Latest / Expert Insights Podcast / #61 – The Role of Ethical Hacking in Proactive Security (Kara Sprague, HackerOne)
Transcript
- 0:06SPEAKER_02: Hello and welcome to the Expert Insights Podcast.
- 0:09SPEAKER_02: I'm Caitlin Harris,
- 0:10SPEAKER_02: Deputy Head of Content at Expert Insights,
- 0:12SPEAKER_02: and your host for this episode,
- 0:13SPEAKER_02: which we're recording live at the RSAC 2025 conference.
- 0:17SPEAKER_02: I'm joined today by Kara Sprague,
- 0:19SPEAKER_02: CEO at HackerOne,
- 0:21to discuss how their community of ethical hackers is helping
- 0:24organizations globally
- 0:25SPEAKER_02: to build long-term resilience,
- 0:27SPEAKER_02: and whether AI is a complement or a disruptor to human testers.
- 0:31SPEAKER_02: Cara, thanks for joining us.
- 0:34So HackerOne is known for bringing ethical hackers and
- 0:37organizations together.
- 0:39Could you walk us through your offensive security platform and
- 0:42how it's helping
- 0:43SPEAKER_02: organizations to identify vulnerabilities before the attackers?
- 0:47SPEAKER_01: Certainly, you're exactly right.
- 0:49HackerOne is a platform and the customers of that platform are
- 0:53businesses,
- 0:54SPEAKER_01: mostly larger enterprises.
- 0:55You know,
- 0:55we have very,
- 0:56very strong coverage of,
- 0:57for example,
- 0:58the Fortune 500.
- 1:00And then the hackers are the community that we connect those
- 1:04businesses with.
- 1:05And we do that through a series of or a set of options around
- 1:09engagements with
- 1:10SPEAKER_01: those researchers.
- 1:12SPEAKER_01: We offer vulnerability disclosure,
- 1:14which is where researchers are invited into a safe space to
- 1:17report vulnerabilities
- 1:18SPEAKER_01: to a company.
- 1:19There's no financial transaction that is involved in those,
- 1:22but rules around like safe harbor
- 1:24exist for the researchers to do that in a way that they are
- 1:27legally protected.
- 1:30SPEAKER_01: We have bug bounty.
- 1:31SPEAKER_01: That's where money actually exchanges place.
- 1:33SPEAKER_01: And our customers would post, for example, a bounty table.
- 1:36SPEAKER_01: I pay out X for a critical vulnerability.
- 1:38SPEAKER_01: I pay out Y for a high.
- 1:40SPEAKER_01: And so researchers can submit.
- 1:41We have both public and private variants of vulnerability
- 1:45disclosure and bounty.
- 1:47SPEAKER_01: We have time-bound engagements.
- 1:49SPEAKER_01: AI red teaming is an example of that, where we are testing the
- 1:53SPEAKER_01: the security, safety, and trust of AI systems and AI models.
- 1:58SPEAKER_01: Then we have pen testing.
- 2:00SPEAKER_01: And then in addition,
- 2:01SPEAKER_01: we have a beta offering called code,
- 2:03which is offering a kind of security testing in line as
- 2:07developers are actually
- 2:08SPEAKER_01: producing code.
- 2:10SPEAKER_02: Absolutely.
- 2:10So it's clear then that offensive security is a lot more than
- 2:14just pen testing,
- 2:14SPEAKER_02: which is kind of the first thing that a lot of people think of.
- 2:17SPEAKER_02: So how do all of these different things that you're doing,
- 2:20all the different components of the platform come together to
- 2:22help build longer
- 2:24SPEAKER_02: term resilience in an organization's defensive strategy?
- 2:28SPEAKER_01: What we see our best customers,
- 2:29our most security sophisticated customers doing is they're
- 2:32using all of these
- 2:33SPEAKER_01: techniques in concert.
