DSS 2026 / Day 1 / Pearl Hall
Out of scope, out of luck: the components audits keep ignoring
Jan Kalivoda
Source: YouTube automatic caption track · Exported 2026-11-16. Timestamps link to the source video. Cue end times were not supplied.
Transcript
0:00Okay. Hello everyone.
0:03Uh, we will finish this session by
0:06a little bit thinking about scope and
0:08how to choose it.
0:10I'm Jan
0:12and
0:13I'm co-founder of Egg Free and security
0:16researcher there.
0:18And if you think about Web Three typical
0:22audit, what comes firstly in mind, it's
0:24probably list of Solidity files where
0:28there is a smart contract repository,
0:30you got it sent from the client, you
0:32estimate it, and you audit contracts and
0:34right, that's it.
0:35But
0:37usually it's much bigger.
0:39So
0:41a lot of bugs are originating elsewhere
0:44or can have some
0:46complex behaviors between them.
0:49So
0:50here is the call out that the scope
0:52should be much bigger.
0:55We were doing a research which is
0:57publicly available. You can also catch
0:59me after the talk if you would like to
1:01show it.
1:02And it's basically some data set of
1:05exploits from January to September.
1:08And you can see that contract bugs have
1:10the highest incidences. So it is
1:13correct that we are focusing on them and
1:15we are auditing them, but the value is
1:18also leaking elsewhere.
1:20As you can see,
1:22for example, backend compromise and
1:23other domains.
1:26Moreover,
1:2871% of these exploits were out of scope
1:32while 23% were in scope, which is not a
1:36very low number. But if you
1:39look closer at it,
1:41you will figure out that only 8% of it
1:44are actually recent audits.
1:46So this is basically showing that audits
1:49are kind of useful, but they have also
1:51expiry.
1:54So I want to highlight here that audit
1:57scope is a security decision and just a
1:59formality.
2:01So
2:02the scope should be defined by our
2:04system, by our protocol, your
2:05application, whatever you have, and you
2:07shall follow the value there, map all
2:09your dependencies, build a trust and
2:12threat model.
2:13If you are not familiar how to do it, I
2:15think definitely there will be several
2:17humans and agents which would like to
2:18help you.
2:21And just from our experience, to give
2:22you some example what it can be,
2:25like you are auditing just a Solidity
2:27files, but it costs you almost zero
2:30time
2:31to just run a Git leaks. And you can
2:34find some critical private key exposure
2:38or anything hidden in some of the
2:40commits, which is exactly why Tyco get
2:44hacked if I am correct
2:46in past.
2:47Then there is a lot of other review
2:49areas such as obviously dependencies,
2:52some malicious version which is pinned
2:54or
2:55unfortunately loaded into some
2:58deployment.
2:59Then if you are reviewing also
3:00deployment scripts, you can find some
3:02inconsistencies between actual smart
3:04contract logic and the intentions from
3:06the customer. It can also help you. And
3:08then there is a lot of integrations or
3:10off-chain services where sometimes you
3:13are offloading some validation
3:15to backend and not in the smart
3:17contracts, then definitely you need to
3:19also look at the backend.
3:21And that's basically it, so we can go to
3:23lunch.
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: af63ec8c48dcded3e6625086761f367596907c3cdd8f01513a89a4cf23ea1bbd