Latest / Expert Insights Podcast / #96. Phishing 3.0: Why AI and Deepfakes Are Rewriting the Attack Playbook with IRONSCALES
Transcript
- 0:04SPEAKER_00: Hello everyone and welcome to the Expert Insights Podcast.
- 0:07SPEAKER_00: Few areas of cybersecurity have been reshaped by AI as quickly as phishing.
- 0:12SPEAKER_00: In just three years,
- 0:13SPEAKER_00: generative AI has rewritten the economics of the attack
- 0:16SPEAKER_00: and defenders are running faster than ever just to keep pace.
- 0:19SPEAKER_00: Today, we're stepping back to look at the bigger picture,
- 0:22SPEAKER_00: how the threat has fundamentally changed
- 0:24SPEAKER_00: and what security leaders should actually be doing about it.
- 0:28SPEAKER_00: I'm joined by Steve Malone,
- 0:29SPEAKER_00: Chief Product and Strategy Officer at Ionscales,
- 0:32SPEAKER_00: an AI-powered email security platform trusted by more than 17,000 organizations.
- 0:38SPEAKER_00: Steve has spent two decades in email security,
- 0:40SPEAKER_00: including senior product leadership roles at Mimecast and Egress,
- 0:44SPEAKER_00: so he's had a front row seat to how this threat has evolved.
- 0:48SPEAKER_01: Steve, welcome to the show.
- 0:50SPEAKER_01: You've been in email security for the best part of two decades.
- 0:54SPEAKER_00: When you look at the phishing threat today versus even three or four years ago,
- 0:58SPEAKER_00: what's different and what are just the same problems
- 1:01SPEAKER_00: that might have been accelerated?
- 1:03SPEAKER_01: So, first of all, thank you for having me.
- 1:04SPEAKER_01: It's great to be here.
- 1:05SPEAKER_01: And when you say 20 years,
- 1:07SPEAKER_01: that sounds terrifying and terrible, but unfortunately or fortunately...
- 1:12SPEAKER_01: It is true.
- 1:13SPEAKER_01: So I've been working within the field of cyber and especially email security now for
- 1:16SPEAKER_01: a very long time.
- 1:17SPEAKER_01: So I've seen the market really evolve over time.
- 1:21SPEAKER_01: I've seen the attack surface evolve and the way
- 1:24SPEAKER_01: that vendors are approaching addressing that attack surface and
- 1:28SPEAKER_01: that threat landscape.
- 1:30SPEAKER_01: So I think When I look at the attack surface,
- 1:34SPEAKER_01: when I look at the threat landscape,
- 1:35SPEAKER_01: what's interesting is that I can frame it in three ways.
- 1:39SPEAKER_01: So email is still the primary channel of attack for any organization.
- 1:44SPEAKER_01: The people within the organization are the biggest risk,
- 1:48SPEAKER_01: but now agents and the use of AI is what I would class as the new risk frontier.
- 1:54SPEAKER_01: So I think what's interesting is that Email as a technology has not moved on greatly
- 2:00SPEAKER_01: in the last 20 plus years,
- 2:02SPEAKER_01: which means that the attack methods also haven't really evolved.
- 2:07SPEAKER_01: The goals of the attackers haven't really evolved.
- 2:09SPEAKER_01: So they're looking for credentials.
- 2:11SPEAKER_01: They're looking to elicit a payment.
- 2:15SPEAKER_01: They're looking for some kind of foothold that they can use email to get within
- 2:20SPEAKER_01: the organizations.
- 2:21SPEAKER_01: And what's interesting is the targets haven't really moved either.
- 2:25SPEAKER_01: So most email attacks are based on actually attacking a person's judgment.
- 2:31SPEAKER_01: So it's the people using email, not necessarily now at the mail server.
- 2:36SPEAKER_01: But what has changed is speed and quality.
- 2:39SPEAKER_01: So the AI revolution over the last couple
- 2:42SPEAKER_01: of years has absolutely revolutionized productivity,
- 2:46SPEAKER_01: but it has also revolutionized the way that attackers work.
- 2:52SPEAKER_01: Before AI, it was rare to find an attack that was high quality, very, very convincing.
- 2:59SPEAKER_01: It took a lot of manual research,
- 3:01SPEAKER_01: but I think now we're in a situation where it costs pennies to create very,
- 3:08SPEAKER_01: very hyper-realistic attacks,
- 3:11SPEAKER_01: to scale those to target multiple organizations and many employees.