- 2:35um so i'll give i'll give an illustration of of uh of one one
- 2:39customer you know of
- 2:40course they have a number of tools in their development
- 2:42pipeline so as the software
- 2:45and other technology is being developed there's a bunch of
- 2:47scanning going on and a
- 2:49bunch of introspection to make sure developers are not putting
- 2:52into into builds
- 2:54code that is not good at the time after build there's generally
- 2:58a homegrown kind of
- 2:59SPEAKER_01: scan
- 3:01SPEAKER_01: that is developed by a purple team,
- 3:02SPEAKER_01: developed and maintained by a purple team that looks for,
- 3:05SPEAKER_01: you know,
- 3:05SPEAKER_01: common issues.
- 3:06SPEAKER_01: Then it goes into pre-release pen testing.
- 3:09SPEAKER_01: And so,
- 3:10SPEAKER_01: you know,
- 3:11SPEAKER_01: our best,
- 3:11most security conscious customers have mandated that all
- 3:14technology that goes out
- 3:15SPEAKER_01: the door has to go through some sort of pen testing before it goes out.
- 3:19And then once it's out the door,
- 3:20then there's a set of adversarial engagements.
- 3:24So oftentimes customers will have coverage of both a VDP as
- 3:28well as bug bounty on
- 3:29SPEAKER_01: specifically high value scope.
- 3:32SPEAKER_01: And they will drive research or engagement to look at new things through,
- 3:36SPEAKER_01: you know,
- 3:36SPEAKER_01: various tactics such as short time bound engagements or
- 3:40SPEAKER_01: campaigns around increasing bounties for a time.
- 3:43SPEAKER_01: Absolutely.
- 3:44So another sort of big topic we're hearing a lot about at the
- 3:47show this year is AI,
- 3:49of course.
- 3:49SPEAKER_02: Yes.
- 3:50And attackers are increasingly using AI to sort of accelerate
- 3:54reconnaissance,
- 3:55develop more sophisticated attacks like polymorphic malware and
- 3:59that kind of thing.
- 4:00How has this rise of AI changed the nature of the
- 4:04vulnerabilities that you're
- 4:05SPEAKER_02: uncovering at Hakawa?
- 4:07SPEAKER_01: So it hasn't yet significantly changed the nature of the vulnerabilities.
- 4:11SPEAKER_01: It's changed the nature of the attacks.
- 4:13So you have now a lot more attackers that are equipped with
- 4:17very sophisticated
- 4:18SPEAKER_01: techniques because AI enables them to scale out much more.
- 4:21SPEAKER_01: And increasingly,
- 4:21we're seeing AI that is approaching levels of some human
- 4:26security researchers and
- 4:27SPEAKER_01: being able to find things.
- 4:28SPEAKER_01: There's hackbots out there,
- 4:29SPEAKER_01: 100% automated things that can go and apply in a non-deterministic way,
- 4:33SPEAKER_01: techniques that will uncover vulnerabilities in code bases.
- 4:36And in fact,
- 4:37we're seeing them start to rise in the leaderboards at
- 4:39HackerOne.
- 4:42SPEAKER_01: In the space of changing the nature of vulnerabilities,
- 4:45AI systems and AI models are an entire new frontier of
- 4:49technology that's going to
- 4:50SPEAKER_01: be required that we test and we need to make more resilient.
- 4:55And so as we see more and more AI adopted into technology and
- 4:59applications,
- 5:00SPEAKER_01: for example,
- 5:01SPEAKER_01: I've seen some projections that in three years,
- 5:02SPEAKER_01: 80% of all applications will have some sort of AI capability behind them.
- 5:08SPEAKER_01: That is a new approach.
- 5:09SPEAKER_01: area of vulnerabilities that is going to have to be looked for.
- 5:13SPEAKER_01: So you can reference the OWAPS top 10 for LLMs,
- 5:16SPEAKER_01: for example,
- 5:18which includes new classes of vulnerabilities that are specific
- 5:20to AI systems.
- 5:22SPEAKER_02: Absolutely.
- 5:23So how are you at HackerOne adapting or how are you planning to
- 5:26adapt your testing
- 5:27SPEAKER_02: capabilities to deal with these AI-enabled threats?
- 5:30SPEAKER_01: So we're looking at this holistically.
- 5:33SPEAKER_01: So on the one hand, there's security for AI.
