectations.
I hear you, but they could have chosen performant crypto or low power crypto or some variant like that. For many lightweight reminds them of low effort ciphers. I see your point though
For me if it comes from NIST, I trust it (even with their past shenanigans) because I know the process they use and also they only produce these standards to be used for federal governent software. So they will do a good job.
This is facetious. Some protection is better than no protection.
If "every little bit helps" is true for the environment, it's also true for cryptography, and vice versa.
> Some protection is better than no protection.
No, not really.
Algorithms tend to fall pretty squarely in either the “prevent your sibling from reading your diary” or the “prevent the NSA and Mossad from reading your Internet traffic” camps.
Computers get faster every year, so a cipher with a very narrow safety margin will tend to become completely broken rapidly.
That classification has more steps.
Some things must be encrypted well enough so that even if NSA records them now, even 10 years or 20 years later they will not be able to decipher them.
Other things must be encrypted only well enough so that nobody will be able to decipher them close to real time. If the adversaries decipher them by brute force after a week, the data will become useless by that time.
Lightweight cryptography is for this latter use case.
The hard part is that anything designed to be breakable in a week is only ~150x away in strength from being broken in an hour. That means you need to be incredibly confident about how strong your algorithm is. It's much easier to eat a little bit of cost such that you think it's invulnerable for thousands of years because that way you don't need to worry about a factor of 2 here and there.
Right. Strength in a shadow contest like cryptography can be, best case, estimated to sixteen-bit orders of magnitude (+-65000x). Just because you can't break it doesn't mean somebody else secretly knows a game changing way to break it. So you keep padding with huge exponential hedges such that if they scan shave a dozen bits off the strength of the scheme, it's still secure under finite resources.
Playing close to the margin is super dangerous.
The environment? As in trees?
The IETF has published TLS cipher suites with zero cryptography, "TLS 1.3 Authentication and Integrity-Only Cipher Suites":
* https://datatracker.ietf.org/doc/html/rfc9150
Lightweight cryptography could be a step between the above zero and the 'heavyweight' ciphers like AES.
They chose Ascon which is a good set of same sponge-based cryptographic functions if you don't have hardware acceleration for AES or the CPU resources for chacha20 which is the intention of the standard. The security is 128-bits (comparable to AES-128).
<https://ascon.isec.tugraz.at/specification.html>
i too approve of their choice of cryptographic functions, if not a little miffed they didnt come to me first
This is pretty cool. But IOT tends to fail hard on key agreement. And nothing here solves that. This seems to pretty much require a pre installed key, otherwise the overhead of securely installing a key would probably nullify the advantage of this encryption.
Right, it's a competition to standardize authenticated encryption constructions, not entire cryptosystems.
By the same token AES is useless as well, because it doesn't address key exchange. This was not the goal of this standardization process.
My point was that AES and SHA are not the reason IOT cryptography is so often broken or missing. Instead its getting the keys onto the system in a halfway secure manner that is the blocking issue.
Hence I'd be a lot more enthusiastic about NIST guidance on these points.
A pairing system as seen in e.g. Zigbee or BLE seems pretty good to me. Not everyone cares to implement it well and there's still no standard for web-based devices, but it's here and it works.
I'd like to see more devices able to pair with NFC, but even that's standardised for Bluetooth, just underused.
Ah, I see. That's indeed an interesting point. At any rate, IOT cryptography can use this standard as a building point, so it is a step in the right direction anyway.
The world that the algorithm targets is exactly where is this doable. MCUs typically have a protected OTP area that makes it a good place to inject keys right in the factory.
Are these for garage doors and doorbells? Those devices could definitely use more security (it's not hard to stuff a proper TLS stack in a microcontroller but manufacturers balk at even putting something as cheap as a ESP32 in their BOM).
Also for currently-constrained platforms for which hardware updates are infeasible, software updates aren't, and cryptographic security is currently compromised due to lack of availability of appropriate constructions.
