DSS 2026 / Day 2 / Pearl Hall
How to Kill Your Own Chain: A Security Playbook for Live Cosmos→EVM Migrations
Ofir Eliasi
Source: YouTube automatic caption track · Exported 2026-11-16. Timestamps link to the source video. Cue end times were not supplied.
Transcript
0:00Um, so, hello everybody.
0:03Um,
0:04I'm going to talk about like a topic
0:06that is um for me very interesting. We
0:09uh I got involved into uh
0:12some scenario that is um
0:15um super complex, but I see it happens
0:17more and more recently and expecting to
0:21have it more on the future. So I think
0:23it's like super interesting and
0:24something that was having
0:26specifically from the security aspects
0:28um
0:29even more.
0:30Um,
0:32first like short introduction about
0:33myself and my background like over 20
0:36years of software development, two times
0:39founder,
0:40um been involved into blockchain
0:42full-time from 2017,
0:44many projects, most of them in
0:47related to infra, MEV, RPC, security
0:52um and so on, research that relates to
0:55blockchain as well.
0:56Um, and that's also what brings me to
0:59this
1:00specific case. Now the CTO of Lava
1:03Network, an RPC provider that is
1:06trustless than decentralized, uh
1:09currently run on running on Cosmos.
1:12And um
1:13we got into the conclusion that um
1:17this chain or this network actually need
1:19to move from uh Cosmos based network
1:23into an EVM
1:25protocol, smart contract of course.
1:28Um, so
1:31first like very shortly about the the
1:33background or the the reason of moving.
1:36As you know, Cosmos based and generally
1:37up chains involve with like many aspects
1:40super complex
1:41and sometimes even expensive. The idea
1:44of of shifting is to cut the the
1:47security layer
1:48to get many of the elements that you can
1:51handle without them
1:52uh out of your scope and in that case
1:55save a lot of
1:57time and effort and resources.
1:59So and obviously we're going to talk
2:01about that uh
2:02like broadly.
2:04Uh I have another talk in two days,
2:07three days actually, um
2:09in Devcon in the main event about why uh
2:12an upchain is not necessary or when
2:15upchain is necessary. In our case it's
2:17not.
2:18Uh but for now we're definitely going to
2:19focus on the security aspect of moving a
2:22chain or a network
2:24uh
2:25into an EVM smart contract protocol
2:28base.
2:29So first
2:30it's huge event. It's a
2:33it's a
2:34it's an earthquake to a to a protocol
2:37that involved with many many element.
2:39Um when you move something out out of an
2:42upchain, it means you take everything
2:46with you. And if it's like different
2:47ecosystem like Cosmos and and EVM,
2:50that means
2:52any of those get a different
2:54interpretation.
2:55You like it's your token and if you have
2:57any mechanism of staking and vesting and
3:01roles, identities
3:03and so on, they all move with you and
3:07should move on the same time.
3:09So and that's involved as I said like
3:11with um a couple of aspects of how you
3:14do deal with that because it's unrelated
3:17uh ecosystems. They have no connection
3:19with them between them. Um of course a
3:22bridge is definitely not not considered
3:25as connected. It can only impact for for
3:28like not not positively on on such case.
3:32But it's also means that you need
3:35other people collaboration. You have a
3:36community, you have token holders, you
3:38have stakers, you have other
3:40participants in the network and you need
3:42them also to be involved with this
3:44migration.
3:46And they
3:47not obey your rules. They sometimes take
3:49their time and sometimes they not exist
3:51or gone. Um
3:52and that's also something that you need
3:54to take into consideration.
3:56Besides that other elements very
3:59very intuitive like identity, a wallet
4:02on Cosmos base uh network is completely
4:05different than EVM.
4:06And that's a challenge that you need to
4:08handle.
4:09So, overall all of this means that you
4:11need to get
4:13everything considered in advance.
4:15Get your um steps planned and make sure
4:19that
4:20the elements that relate to you you're
4:22handling them. What whatever is not or
4:24depends on other, you know how to handle
4:27it properly and to take into
4:28consideration any scenario because even
4:32when what is uh considered is like
4:34unreasonable, like people giving up
4:36their funds,
4:37it's obviously available on such case.
4:42Um so let's talk about like the exact
4:44plan. This is again this is my
4:46interpretation.