- 5:36SPEAKER_01: And so already we have in place with a number of our customers,
- 5:40SPEAKER_01: AI red teaming engagements,
- 5:42SPEAKER_01: where both for model providers,
- 5:44as well as for customers that are taking third party models and
- 5:47incorporating them
- 5:47SPEAKER_01: into their offerings.
- 5:49SPEAKER_01: We are providing the security, safety and trust testing for those.
- 5:53And what we found is that the security researcher community is
- 5:57very,
- 5:57very effective.
- 5:58The same creativity and ingenuity that makes them effective in
- 6:01finding security
- 6:02SPEAKER_01: vulnerabilities makes them very effective in testing AI systems.
- 6:06So as an example,
- 6:07Anthropic,
- 6:08we are a major partner with Anthropic for testing their models.
- 6:12SPEAKER_01: Then there's the bucket of AI for security.
- 6:17And so in that space,
- 6:20we have already in our platform,
- 6:22our agentic co-pilot called Hi.
- 6:26And what that is doing is that is streamlining workflows and
- 6:29making it much,
- 6:30much easier for our
- 6:32SPEAKER_01: users to get sophisticated things done.
- 6:35SPEAKER_01: For example,
- 6:35generating report on mitigation items and making sure that they
- 6:41have the right
- 6:41SPEAKER_01: recommendations and benchmarks for how to run their programs.
- 6:44SPEAKER_01: We're expanding the use of that.
- 6:46SPEAKER_01: There's another bucket of AI for security, which is
- 6:50SPEAKER_01: connecting those different types of tests together.
- 6:53SPEAKER_01: So it's one thing to say,
- 6:54SPEAKER_01: look,
- 6:54I have a pipeline from when my technology is developed to when
- 6:58it's in production,
- 6:59SPEAKER_01: and I have different testing techniques at each part.
- 7:03But what we see is a lot of value in ultimately reducing the
- 7:08bug density in
- 7:09SPEAKER_01: technology is connecting these things.
- 7:11SPEAKER_01: So you find a bunch of things in your bounty program.
- 7:13SPEAKER_01: Well,
- 7:13that means that you should be adjusting the pen test to make
- 7:15sure that you're
- 7:16SPEAKER_01: looking for those things in your pen test.
- 7:18SPEAKER_01: You find a bunch of things in your pen test.
- 7:19You need to continue shifting left and go further upstream and
- 7:22make sure that in
- 7:22SPEAKER_01: your code scanning,
- 7:23SPEAKER_01: you're also looking for those things and you're finding them earlier.
- 7:25SPEAKER_01: The earlier you can find vulnerabilities,
- 7:27SPEAKER_01: the much,
- 7:28much less expensive it is to fix and the much less risk you
- 7:32have to the overall
- 7:33SPEAKER_01: company.
- 7:34So that's another place where we're using AI to stitch those
- 7:36things together.
- 7:37And then the last point I would make about where we're using AI
- 7:40for security is in
- 7:41SPEAKER_01: the space of augmenting and complementing the human testing.
- 7:46And so what that means is enabling capabilities for fully
- 7:49automated scanning and
- 7:51hackbots to interact with our customers' programs alongside
- 7:55human researchers.
- 7:56SPEAKER_01: Absolutely.
- 7:56And I think that sort of combination of the AI and the human
- 8:00testing is something
- 8:01SPEAKER_02: that people are really interested in at the moment.
- 8:03Especially when they're not sure if they can fully trust AI or
- 8:06quite fully pass the
- 8:07SPEAKER_02: control over to it.
- 8:08SPEAKER_02: Yeah.
- 8:09SPEAKER_02: Yeah, absolutely.
- 8:11So taking a step back and looking at the cybersecurity
- 8:14landscape a little bit more
- 8:15SPEAKER_02: broadly,
- 8:16how do you see the future of cybersecurity changing as AI
- 8:20continues to evolve?
- 8:21I think cybersecurity is a fairly unique space when you start
- 8:27thinking about the
- 8:28SPEAKER_01: application of AI.
