DSS 2026 / Day 2 / Pearl Hall
Quantum-Safe Is Not Just Post-Quantum Signatures
Abhinav Sharma
Source: YouTube automatic caption track · Exported 2026-11-16. Timestamps link to the source video. Cue end times were not supplied.
Transcript
0:00Perfect.
0:01So,
0:02hello everyone. My name is Abhinav.
0:04I'll be talking something other than AI.
0:08Of course, why not?
0:10I'll be talking about how quantum safe
0:12is not just about post quantum
0:14signatures
0:15and we have five more layers including
0:19that that we need to take care of.
0:21So,
0:23the general reflex around
0:25securing ourself from
0:28post quantum world.
0:30We only take care
0:33of one failure mode which is just that
0:35will swap the signatures with quantum
0:37proof signatures.
0:38But
0:39I want to make sure that there are four
0:41more layers that we could get our
0:43awareness on and what are they and what
0:45we need to take care of.
0:48The modern cryptography relies on one
0:51assumption which is that certain
0:53mathematical problems
0:55are very hard to solve
0:58so that even if we have enormous
1:00computing power
1:01the goal to get the right answer is
1:04practically infeasible.
1:05And
1:06quantum computing challenges this
1:08assumption by saying hey
1:10what if we change the architecture of
1:12the computation itself? So, instead of
1:14representing our information in the bits
1:16of zeros and one what we'll do is we'll
1:19use something called as qubits and this
1:21qubit represents
1:22a simultaneous state
1:25by using a principle called as
1:27superposition. I won't go into the
1:28details due to the lack of time, but
1:31current estimate says that we need
1:32around 1200 logical qubits and around
1:35less than 9 minutes to crack
1:38one private key
1:39and less than 9 days to crack top 1000
1:43Ethereum accounts.
1:46Basically decimation.
1:49There are five layers that we need to
1:50take care of.
1:52The first one is ownership that we all
1:53know which is what?
1:55Cracking your private key. So, each one
1:57of us must have a wallet address that
1:59wallet address is nothing but a public
2:02key.
2:03And quantum computers can break your
2:05private key by using that public key. So
2:07when you're transacting, you're exposing
2:09your public key.
2:11When I'm talking about ownership, I'm
2:12not just talking about these signatures,
2:14but also the admin accounts that are
2:17being used to transact multi-billion
2:20cryptos from one point to another.
2:23And so the fix here will be to focus
2:25more on account abstraction that makes
2:27ownership much more flexible and key
2:30rotation followed by quantum proof
2:32signatures.
2:33The second one will be the code base. So
2:35in order to stay under the gas limit,
2:38smart contracts
2:40give their heavy lifting to something
2:42called as Nates pre-compiles, which is
2:44EC recover, we have BLS 12 pairing and
2:47many more.
2:48These are also dependent on the elliptic
2:51curve cryptography, which means that if
2:53I can break them
2:55or an attacker who has a quantum
2:56computer can break them,
2:58the impact can become that I can just
3:00forge a proof on a rollup, an invalid
3:03proof.
3:04So the fix here will be to use
3:05post-quantum pre-compiles and slowly
3:08move forward to the lattice-based
3:09designs.
3:11Third one, which is the consensus. So
3:13validators are using the BLS key
3:16which are
3:18curve friendly
3:19and
3:20if an attacker can sort of get their
3:23hands on cracking the validator key,
3:26they can forge attestations at a large
3:29scale. We need
3:30two-thirds of the majority and believe
3:32me, this is not about stealing funds. It
3:35is about the network's ability to agree
3:37on what is true.
3:39The fix here will be to harden
3:41the validator key rotation and
3:43post-quantum signature aggregation as it
3:45matures.
3:46Data availability. So
3:48rollups
3:51they upload, for lack of better words,
3:53their data
3:55and these validators do not just extract
3:58it and verify it. They are dependent on
4:00something called as KGZ commitments.
4:02These commitments are essentially the
4:04proof that what is being transported is
4:07actually what we want.
4:10The
4:11quantum attacker can basically forge
4:14these since KZG commitment also uses
4:17elliptic curve cryptography.
4:19If this goes
4:21of the if this breaks, that means I can
4:24forge the proofs that are invalid in
4:25nature
4:26and that means the entire roll up
4:28infrastructure needs to
4:30be quantum safe.
4:33The fifth and I think the
4:36most underestimated one is the privacy.
4:39So all of these protocols that
4:41essentially say that they are
4:42anonymizing the user
4:44it breaks into two parts. First is the
4:46key exchange that is responsible for
4:49sharing the secret between two users. If
4:51this goes down, that means we can
4:54deanonymize a user followed by checking
4:57all of this his old transactions as
4:59well.
5:00The second is the same called as
5:02pedersen commitments which is
5:04essentially
5:06hiding the amount of value that is being
5:08transported.
5:10If that goes down, which means it is not
5:12just about the value that I can get from
5:14this, but I can also force the invalid
5:17proofs which is essentially
5:18counterfeiting
5:20minting money out of thin air.
5:22So
5:24in order to be
5:25in order for us to say that we are
5:27quantum safe
5:29there were five layers that we need to
5:30take care of and believe these are not
5:33just
5:34race against a particular transaction as
5:36some of the things like the KZG
5:38commitments need a trusted setup and one
5:41offline compute can crack that
5:43which means it is not just about race
5:44against time, but we need to have a five
5:47layer checklist as a security researcher
5:50that could focus on auditing the stack
5:53and focus more on the
5:56thinking statement that this is not not
5:58just about
5:59a checkbox but a checklist.
6:02This marks the end of my talk. I hope
6:04you have a great time.
6:05Thank you.
Source and accuracy details
- Auto-generated YouTube captions; native source language tag preserved; not human-verified against audio.
- Names, organizations, technical terms, code and numbers may be misrecognized.
- Source wording and disfluencies are preserved; no editorial corrections were applied.
- Export contains whole-second cue starts only; cue ends and subsecond timing are unavailable.
- A full caption-track export does not establish that every spoken word is represented.
Source export SHA256: 0cd7f453dc15418cf0e8c9928003efb65cb4fa8c588118d2574a9bb5e59d1288