DSS 2026 / Day 1 / Pearl Hall
Why Crowdsourced Security Is Dying (Code4rena, Hats Finance, and What Comes Next)
Dmytro Matviiv
Source: YouTube automatic caption track · Exported 2026-11-16. Timestamps link to the source video. Cue end times were not supplied.
Transcript
0:00Yeah, let's let's do it. I'm Dmytro. I'm
0:02from Ukraine, and actually I hold my
0:04life in cybersecurity, as everyone is
0:06here. And today I would like to talk
0:08about especially crowd source security
0:10since actually you do crowd source for
0:12the last nine years.
0:14So, [snorts]
0:16yeah.
0:17Today I would like to focus only on a
0:19stuff that related to contest and crowd
0:23source security, yeah? So, if you
0:25actually remember, we had a platforms
0:28before. It was Hacker One, it was
0:29Bugcrowd, and etc. Then actually new
0:31platform is appear on the market as
0:33Immunify, HackenProof, and etc. And best
0:37years for everyone probably was 2021st
0:40where we actually had new models on the
0:43market which is contest, yes? And then
0:46we have a new reality in 2024 and 2026.
0:50So,
0:51here
0:52we should find the common name, yeah?
0:55Crowd source audits, contest, hackathon,
0:58wherever everyone had trying to create
1:01own name for this space, but the idea is
1:05is one. Many researchers, one codebase,
1:08and one goal, find vulnerabilities, and
1:10of course earn money, but not on the
1:12slide. And so I call this chapter is
1:15beginning, beginning of everything.
1:18So, what's the problem we trying to
1:20solve this time? So, first of all, there
1:23was not enough
1:25security researchers on the market, yes?
1:27New smart contract languages and etc.
1:29appear, and everyone wants to know that
1:32everything is perfect.
1:34Cyber security firms were were fighting
1:36for security specialist, and everyone
1:38was saying, hell, no, we actually know
1:40how to do move audits, rust audits, and
1:43etc. But at the end everyone was
1:46find out that this is was the same
1:48specialist cross-sharing through each
1:51auditors. Also, at the same time, there
1:54was lots of security good engineers who
1:56had a full-time job as a developer who
1:59actually built a code base and they
2:02like,
2:03hmm, why I cannot just doing some
2:05part-time job legally? They asked it
2:08actually the HR to make some
2:12extra legal agreement to participate in
2:15etc. So market was growing more and more
2:17and everyone wants to have transparency
2:19because like if some languages appear
2:21have to learn bugs, have to actually be
2:24faster than bad actors.
2:26So what's happening then?
2:28Best times when actually Code for Arena
2:31was announcing about first contest for
2:33more than $100,000,
2:35everyone was talking about this that
2:37hey, we are going to earn money, not
2:39much specialist and at the beginning it
2:41was a few guys who participated and they
2:43earned good money.
2:45So if you actually remember it's public,
2:47all data is public, yes, so nothing is
2:50to hide here so you can make a photo and
2:51etc. So with those year actually was
2:55documented more than 160 crowdsource
2:58audits on a market
3:00and more than nine
3:02million plus actually rewarded bounties.
3:05And there was more than 200 unique with
3:08critical vulnerabilities. Everyone
3:10actually could learn. So many good
3:12actually initiatives was around. I
3:14remember even Pashov groups
3:17I met guy at the conference and then
3:20they collected together good talent is
3:22and created a new firm security. So lots
3:26of talented actually created on those
3:28time. And the business model was pretty
3:30nice, yes. So
3:32companies so we talk about Code for
3:34Arena for example, they earn around
3:3620-30% from each actually contest. So by
3:39having contest for 100,000 you could
3:42earn actually as a business
3:44to 20-30,000 dollars actually
3:47and work. So if you remember there was
3:50an article, unfortunately my designer
3:52didn't put it, but actually there was
3:54article there where Code Arena was
3:56promoting a guy who actually earned more
3:58than 1 million just for 40 months at the
4:02Code Arena for
4:04participating in the contest.