4:47Um I split it into all the steps that
4:50can be involved, but definitely you can
4:53merge some step or do them together. You
4:56can um
4:57give up some of them or have something
5:00that be different, but
5:01the common sense or what mostly
5:04available and others also did having
5:07such a scenario is is considered like
5:11it's very similar to uh what I'm going
5:12to show right now. So first, as I said,
5:15identity is one of the the biggest
5:17challenges. You need to map where if you
5:19are if you want to say that like in a
5:20very uh
5:22um naive way, you need to map any wallet
5:25from the app chain Cosmos in our case
5:28into EVM. How to do that? You can go
5:30very straightforward and sign something
5:33on Cosmos that will be um um executed on
5:36chain on the EVM chain that you're
5:38heading.
5:39Um but that means you need to
5:40collaborate you need others
5:42collaboration and you cannot set them a
5:45timeline that will be very strict. That
5:47's why on that case you take some uh
5:49some gap and having this like couple of
5:52days, sometimes week, and sometimes even
5:53a month.
5:55After that we're going to handle each
5:56security aspect of those like later, but
5:58just like briefly to go through that.
6:00After that, of course, um shifting
6:04everything you have
6:05from upchain into EVM. It means taking a
6:09snapshot involved with all the data that
6:11is related the current up-to-date data
6:14and and put it on chain
6:17on the target chain. It's including of
6:20course the token and the token holders,
6:23stakers,
6:25roles, identities as we said everything
6:27just shift
6:29into that chain.
6:31After that once you did that and
6:34deployed on the other chain you need to
6:36actually deploying is actually the
6:38loading.
6:39You need to open the the new chain the
6:40EVM chain up for running and go back to
6:44business
6:45on the EVM chain and move on from there.
6:48One step after is to stop the the old
6:51chain or the EVM upchain the sorry the
6:54upchain. By the way ignoring it is also
6:56an option but you know we cannot ask
6:58validators to stop or you can ask but
7:00they not need to not necessary they're
7:02going to do what you are saying.
7:05Just ignoring all I cancel it completely
7:08and the last step is
7:11the rest of the time we call it forever
7:13when on that
7:16time window things can also happen and
7:19messed up.
7:21So let's go one by one. Registration as
7:23we said not okay not our
7:26control we need collaboration of the
7:28community get them to sign up.
7:30It's something that gonna run for a
7:32while.
7:33Okay
7:35that means you have some kind of a
7:37contract or a way to map
7:40old identity into new identity and
7:44obviously on that area it's a huge risk
7:46that relates to the community. Other
7:49people
7:50not you that gonna do the mapping you
7:53ask them to collaborate and everything
7:56that can be like
7:57misinformation, disinformation, all kind
8:00of confusion is is the is the most
8:04significant significant risk about it
8:06but as well is stealing identity and
8:09thing that can be related to that.
8:12Of course, you should limit that into a
8:14specific time window and make sure
8:17to get that collaboration.
8:19We need to remember that on that case
8:22there is the scenario that you won't get
8:23the collaboration that you're asking
8:25for. What you do in that case? How you
8:26handle that? This is also part of the
8:28scenarios that you need to handle and
8:30have a backdoor
8:32to somehow keep running on the the the
8:36old chain.
8:38But besides this part,
8:41the rest of the parts mostly have no
8:43rollback or no way to go back and that's
8:46also something we need to put on focus.
8:49The snapshot, as I said, taking like a
8:52decent amount of data sometimes can be
8:54like huge, so in some cases you need to
8:57split them
8:59or to execute that like in steps in a
9:02way that you actually frozen the time
9:05like freezing the time
9:07and take it to the new chain.
9:12Of course, like this is a step that you
9:14actually control. You can do immediately
9:16or on the time that you decide to do
9:18that,
9:19but the you need to remember that the
9:22window between between the the snapshot
9:25and the next one that is actually
9:27loading that snapshot into the new chain
9:30is the the time where most of the mess
9:33happening.
9:34You actually have two chains running.
9:37You have
9:41participants or like community members
9:43that operating both
9:45or none
9:46and a lot can happen. You also have a
9:49time where people can
9:52run the the old chain since you don't
9:55want to have like too much of a down
9:56downtime
9:58and still expecting getting rewards on
10:01the new chain. It can be a transfer, it
10:02can be anything like that.
10:04We need to also remember that this on
10:06this time frame we also cutting anything
10:09that can be
10:10um a confusion or a combination of both
10:12like bridges. If we have a bridge from
10:14the
10:15um the old upchain to uh the new EVM
10:17chain that need to be on hold and
10:21actually to be uh
10:23killed in a way that you not get like
10:25double spend.