- 8:28SPEAKER_01: And I say that because already in the industry,
- 8:32cybersecurity was a little bit of a standout in that the
- 8:34lifecycle of cybersecurity
- 8:36SPEAKER_01: companies is much,
- 8:37SPEAKER_01: much shorter than most other tech companies.
- 8:40SPEAKER_01: And that's because the technology,
- 8:42the underlying technology that you need to be testing and
- 8:46looking for issues in
- 8:47SPEAKER_01: continues to evolve and change.
- 8:49And the techniques that attackers are using to exploit issues
- 8:53in that technology
- 8:54SPEAKER_01: are also evolving and changing.
- 8:56And I believe this is a space where there's going to continue
- 8:58to be a gap between
- 8:59what the human capability is and finding issues and what AI is
- 9:03able to do will
- 9:04SPEAKER_01: perpetuate for quite some time,
- 9:06SPEAKER_01: possibly even indefinitely.
- 9:09SPEAKER_01: simply because of the underlying dynamics.
- 9:11I think that is a very different dynamic than what you see in
- 9:13other realms.
- 9:14SPEAKER_01: I was sharing with somebody this morning,
- 9:17SPEAKER_01: you know,
- 9:17there's only so many humans that you need identifying cats in
- 9:20pictures before you
- 9:21have a machine that is able to identify cats in pictures better
- 9:24than most humans.
- 9:25SPEAKER_01: And it's not like you expect in 10 years that it's going to get,
- 9:28SPEAKER_01: you know,
- 9:28SPEAKER_01: significantly different.
- 9:29SPEAKER_01: Security is not like that.
- 9:31SPEAKER_01: And the underlying foundation of it changes so rapidly that the human,
- 9:35SPEAKER_01: I think,
- 9:35SPEAKER_01: is always going to be an incredibly important part of the loop.
- 9:37SPEAKER_01: Absolutely.
- 9:38SPEAKER_02: So kind of as an extension of that,
- 9:41what should CISOs and security leaders be thinking about today
- 9:44or what practical
- 9:45steps can they be taking today to kind of prepare themselves
- 9:49for that future?
- 9:51SPEAKER_01: So I think I would encourage CISOs to be taking a fairly,
- 9:55to zoom out and take a pretty holistic view of the tech stack
- 9:59that they are using
- 10:00SPEAKER_01: to secure.
- 10:01SPEAKER_01: You can look at all the vendors here at RSA,
- 10:04and there are so many vendors and not enough spend to throw at
- 10:08all of the vendors.
- 10:10SPEAKER_01: And each of them protects some sort of niche part of your environment.
- 10:14SPEAKER_01: There's a lot of overlap.
- 10:16But I would encourage CISOs to be really taking a step back and
- 10:19looking at this
- 10:19SPEAKER_01: holistically and saying,
- 10:20what is the right architecture that I need to put in place from
- 10:24a technology
- 10:24SPEAKER_01: perspective to be able to secure myself for the future of AI?
- 10:28SPEAKER_01: And they need to start implementing and moving towards that now.
- 10:32I would encourage CISOs to think a lot more about preventative
- 10:36and proactive security measures.
- 10:38SPEAKER_01: And I would put bounty,
- 10:40SPEAKER_01: vulnerability disclosure,
- 10:41SPEAKER_01: pen testing,
- 10:41that is all in the bucket of preventive and proactive because
- 10:45it's stuff that
- 10:45SPEAKER_01: you're doing before an incident happens.
- 10:49And I would also encourage CISOs to be thinking a lot more
- 10:51about welcoming in and
- 10:53SPEAKER_01: engaging communities outside of their walls.
- 10:56There's something like a 4.8 to 5 million human gap in talented
- 11:03cybersecurity
- 11:04SPEAKER_01: people around the world today.
- 11:05SPEAKER_01: And that gap is only widening every single year.
- 11:08And so CISOs need to be really strategic about how do they
- 11:11start leveraging
- 11:12SPEAKER_01: expertise from outside.
- 11:14SPEAKER_01: Absolutely.
- 11:14SPEAKER_01: Absolutely.