- 3:17SPEAKER_01: And all of the old red flags have disappeared.
- 3:19SPEAKER_01: So people used to talk about looking for bad grammar,
- 3:24SPEAKER_01: odd phrasing, generic sign-off, generic greetings.
- 3:28SPEAKER_01: But all of that has now been removed.
- 3:30SPEAKER_01: AI has polished all of those pieces out of attackers, or out of these attacks.
- 3:36SPEAKER_01: So attackers are using AI to probe and then to launch these very,
- 3:41SPEAKER_01: very personalized, hyper-realistic attacks.
- 3:44SPEAKER_01: And that's really the big change.
- 3:46SPEAKER_00: Absolutely.
- 3:47SPEAKER_00: And I want to kind of touch on that side of things a little bit more.
- 3:51SPEAKER_00: Your new business cost of phishing report that Ironscale's released this year found
- 3:57SPEAKER_00: that sort of almost counterintuitive point
- 4:00SPEAKER_00: that teams are handling each phishing attack faster than ever,
- 4:03SPEAKER_00: yet phishing is costing them more.
- 4:05SPEAKER_00: Could you go a little bit more into that side of things and what it tells us about
- 4:09SPEAKER_00: the way most organizations are approaching the problem?
- 4:12SPEAKER_01: Yes, definitely.
- 4:13SPEAKER_01: So that was our Osterman research report.
- 4:16SPEAKER_01: So we surveyed 128 organizations or security professionals or organizations.
- 4:22SPEAKER_01: They were organizations of between 1,000 and 5,000 employees, so fairly decent size.
- 4:26SPEAKER_01: And we We benchmarked against our first report that we did with Osterman,
- 4:32SPEAKER_01: which was back in 2022.
- 4:34SPEAKER_01: And we did that because 2022 was before ChatGPT came into the world.
- 4:40SPEAKER_01: So it was a nice before and after with the same questions.
- 4:44SPEAKER_01: So what we found was interesting is that security teams have become faster.
- 4:49SPEAKER_01: So the handling time per incident actually fell.
- 4:55SPEAKER_01: The cost per phishing email fell.
- 4:58SPEAKER_01: So those are actually gains.
- 5:00SPEAKER_01: But what was interesting was that the total bill went up anyway.
- 5:03SPEAKER_01: So phishing now consumes, according to our report, 36.5% of security team hours.
- 5:09SPEAKER_01: And that's up from 33% or just over 33%.
- 5:13SPEAKER_01: So that's an interesting dichotomy.
- 5:15SPEAKER_01: And the reason that we saw was that Efficiency gains are quite linear,
- 5:21SPEAKER_01: but volume growth is not.
- 5:23SPEAKER_01: So even though the unit cost fell, the actual unit count.
- 5:28SPEAKER_01: So the volume of attacks rose much, much, much faster.
- 5:33SPEAKER_01: And this is really the challenge.
- 5:35SPEAKER_01: Attackers control the volume dial on these attacks.
- 5:39SPEAKER_01: And because the barriers to entry for creating these attacks have fallen,
- 5:43SPEAKER_01: attackers can create,
- 5:45SPEAKER_01: as I was saying earlier,
- 5:46SPEAKER_01: attackers can create much more realistic attacks at much, much, much higher volume.
- 5:51SPEAKER_01: So that really describes what we're seeing.
- 5:54SPEAKER_01: Everybody in the security world has got faster,
- 5:58SPEAKER_01: but attackers have got faster at a faster rate at launching these attacks.
- 6:03SPEAKER_01: So I think what we're seeing is you can't really win on cycle time when
- 6:08SPEAKER_01: the adversary is the one that is setting the volume.
- 6:11SPEAKER_01: So the way that we see this or the way that we interpret this is that
- 6:15SPEAKER_01: the right approach Absolutely.
- 6:17SPEAKER_01: Absolutely.
- 6:40SPEAKER_00: Thinking of that current era as Ironscales are framing it at the moment
- 6:45SPEAKER_00: as phishing 3.0, what does that actually mean for a security leader?
- 6:49SPEAKER_00: And why do the defenses that worked in the last era start to break down?
- 6:53SPEAKER_00: What are some of the new things that they should be looking for?
- 6:56SPEAKER_01: So that's a great question.