You'd change the coin cell in an ESP32 powered garage door fob monthly... Unacceptable to most.
Any remote control's microcontroller would be in deep sleep most of the time, and computationally expensive key exchange would be rare.
I have a few of these ESP-8266 remotes, and battery life is fine. https://github.com/mcer12/Hugo-ESP8266
I wonder whether this is backdoored by NSA as well
There is a lot of cryptanalysis on it: https://eprint.iacr.org/search?q=ASCON Furthermore, it is not designed by NIST.
Yup, NIST has no business putting their names on anything cryptographic anymore.
Did I miss something recent? Or are we going back to the DES days?
While not exactly recent, I think most people have Dual_EC_DRBG in mind these days when it comes to NIST and backdoors
https://en.wikipedia.org/wiki/Dual_EC_DRBG
they really are shits arent they
Wikipedia says Ascon has 320 bits of state and uses 5 bit s-boxes. That’s tiny compared to sha-256 or Blake2. One would think a pre image attack would be much more tractable at that scale.
https://en.m.wikipedia.org/wiki/Ascon_(cipher)
No? SHA-256 has only eight 32-bit words of state which is even less than Ascon. BLAKE2s looks the same.
https://en.wikipedia.org/wiki/SHA-2
https://datatracker.ietf.org/doc/html/rfc7693
This says very little about the strength of the cipher.
The initial state of ChaCha20 also has only at most 320 unknown bits (512 bits - 128 constant bits - 64 bits of a nonce). Actually you normally also know the counter, so there are only 256 unknown bits.
Of course the actual strength of the cipher cannot exceed the size of the state, but the design strength must be much lower for this cipher. It competes with AES-128, which is designed for an 128-bit strength.
320 bits of state is more than enough for a cipher that must have an 128-bit strength, or even for a cipher designed for a 256-bit strength, like AES-256 or ChaCha20.
Sponges divide state into rate (r) and capacity (c). They "absorb" incoming bytes, perform permutation (moving bits around without losing any of them) on the whole state, and "squeeze" out bytes from the rate, while capacity remains hidden.
For the secure hash function, the capacity should be at least twice the target, that is for 128-bit security you need 256 bits of capacity. ASCON hash uses 256 bit capacity and 320-256 = 64 bit rate, so to get a 32-byte hash of a 8-byte string (without padding), you'll need to do at least 4 permutations.
If you can design a secure permutation that permutes 257 bits, you can make a secure, but impractical hash function from it by setting the rate to 1 bit.
For the duplex mode that's used for authenticated encryption, capacity can be lower, because it's keyed -- it's 192 bits in ASCON.
This assumes the permutation of the 320-byte state itself is secure, of course.
If these primitives are less resource intensive than what we use today with the same level of security, then why don't we just use these everywhere? If they are not as secure, then why would be use these anywhere?
Your front door isn't a hardened steel vault door... I'm assuming you still lock it.
This is a bad analogy, even it it's apt. Only in that an encryption scheme can be technically weaker than another, but still plenty secure. As long as it does the job and hasn't been broken.
For example, part of the spec is simply hashing and packet signing... not necessarily making the packets secret, but authenticating the origin/source. It isn't necessarily about creating the most secure vault in the world, but securing what might otherwise be completely insecure channels of communications/attack. It's also not about a $1000 smart phone, but a micro controller that's a fraction of a dollar on devices that have a total BoM that is very low cost or small.
Why not use chacha20 or xtea for embedded devices? They are lighter than AES..
While I know people where a little sceptical of it. I honestly liked the speck cipher that was published.
This cipher is a lot more heavy.
A land shanty is just called a song. It's protective cryptography or not. Binary.
"Lightweight" here doesn't stop it from proper cryptography. In this case, the qualifier doesn't take away from the original meaning, just focuses it on a subset. Like "public-key cryptography".
Really glad a sponge function won, they are a big step forward in terms of crypto engineering!
s/NIST/NSA/g
[dead]