- 11:15It's hard for me not to get excited talking about HackerOne
- 11:17because I think
- 11:18SPEAKER_01: everything the company does is just so incredibly cool.
- 11:21And we talked a lot about how I believe through this
- 11:24combination of offensive
- 11:25security programs and by stitching them together and applying
- 11:28AI in the ways that
- 11:28SPEAKER_01: we talked about,
- 11:29I believe that there is a point that we are going to get to
- 11:32where I can go to a
- 11:33SPEAKER_01: customer and say,
- 11:34SPEAKER_01: okay,
- 11:34SPEAKER_01: we looked at this asset.
- 11:35SPEAKER_01: This asset has this amount of vulnerabilities.
- 11:37SPEAKER_01: We believe in it from some
- 11:38SPEAKER_01: initial scans.
- 11:40SPEAKER_01: And we think through a combination of these testing techniques,
- 11:43we can burn down the vulnerabilities in that piece of
- 11:45technology over this amount
- 11:46SPEAKER_01: of time and with this amount of money.
- 11:48And that kind of determinism is precisely the thing that's
- 11:51going to give CISOs and
- 11:53other executive buyers a lot more confidence that they're not
- 11:56just pouring money
- 11:57SPEAKER_01: into a black hole when they're investing in cybersecurity.
- 11:59SPEAKER_01: And I think we are very close to being able to do that.
- 12:01SPEAKER_01: So that's super exciting.
- 12:03SPEAKER_00: Can I just jump in with a little question as well?
- 12:05SPEAKER_00: You may.
- 12:05SPEAKER_00: Just from some of our conversations recently,
- 12:07we were hearing lots about zero days and kind of the market for
- 12:11that and how
- 12:12they're being kind of traded and sometimes brought up by
- 12:15insurance companies.
- 12:15SPEAKER_00: I don't know,
- 12:16do you have a kind of perspective on or maybe on the regulation
- 12:20side on how we
- 12:21SPEAKER_00: should approach zero days?
- 12:24SPEAKER_01: So I have a few thoughts.
- 12:25SPEAKER_01: Closure requirements to governments.
- 12:27What we have found is that those can have a chilling effect in
- 12:31organizations
- 12:32because what it results in is people don't want to know about
- 12:35their vulnerabilities
- 12:36SPEAKER_01: because if they find out about them,
- 12:37SPEAKER_01: they have to report them.
- 12:39SPEAKER_01: And so that I think has to be worked through.
- 12:43What we are advising and what we would advise policymakers is
- 12:47that they need to be
- 12:48creating these safe spaces for third parties to be reporting
- 12:51vulnerabilities.
- 12:52It needs to be a safe space for organizations to be hearing
- 12:55that they have
- 12:56vulnerabilities and they need to have enough both time but also
- 13:01agency and
- 13:03SPEAKER_01: requirement to then go fix the things that are found.
- 13:06And then how that fits into like disclosures back to government
- 13:10organizations,
- 13:11SPEAKER_01: I think is a separate piece of how to work that out.
- 13:13Tara,
- 13:14I also want to talk about a pretty exciting milestone for
- 13:17HackerOne.
- 13:18SPEAKER_02: You recently reached your 50th millionaire on the platform.
- 13:21Could you tell us more about the economic opportunity that you
- 13:24provide to your community?
- 13:25SPEAKER_01: You've got these, like, outrageously talented people.
- 13:30SPEAKER_01: And I'm talking about, like, this could be savants, like teenagers.
- 13:33SPEAKER_01: They have a computer, they have an internet connection.
- 13:36SPEAKER_01: They teach themselves how to hack.
- 13:38And we're giving them an opportunity to do that in a very
- 13:40productive and meaningful way.
- 13:42SPEAKER_01: and at the same time giving them an opportunity to make money.
- 13:44So we have many people that are earning really,
- 13:47really,
- 13:47really great dollars.
- 13:48SPEAKER_01: Like they are making livelihood for their family,
- 13:52SPEAKER_01: you know,
- 13:53SPEAKER_01: helping their parents with economics,
- 13:54SPEAKER_01: sending themselves to college,
- 13:56SPEAKER_01: all on the back of bug bounties.