4:06And then actually became a boom. So when
4:09something is appear, everyone trying to
4:11step in as a business, work, find
4:13models, and etc. And this is what was
4:16happening. So
4:17Spearbit, which is Cantina, they
4:19actually raised money. There was a
4:21public announced that they will going to
4:23crowdsource security market and improve
4:25this market. There also actually
4:28was opportunity for for this because
4:30everyone was knowing that Code Arena was
4:32charging 20 30% for
4:36listing companies and they could
4:37actually build strategy to get
4:40investments into it. And then actually,
4:43if you remember, there was a boom of
4:46millions contest. So it was actually
4:49make a DAO on the Sherlock. There was
4:50actually Blast, Uniswap,
4:53Arl, etc. So there was like a boom of 1
4:56millions
4:57platforms and even companies who wants
5:00to have a audit, they were fighting to
5:01be listed at the platforms and actually
5:05show that they have money to make a
5:07contest. And everyone was thinking that
5:10contest is cheaper than doing internal
5:12audits, but the reality always was and
5:16and is that contest is always expensive
5:19than actually doing private audits.
5:22And actually, so everyone wanted to be
5:24in, yeah? New talent, learning boom.
5:27Everyone was just making some courses,
5:30training actually pipelines, and etc.
5:33And then actually, if you remember,
5:35there was a Cantina elite security club.
5:38They announced it that if someone is
5:40joining, they cannot participate in
5:42other publicly contest and etc. Everyone
5:46was trying to step in into this contest.
5:49And then was a turning point which is
5:52actually happening. So
5:54crowd become too crowded and more
5:57researchers, same price, submissions
5:59overloaded, early AI entering to the
6:02games, more duplicates. So, everyone
6:05trying to rethink um
6:08technology. So, if you actually take a
6:11look, so previously there was around 500
6:14that price pool for submissions. So, in
6:172026 there was less than five times
6:21public contest and price for per
6:23submission was just $80.
6:26So, there was a perfect storm, yes? So,
6:29crypto winter, platform competitions,
6:31hungry researchers, um
6:34AI wasn't the main boom yet. So,
6:36actually
6:38what's happening the next? I will take
6:40one minute extra, okay? So, zero
6:43yes. So, California actually release
6:46zero fee for the companies. So, come and
6:49not paying. But the reality, if you
6:51actually company and you have register
6:54your entity, let's say in US or
6:56somewhere, you also have to pay money
6:58salary. I'm talking as a CEO for you,
7:00yeah? So, to keep company at least for I
7:03don't know, 10 people in US is
7:06more than $100 per month. So, actually
7:10what's happening? So, companies,
7:12platforms start
7:13competing with each other. Everyone
7:15trying to show zero
7:17bug fee model,
7:19conditionals pools. If you remember,
7:21let's say someone was saying, hey, you
7:23have to find 10 critical
7:27So, that was the reality. Then severity
7:29based when like previously the main idea
7:31was there is a pool and you take pool
7:33even you have low, medium bug. Then
7:35actually critical, high, low and
7:37everyone has to participate. So, what's
7:39happening then? Two pioneers die, yeah?
7:44Because of business model first of all.
7:46Second, it's my last part, cows. Cows
7:49actually was related to what? Everyone
7:52trying to do with AI, everyone trying to
7:54participate in new models appear on the
7:56market, pay submission and etc. And
7:58where we move, last slide from my side.
8:01So, we move to
8:03from, hey, we need more talents to we
8:06have AI. So, next what we will see is
8:09actually we move to the way when AI
8:13researchers and actually researchers
8:15will burn their AI tokens to save tokens
8:18of companies to actually make AI audits.
8:22Yeah, thank you.
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: 1347ebc36c32a47193fe2e0d139306b7ed7887d8aef49c3ec9826761c2d23b84