- 6:58SPEAKER_01: And as I said earlier, I've been in this industry for a long time.
- 7:02SPEAKER_01: So I've seen the email security threat landscape and also vendor landscape evolve.
- 7:08SPEAKER_01: And when we talk about phishing 3.0,
- 7:09SPEAKER_01: we're Essentially talking about the evolution of the way that attackers use email
- 7:17SPEAKER_01: to breach organizations.
- 7:18SPEAKER_01: So if you think about it in the most simple terms,
- 7:21SPEAKER_01: which is 1.0,
- 7:22SPEAKER_01: 2.0 and 3.0,
- 7:24SPEAKER_01: using terminology that IT professionals So the first wave
- 7:30SPEAKER_01: of phishing was really things like bad content.
- 7:33SPEAKER_01: So going way, way,
- 7:35SPEAKER_01: way back when emails had an infected attachment or a weaponized attachment,
- 7:40SPEAKER_01: maybe an executable file attached to the email.
- 7:43SPEAKER_01: That moved on to things like malicious links being used in emails
- 7:49SPEAKER_01: and even Microsoft Office files being weaponized through the use of unsafe macros.
- 7:54SPEAKER_01: But all of that 1.0 phishing was essentially reliant on there being something within
- 8:01SPEAKER_01: the email that would download a piece of ransomware or direct a user to
- 8:07SPEAKER_01: an external website.
- 8:09SPEAKER_01: Now, the fishing world evolved.
- 8:12SPEAKER_01: And when we talk about fishing 2.0,
- 8:14SPEAKER_01: that was the evolution where attackers realized that Vendors were getting better
- 8:20SPEAKER_01: at blocking emails with attachments or with links or,
- 8:23SPEAKER_01: you know, with any kind of executable content.
- 8:25SPEAKER_01: And that was where these attacks moved on to things like BEC,
- 8:28SPEAKER_01: business email compromise,
- 8:30SPEAKER_01: impersonation attacks,
- 8:32SPEAKER_01: especially of senior execs,
- 8:34SPEAKER_01: attacks that used social engineering rather than code within the emails
- 8:41SPEAKER_01: to persuade users to take unsafe action.
- 8:44SPEAKER_01: So, you know, Please, can you send me the payroll details for these individuals?
- 8:51SPEAKER_01: Or I'm the CFO.
- 8:53SPEAKER_01: I really need to make this payment, but I'm out of the office.
- 8:57SPEAKER_01: So can you make this payment for me?
- 8:59SPEAKER_01: Here are the details.
- 9:00SPEAKER_01: All of those attacks relied purely on persuasion and they were harder to detect.
- 9:05SPEAKER_01: So what we're seeing now is this era of phishing 3.0,
- 9:09SPEAKER_01: which is essentially the evolution of all of those attacks.
- 9:14SPEAKER_01: Which has come about because of AI, but also the framing of these attacks as multi-channel.
- 9:21SPEAKER_01: So email attacks used to be very, very focused on email only.
- 9:25SPEAKER_01: We've seen an increase and an uptick in both speed and quality of attacks,
- 9:31SPEAKER_01: but we've also seen all of these attacks becoming multi-channel or multi-modal
- 9:35SPEAKER_01: as well.
- 9:37SPEAKER_01: So what that means is that a single campaign can start with email,
- 9:42SPEAKER_01: can initiate an account takeover,
- 9:45SPEAKER_01: can allow an attacker to move into a collaboration tool like Microsoft Teams,
- 9:49SPEAKER_01: and then use the access They have within teams to use a deepfake persona,
- 9:55SPEAKER_01: either both voice or video,
- 9:57SPEAKER_01: to again persuade people in the organization to take unsafe actions.
- 10:02SPEAKER_01: So really moving across the systems that people rely on.
- 10:06SPEAKER_01: And the reason that the old type of defenses break is that There is no content to inspect now.
- 10:15SPEAKER_01: We don't have executables within mails.
- 10:18SPEAKER_01: There are no attachments.
- 10:19SPEAKER_01: Email reputation,
- 10:21SPEAKER_01: signature models aren't able to keep up
- 10:24SPEAKER_01: with these high-powered AI-generated attacks.
- 10:27SPEAKER_01: And the other piece,
- 10:28SPEAKER_01: going back to how I framed the threat landscape earlier,
- 10:32SPEAKER_01: is that if you are attacking a human,
- 10:34SPEAKER_01: you Then that assumed that there was some kind of attacker error or human error.