10:28Um one more element that will let the
10:30security no one actually promise you
10:32that
10:34uh deploying this transaction or
10:36floating the the data will actually
10:38succeed. Um it can be like technical
10:41issues and bugs but can also be someone
10:44that is like specifically trying to
10:46target this transaction and do
10:48everything he can in order to fail it.
10:50Why?
10:51Many reasons but definitely it's doable
10:55not that hard to do that and sometimes
10:57someone to will try to four trying this
10:58transaction and make things in a way
11:00that um can benefit out of it. So this
11:04time frame is mostly uh sensitive in a
11:07way that you need to cover the gap
11:09sometimes afterwards.
11:13Then last step when you open it back to
11:16uh business on the new EVM chain uh we
11:19need to remember one thing.
11:21Um the idea or the motivation is to have
11:24as minimal as downtime
11:27as we can. This protocol is running
11:29someone is is using it
11:31and
11:32in this case
11:34um
11:35like the idea is to decrease the time
11:37between the freeze and opening the new
11:41protocol.
11:42Um again some
11:45will will some protocol decide to still
11:47operate on the old chain until the open
11:50and not stop operating on the freeze and
11:54on that case you need uh again you need
11:55some way to make up on that time to
11:58who's um um decide to still uh execute
12:03and you need to incentive them. It can
12:04be like by the foundation, it can be
12:06manually but it can be part of a format
12:08that later do catch up and get a second
12:12wave of of
12:14of deployment of data that relates to
12:16the old chain.
12:19Here mostly
12:21they dare to get back to normal,
12:23to get the users involved and to have
12:26them
12:28into like the routine. Remember, if you
12:30have staking, for example, and that's
12:33the interest of the
12:35the protocol to keep those staking,
12:38you definitely don't want to get them
12:39out of staking and ask them to restake
12:41because
12:42it's going to be involved with the
12:43churn. People not necessarily all of
12:45them will stake again. So the best idea
12:49on this case is to keep all to copy
12:53every element or every aspect of it
12:55into
12:57the new chain. By the way, in in Lava's
12:59case,
13:00what happened is that we besides
13:03staking, we also have vesting.
13:05And on the Cosmos based network, we
13:07allow to who has token that are vested
13:11to stake and to participate in the
13:14in the protocol process.
13:16So in that case,
13:19it's the idea is to take everything that
13:21relates to that and put it on chain and
13:23keep and preserve the state of staking
13:25and participation in the the the process
13:29of the protocol.
13:30Last one,
13:32of course, stopping is not mandatory or
13:34should be defined according to the
13:38the involvement of
13:41the community, but definitely ignoring
13:43is also an option. We just need to
13:46define it like officially that this
13:48chain is irrelevant anymore. On that
13:50case,
13:51many can has funds lost, but you can
13:54handle that technically by stopping or
13:56having a new version that will stop any
13:59functionality that relates to the chain.
14:01But again, that's not not like the
14:04biggest challenge we have.
14:06Last one,
14:08since then, after we open and stop the
14:10the old chain running forever, it seems
14:13like what the what issue can
14:15uh what we can have on
14:17uh like the day-to-day ongoing on the
14:19new chain. Well, remember that not
14:22necessary everybody claimed their
14:23identity, not everybody claimed their
14:26funds. And that's still
14:29um something a scenario that we need to
14:32um consider because
14:34um
14:35eventually someone can get this funds or
14:39try to hack them and get them um
14:42like instead of those that actually own
14:45them.
14:46Um
14:47yes um so we kind of ran out of time. So
14:52I think like the main message from this
14:55um
14:56talk is
14:58definitely
14:59um um the awareness of having such steps
15:04and the the ability to split to split
15:07the the process into time frames and and
15:10time windows
15:11and to handle each one of them
15:13separately.
15:14Um
15:15um so the idea is like of course the the
15:18to kill the chain but not kill the
15:20users, make them stay and keep the the
15:22protocol running on the new chain.
15:26Questions?
15:29Good, we ran out of time.
15:30Um so about that my my schedule is like
15:34definitely I have one more this today
15:37later, but as I said I'm going to talk
15:39about this topic like more about the
15:41reasons and how to handle it technically
15:43later this week.
15:45Thank you everybody.
Source and accuracy details
- 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.
- Automatic YouTube captions; not human-verified against the audio.
Source export SHA256: 598694e23a435766dd4ebd0a73a03dfbb110ab5ed1986ad439138b70883c27f9