DSS 2026 — Out of scope, out of luck: the components audits keep ignoring Jan Kalivoda Source video: https://www.youtube.com/watch?v=uYXJQmCagt4 Automatic captions — not human-reviewed. Complete spoken-audio coverage is unverified. Source language tag: en (English). Source export SHA256: af63ec8c48dcded3e6625086761f367596907c3cdd8f01513a89a4cf23ea1bbd [0:00] Okay. Hello everyone. [0:03] Uh, we will finish this session by [0:06] a little bit thinking about scope and [0:08] how to choose it. [0:10] I'm Jan [0:12] and [0:13] I'm co-founder of Egg Free and security [0:16] researcher there. [0:18] And if you think about Web Three typical [0:22] audit, what comes firstly in mind, it's [0:24] probably list of Solidity files where [0:28] there is a smart contract repository, [0:30] you got it sent from the client, you [0:32] estimate it, and you audit contracts and [0:34] right, that's it. [0:35] But [0:37] usually it's much bigger. [0:39] So [0:41] a lot of bugs are originating elsewhere [0:44] or can have some [0:46] complex behaviors between them. [0:49] So [0:50] here is the call out that the scope [0:52] should be much bigger. [0:55] We were doing a research which is [0:57] publicly available. You can also catch [0:59] me after the talk if you would like to [1:01] show it. [1:02] And it's basically some data set of [1:05] exploits from January to September. [1:08] And you can see that contract bugs have [1:10] the highest incidences. So it is [1:13] correct that we are focusing on them and [1:15] we are auditing them, but the value is [1:18] also leaking elsewhere. [1:20] As you can see, [1:22] for example, backend compromise and [1:23] other domains. [1:26] Moreover, [1:28] 71% of these exploits were out of scope [1:32] while 23% were in scope, which is not a [1:36] very low number. But if you [1:39] look closer at it, [1:41] you will figure out that only 8% of it [1:44] are actually recent audits. [1:46] So this is basically showing that audits [1:49] are kind of useful, but they have also [1:51] expiry. [1:54] So I want to highlight here that audit [1:57] scope is a security decision and just a [1:59] formality. [2:01] So [2:02] the scope should be defined by our [2:04] system, by our protocol, your [2:05] application, whatever you have, and you [2:07] shall follow the value there, map all [2:09] your dependencies, build a trust and [2:12] threat model. [2:13] If you are not familiar how to do it, I [2:15] think definitely there will be several [2:17] humans and agents which would like to [2:18] help you. [2:21] And just from our experience, to give [2:22] you some example what it can be, [2:25] like you are auditing just a Solidity [2:27] files, but it costs you almost zero [2:30] time [2:31] to just run a Git leaks. And you can [2:34] find some critical private key exposure [2:38] or anything hidden in some of the [2:40] commits, which is exactly why Tyco get [2:44] hacked if I am correct [2:46] in past. [2:47] Then there is a lot of other review [2:49] areas such as obviously dependencies, [2:52] some malicious version which is pinned [2:54] or [2:55] unfortunately loaded into some [2:58] deployment. [2:59] Then if you are reviewing also [3:00] deployment scripts, you can find some [3:02] inconsistencies between actual smart [3:04] contract logic and the intentions from [3:06] the customer. It can also help you. And [3:08] then there is a lot of integrations or [3:10] off-chain services where sometimes you [3:13] are offloading some validation [3:15] to backend and not in the smart [3:17] contracts, then definitely you need to [3:19] also look at the backend. [3:21] And that's basically it, so we can go to [3:23] lunch.