- 10:40SPEAKER_01: And we historically trained people to spot mistakes within emails.
- 10:46SPEAKER_01: But again, with attacks becoming more realistic, that has also become more difficult.
- 10:52SPEAKER_01: So essentially for security leaders,
- 10:54SPEAKER_01: the response to this,
- 10:56SPEAKER_01: the response model shifts away from purely blocking the to detect and respond,
- 11:01SPEAKER_01: but now moving forwards to being able to preempt attacks.
- 11:06SPEAKER_01: So essentially,
- 11:06SPEAKER_01: you need to know what an attack designed for your specific organization looks like
- 11:12SPEAKER_01: before it arrives.
- 11:13SPEAKER_00: Absolutely.
- 11:14SPEAKER_00: I think that's a really important question there around the Sort of employee mistake
- 11:20SPEAKER_00: or employee training side of things because historically we would teach
- 11:25SPEAKER_00: the end users to spot the red flags in an email or trust voice rather
- 11:30SPEAKER_00: than just text.
- 11:32SPEAKER_00: So do you think that we're having to move away from that user training or is
- 11:36SPEAKER_00: that still an important piece here within the overall solution to this?
- 11:41SPEAKER_01: So I think it has a place and the security awareness
- 11:45SPEAKER_01: and training market has itself evolved greatly over the last seven or eight years,
- 11:52SPEAKER_01: moving away from purely training-based offerings to talking
- 11:56SPEAKER_01: about secure behavior management or human risk management.
- 12:00SPEAKER_01: But essentially, the way that we think about making humans secure is Really needs to change.
- 12:06SPEAKER_01: So you mentioned that whole red flag checklist.
- 12:09SPEAKER_01: So, you know, in the old days, we taught people to look for typos within an email.
- 12:15SPEAKER_01: We asked them to look for a sign-off that maybe wasn't familiar.
- 12:20SPEAKER_01: So Stephen, rather than Steve, as an example, you know, a sense of urgency.
- 12:25SPEAKER_01: All of those things were the artifacts of very basic and cheap attacks.
- 12:30SPEAKER_01: And we also used to give,
- 12:32SPEAKER_01: and when I say we,
- 12:33SPEAKER_01: I mean the security industry, We used to give advice, well, verify.
- 12:39SPEAKER_01: So if you receive an email asking you to transfer a million dollars,
- 12:42SPEAKER_01: pick up the phone and verify that.
- 12:45SPEAKER_01: But again,
- 12:45SPEAKER_01: that assumes that the attacker hasn't already put that into the attack framework
- 12:51SPEAKER_01: and is not using a deepfake for either voice or video to scam you when you make
- 12:57SPEAKER_01: that verification call.
- 12:58SPEAKER_01: So again, even that longstanding advice is now exploitable.
- 13:03SPEAKER_01: You know, again,
- 13:04SPEAKER_01: as part of the survey work that we've done, we've done a lot of work on deepfakes.
- 13:09SPEAKER_01: Our stat is that 62.5% of respondents say
- 13:12SPEAKER_01: that deepfake attacks are already being disruptive in their organizations.
- 13:18SPEAKER_01: So, this is an emerging but very rapidly emerging attack factor.
- 13:23SPEAKER_01: So, Going back to what I was saying earlier,
- 13:27SPEAKER_01: the way that we asked people to verify trust now really becomes one of
- 13:32SPEAKER_01: the least verifiable ways of trusting.
- 13:35SPEAKER_01: Pick up the phone or jump on a call with somebody isn't now trustworthy.
- 13:39SPEAKER_01: What's interesting is that the social engineering element of this hasn't really changed.
- 13:45SPEAKER_01: So still some of those things like pretext, authority, urgency, they give context.
- 13:53SPEAKER_01: But the challenge is that AI has made the delivery of those things completely flawless.
- 13:59SPEAKER_01: That's really the problem.
- 14:01SPEAKER_01: So essentially,
- 14:02SPEAKER_01: when we're thinking about how we teach our end users to be secure now,
- 14:07SPEAKER_01: rather than saying you need to spot the fake,
- 14:10SPEAKER_01: what we need to be telling users is, okay,
- 14:13SPEAKER_01: you need to verify the process that you're being asked to undertake and And in
- 14:20SPEAKER_01: a much more robust way.
