DSS 2026 / Day 2 / Pearl Hall
Can security judgement be financialized without corrupting it?
Don Josey
Source: YouTube automatic caption track · Exported 2026-11-16. Timestamps link to the source video. Cue end times were not supplied.
Transcript
0:00Hey, I'm Dawn with Glosify and today I'm
0:03here to talk about
0:04what we built and specifically whether
0:07security judgment could be financialized
0:09without corrupting it. Um so
0:13basically
0:14uh at Glosify we built uh a protocol
0:17which is a swarm that allows um it's a
0:21it's a protocol that allows you to
0:22coordinate economically a swarm of
0:24agents that attack security challenges
0:27basically. So what do I mean by that?
0:29We're talking about attacking not
0:31operational security um but code bases
0:35and dependency graphs within kind of our
0:37software ecosystem. Um and uh yeah I
0:41mean uh modern software code bases are
0:43built of these maps of you know
0:47uh all these
0:49dependencies where you know somewhere up
0:51at the top of the stack if you have one
0:53thing one failure it would propagate
0:55downstream you know into a lot of
0:58different services and our argument is
1:00that structurally kind of we underfund
1:03um paying for uh research in a lot of
1:06these you know frameworks like for
1:07example PHP 6 or whatever that just
1:09failed you know um and looking out for
1:12vulnerabilities that might uh surface
1:14kind of in the long tail. So what can we
1:17do to address that? That's the question.
1:19And um so we're uh
1:23basically trying to align incentives
1:24this is especially relevant for DeFi um
1:27which a lot of these apps you know rely
1:29on uh these dependency graphs. So we
1:34built this protocol and the idea is
1:36coordinate continuous swarms that are
1:39looking at code bases and building a
1:41graph of security intelligence. So this
1:44is continuous as opposed to kind of like
1:46the one time uh you know bug bounty or
1:50audit you know paradigms which are super
1:52important but this is intended to kind
1:54of complement those tasks basically and
1:57bring humans and agents to the of
1:59attacking, you know, security.
2:02Um, and how do we do that specifically?
2:04It's a ranking process. I describe it as
2:06like a cousin of a prediction market
2:08basically, where we take in data that is
2:10like an oracle on failures of open
2:13source code that surface up. And then
2:16you in the ecosystem as a researcher
2:19would have an account and do staking and
2:22get, you know, rewards if you do a good
2:24job of predicting these failures
2:26basically. So
2:28um and uh yeah. And your kind of key
2:32account unit in the system is a vault
2:34and EIP 4626 style vaults, which um as
2:40uh for those not familiar, these allow
2:42you to uh basically
2:45have users who would back your work. Um,
2:48so you can attract liquidity into this
2:50vault uh as a researcher and kind of
2:52attract attention if you do a good job
2:54of doing the research. And then uh your
2:57funds are staked in the vault in such a
2:59way where your rewards payout based on
3:02your performance, you know, multiplied
3:04by your staking. So if you do a good job
3:06finding vulnerabilities,
3:08uh that amplifies your reward payout
3:10basically. And this creates like a
3:11flywheel of intelligence is the idea
3:13where researchers who do a good job
3:15attract more capital and they get paid
3:17more because, you know,
3:19they are successfully finding
3:20vulnerabilities. So um and um yeah. We
3:24kind of smooth the rewards out uh so
3:26they're not coming in kind of uh
3:29in chunks. They basically like buffer
3:30in, which kind of is supposed to fight
3:33like basically people gaming the system
3:35and kind of just like jumping in ahead
3:37of a reward. Also the system's based on
3:39uh staking but not slashing with the
3:41idea being that that creates a mechanism
3:44where uh basically
3:46people feel more comfortable backing
3:48these pools because it's not going to be
3:49like some adversarial scam where the
3:51pool runner is going to attack them or
3:53something. Um, we also built a basically
3:55a cryptographic bounty for vulnerability
3:58disclosure, which you can attach to your
4:00vault basically a signature of a
4:02vulnerability, which goes against the
4:04timestamp, and later, when the
4:05vulnerability's found, be awarded
4:07retroactively for the payout. Um,
4:11so the whole idea here is to kind of
4:13harness collaborative and cooperative
4:15energies in one place for vulnerability
4:17disclosure in a way that's dynamic,
4:19continuous, where people can get paid
4:21for work and build agents and, you know,
4:23basically find security work through
4:26attention and reputation they get in the
4:27system. There's a token, you know,
4:29that's locked in the contract, so
4:30everyone is collaborating and locking
4:32that up to secure the network while also
4:35competing by finding kind of different
4:37layers to build this data stream. Um,
4:41and yeah, once the data is produced, uh,
4:44you know, we can build future mechanisms
4:47for insurance and basically route this
4:50data back in like a feedback loop to
4:52users within a software supply chain so
4:54they can secure their own systems. So
4:57this is what we're doing. We've built
4:58this. This is live on Sepolia now. Check
5:00the app out. Um, break it, please, uh,
5:04and um, sign up for our mailing list and
5:07hear more. We're trying to launch in the
5:09next few months and yeah, get our audit
5:11launch, everything done. And yeah, so
5:13that's it. Glossify. Thanks.
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: 84098c983d11cdd501f32afc2fa76aa4561ad3fac76260e032c6e0852df8fb15