- 13:59SPEAKER_01: I am particularly proud.
- 14:00We've paid out in our 12-year history $380 million in bounties
- 14:04to researchers around the world.
- 14:06And we have just passed the mark of 50 researchers have made
- 14:09more than a million
- 14:10SPEAKER_01: dollars on our platform.
- 14:11SPEAKER_01: The top top earner has made more than four million.
- 14:15SPEAKER_01: And it's just it's just a phenomenal model.
- 14:18SPEAKER_02: Yeah, absolutely.
- 14:19SPEAKER_02: It's incredible seeing.
- 14:21The impact that you're having on these people's lives and sort
- 14:24of encouraging them
- 14:25SPEAKER_02: to get involved,
- 14:26SPEAKER_02: but also put those skills to use in a good way.
- 14:28SPEAKER_02: Yes.
- 14:29SPEAKER_01: We got one person,
- 14:29like one of our hackers or probably more than one of our
- 14:31hackers was a doctor,
- 14:32SPEAKER_01: a full time doctor.
- 14:33SPEAKER_01: And then he hacks kind of on his side job.
- 14:36SPEAKER_01: I was talking to some guy, a guy based in Argentina.
- 14:40who you know he he had a family crisis and he had to step back
- 14:44and he was using the
- 14:44flexibility of being able to hack and now he's earning more
- 14:47than he ever did when
- 14:48he had like a full-time job um i was talking to a government
- 14:52organization just
- 14:53yesterday someone in asian asia pacific and they were telling
- 14:56me about um how they
- 14:57now have a program where their police are engaging with um
- 15:01youths that are showing
- 15:02you know starting to tend towards doing hacking but not in a
- 15:04very productive way
- 15:06SPEAKER_01: and they're they're having an effective program to basically
- 15:09bring those people over into doing fruitful security research
- 15:14as bounty hunters.
- 15:15SPEAKER_01: There's just so many cool, cool stories.
- 15:17SPEAKER_02: Yeah, absolutely.
- 15:18SPEAKER_02: And it's that nice balance between the technical side of it,
- 15:21SPEAKER_02: but also real creativity and finding new ways to do it.
- 15:25SPEAKER_02: Yeah.
- 15:26SPEAKER_02: There's hope for cybersecurity.
- 15:28SPEAKER_02: That's good to know.
- 15:30SPEAKER_02: So I just have one last question for you today.
- 15:33What is something that's really exciting you in the
- 15:35cybersecurity space at the moment?
- 15:38I'm very,
- 15:39very excited about what the impact can be of AI applied into
- 15:43this space.
- 15:43SPEAKER_01: Like this is a space where because of all of the tool sets,
- 15:46SPEAKER_01: all the tools are looking for different things,
- 15:48SPEAKER_01: but most of them have problems of false positives.
- 15:51And so unfortunately,
- 15:53you know,
- 15:53I talked about the dearth of cybersecurity talent.
- 15:55We've got our most talented cybersecurity folks that are
- 15:58basically buried in
- 15:59drudgery of digging through piles and piles of false positives
- 16:02and trying to figure
- 16:03SPEAKER_01: out,
- 16:03SPEAKER_01: you know,
- 16:04SPEAKER_01: this one I need to take action on and this one I don't.
- 16:07And I think AI can have a huge impact in enabling those folks
- 16:11to be much more
- 16:11SPEAKER_01: effective,
- 16:12make their time spent in a much more rewarding way and
- 16:16basically be a force
- 16:17SPEAKER_01: multiplier for what we need.
- 16:19SPEAKER_01: And I think when we enable our defenders to do that,
- 16:22SPEAKER_01: we can quickly overtake the cyber criminals.
- 16:25SPEAKER_02: Absolutely.
- 16:25SPEAKER_02: I think that's a really strong, positive message to end on.
- 16:28SPEAKER_02: Thank you so much for taking the time to speak with us today.
- 16:31SPEAKER_02: I've really enjoyed the conversation.
- 16:32SPEAKER_01: Really appreciate the time as well.