- 14:21SPEAKER_01: So call back on a number that you have from your own records,
- 14:26SPEAKER_01: not the one that you're seeing in the email that has been sent to you.
- 14:30SPEAKER_01: And also organizations need to think in a more procedural way.
- 14:34SPEAKER_01: At a business level on how they handle certain high value processes.
- 14:39SPEAKER_01: So as an example,
- 14:41SPEAKER_01: for large transactions,
- 14:43SPEAKER_01: organizations need to think about who has the authority to make bank detail changes,
- 14:48SPEAKER_01: credential resets,
- 14:49SPEAKER_01: that kind of stuff, wire transfer approvals, requests for HR data.
- 14:54SPEAKER_01: What is the business process and who is involved in that process?
- 14:58SPEAKER_01: And the controls
- 15:02SPEAKER_01: Shouldn't be reliant on somebody noticing that something feels wrong.
- 15:07SPEAKER_01: There should be a defined process for anything high value relating to data or money
- 15:11SPEAKER_01: within the organization.
- 15:13SPEAKER_00: Absolutely, absolutely.
- 15:15SPEAKER_00: So thinking about all of those challenges,
- 15:17SPEAKER_00: the extreme high volume of attacks,
- 15:20SPEAKER_00: the flawless AI-generated content and deepfakes,
- 15:23SPEAKER_00: and also these cross-channel attacks,
- 15:25SPEAKER_00: I want to talk a little bit about how you tackle this at Iron Scales.
- 15:29SPEAKER_00: The platform's built on AI and integrates directly with Microsoft 365 and Google Workspace.
- 15:36SPEAKER_00: For a security leader trying to get ahead of these threats and these challenges,
- 15:40SPEAKER_00: what does iron scales do differently from the defenses
- 15:43SPEAKER_00: that most organizations already have in place?
- 15:47SPEAKER_01: The way that organizations think about email security products has changed, again,
- 15:53SPEAKER_01: massively over the last 20 years from on-premise exchange servers
- 15:59SPEAKER_01: with on-premise email appliances or email security appliances
- 16:03SPEAKER_01: and antivirus appliances and spam appliances,
- 16:06SPEAKER_01: moving through the era of secure email gateways,
- 16:10SPEAKER_01: consolidating all of those services.
- 16:13SPEAKER_01: And then consolidating those services into the cloud.
- 16:16SPEAKER_01: And then Microsoft 365 and Google really expanding their core platforms,
- 16:23SPEAKER_01: making secure email gateways less relevant because they add layers
- 16:28SPEAKER_01: of additional cost and additional functionality duplication.
- 16:32SPEAKER_01: So The landscape or the ecosystem has changed and we fit into the category of native API.
- 16:41SPEAKER_01: So we are natively API integrated into either Microsoft 365 or Google.
- 16:48SPEAKER_01: The benefit there, the organization doesn't need to change their MX record.
- 16:52SPEAKER_01: There isn't a gateway.
- 16:53SPEAKER_01: There are no agents to install.
- 16:55SPEAKER_01: So we're very fast and easy to install.
- 16:58SPEAKER_01: And most importantly, we run alongside what the organization already has in place.
- 17:03SPEAKER_01: So if they have Microsoft 365 and a SEG, we provide defense in depth.
- 17:08SPEAKER_01: If they have Microsoft 365 plus Defender and no secure email gateway,
- 17:13SPEAKER_01: we can also provide a layered set of defense in depth as well.
- 17:17SPEAKER_01: So from a deployment point of view, we sit right in the middle of that layer of flexibility.
- 17:23SPEAKER_01: Our detection approach as well is very AI native and AI first.
- 17:27SPEAKER_01: So again,
- 17:28SPEAKER_01: rather than relying on signatures and policies,
- 17:32SPEAKER_01: adaptive AI essentially builds like a communication baseline, if you like.
- 17:37SPEAKER_01: So a social graph for every user.
- 17:39SPEAKER_01: So the question that we ask is,
- 17:42SPEAKER_01: Is this normal communication for the relationship that this person has rather
- 17:47SPEAKER_01: than is this single email malicious?
- 17:51SPEAKER_01: So that's how we catch BEC, right?
- 17:52SPEAKER_01: That's the right approach.
- 17:53SPEAKER_01: It's looking at communication chains rather than individual mails.
- 17:58SPEAKER_01: Again, going back to what we were talking about earlier on
- 18:01SPEAKER_01: about security teams actually having more to deal with.
- 18:06SPEAKER_01: We address that with our agentic virtual SOC.
- 18:10SPEAKER_01: So what we try to do is handle as much of the remediation as possible
- 18:15SPEAKER_01: without having humans involved.
- 18:18SPEAKER_01: So we want to take away the workload from SOC teams and from security professionals.
- 18:24SPEAKER_01: That drops the response time down and it's how we balance out that mismatch
- 18:28SPEAKER_01: of statistics that we saw in the report earlier on.
- 18:31SPEAKER_01: And we also have a set of agents that really help to bring in that what we call
- 18:36SPEAKER_01: the preemptive layer that I talked about earlier on.
- 18:39SPEAKER_01: So we have a red teaming agent that this is one of our most exciting pieces
- 18:44SPEAKER_01: of innovation over the last year.
- 18:46SPEAKER_01: The red teaming agent essentially researches your organization on your behalf.
- 18:52SPEAKER_01: Using external data sources, using things like OSINT.
- 18:57SPEAKER_01: And then the agent attacks or virtually attacks your instance of iron scales,
- 19:04SPEAKER_01: works out if there are any weak points,
- 19:06SPEAKER_01: and then we preemptively harden the defenses for your organization based on what
- 19:12SPEAKER_01: that agent has found.
- 19:14SPEAKER_01: So rather than relying on one detect The other thing that makes us important in
- 19:18SPEAKER_01: the industry, I guess, is that we're a consolidation point.
- 19:31SPEAKER_01: We're focused on detection.
- 19:33SPEAKER_01: We're also focused on account takeover.
- 19:35SPEAKER_01: We're focused on human risk management through the lens of security awareness,
- 19:40SPEAKER_01: security training, demarc, encryption, deepfake protection.
- 19:44SPEAKER_01: So we're essentially filling in the gaps where those underlying platforms are weak.
- 19:49SPEAKER_01: So where Microsoft or Google are not strong, we fill the gaps, we augment those platforms.
- 19:56SPEAKER_00: Absolutely.
- 19:56SPEAKER_00: Thank you.
- 19:57SPEAKER_00: So you've spoken there through...
- 20:01SPEAKER_00: I want to touch on that human risk management piece
- 20:06SPEAKER_00: because we did mention security awareness training earlier and how that has been
- 20:11SPEAKER_00: a part of the fight for years, but it does need to change.
- 20:14SPEAKER_00: So Ionscales takes an approach with adaptive AI generated simulations that are based
- 20:19SPEAKER_00: on real reconnaissance.
- 20:20SPEAKER_00: Could you talk me through that a little bit more and how
- 20:23SPEAKER_00: that human layer fits alongside the technology?
- 20:26SPEAKER_01: Definitely.
- 20:28SPEAKER_01: People are users of email and it's the people in the organizations who are making mistakes.
- 20:36SPEAKER_01: It's very easy to point the finger of blame at people within organizations,
- 20:42SPEAKER_01: but ultimately it's our responsibility to help make people as cyber aware
- 20:48SPEAKER_01: as possible.
- 20:49SPEAKER_01: But those old methods aren't working.
- 20:51SPEAKER_01: So If you think about the classic approach to security awareness and training,
- 20:56SPEAKER_01: an organization would create a very generic phishing testing template
- 21:01SPEAKER_01: and they would send that individual mail out to everybody in the organization.
- 21:06SPEAKER_01: So the same mail for all users.
- 21:08SPEAKER_01: Most of the time,
- 21:10SPEAKER_01: those very generic templates were something that would maybe be roughly applicable
- 21:16SPEAKER_01: to the organization and The right way to approach this now is essentially to use AI.
- 21:20SPEAKER_01: And again, we have an agency.
- 21:33SPEAKER_01: We've kind of reversed that.
- 21:34SPEAKER_01: We're doing the
- 21:48SPEAKER_01: The same research or the agent is doing the same research on behalf of our customers.
- 21:52SPEAKER_01: That agent essentially uses OSINT and external data to research users within
- 21:57SPEAKER_01: the organization and then enables you to send very personalized attacks
- 22:01SPEAKER_01: to each user individually.
- 22:03SPEAKER_01: Again, that cuts down the time the security teams need to spend
- 22:08SPEAKER_01: on security awareness training,
- 22:10SPEAKER_01: but it also increases the efficiency of those attacks.
- 22:14SPEAKER_01: So essentially, we're sending the right simulation to Absolutely.
- 22:17SPEAKER_01: Thank you.
- 22:18SPEAKER_01: Absolutely.
- 22:19SPEAKER_01: Thank you.
- 22:35SPEAKER_00: The sort of deepfake side of things because Ironscales has launched
- 22:39SPEAKER_00: a deepfake protection for Microsoft Teams.
- 22:41SPEAKER_00: And as you said earlier, they are one of the fastest moving emerging threats.
- 22:45SPEAKER_00: So how does this new deepfake protection work in practice?
- 22:50SPEAKER_00: Why is that real-time verification becoming so important?
- 22:55SPEAKER_01: This is one of the most exciting products that we have in the portfolio.
- 22:58SPEAKER_01: One thing to be very aware of is that it links very tightly with our other products.
- 23:06SPEAKER_01: And the main reason for that is because we're protecting the platforms that our customers use.
- 23:11SPEAKER_01: So if you think about an organization using Microsoft 365,
- 23:14SPEAKER_01: they're sending and receiving email through that platform,
- 23:17SPEAKER_01: but they're also very likely using Microsoft Teams for chat and
- 23:23SPEAKER_01: for video collaboration.
- 23:24SPEAKER_01: And now that it's becoming much easier for Teams to be used with external parties,
- 23:30SPEAKER_01: both for video, voice and chat, It opens up that part of the attack surface.
- 23:37SPEAKER_01: But if you think about what I was saying earlier,
- 23:39SPEAKER_01: that attacks are multimodal or multichannel,
- 23:42SPEAKER_01: an attack can start in email,
- 23:45SPEAKER_01: but that email can attempt to gain credentials,
- 23:50SPEAKER_01: essentially an account takeover attack.
- 23:53SPEAKER_01: If that account takeover attack is successful,
- 23:56SPEAKER_01: It's very likely that the attacker will gain access to
- 24:01SPEAKER_01: the Microsoft 365 user account and then move laterally into other applications,
- 24:05SPEAKER_01: Microsoft Teams being a great example.
- 24:08SPEAKER_01: So if Teams and video form part of that attack chain,
- 24:13SPEAKER_01: So essentially,
- 24:14SPEAKER_01: what we have built is the ability to do real-time identity verification of users
- 24:21SPEAKER_01: in Teams meetings, essentially.
- 24:23SPEAKER_01: So it's both visual and audio analysis.
- 24:26SPEAKER_01: And that's created against a learned baseline,
- 24:30SPEAKER_01: I guess, is the best way that I can describe that.
- 24:32SPEAKER_01: So we do automatic profiling.
- 24:34SPEAKER_01: We're baselining.
- 24:36SPEAKER_01: The people that are in those meetings based on both audio as much as video.
- 24:41SPEAKER_01: So it's far easier to spoof voice.
- 24:43SPEAKER_01: So obviously that's a really common use case.
- 24:45SPEAKER_01: We don't record anything.
- 24:47SPEAKER_01: We're not logging anything.
- 24:49SPEAKER_01: So this isn't a solution that has privacy concerns.
- 24:52SPEAKER_01: And we try to be...
- 24:54SPEAKER_01: As unintrusive as possible with the product when customers are using it.
- 24:59SPEAKER_01: So even through the response that we have,
- 25:02SPEAKER_01: so rather than just kicking people out of the meetings,
- 25:06SPEAKER_01: we have a process of graduated response.
- 25:09SPEAKER_01: So we warn either the meeting owner or the participants that there may be somebody
- 25:13SPEAKER_01: in the meeting that we can't be sure of the identity of.
- 25:18SPEAKER_01: And then we intervene if no action is taken.
- 25:21SPEAKER_01: So the most important thing, though, is that this is real time.
- 25:24SPEAKER_01: And that's really the requirement for this scenario.
- 25:27SPEAKER_01: The fraud is happening in that live call.
- 25:30SPEAKER_01: So if you...
- 25:32SPEAKER_01: If you were to review that call afterwards, it's too late.
- 25:35SPEAKER_01: Those things have already happened.
- 25:36SPEAKER_01: And we're extending the deepfake product currently to cover what we see as one of
- 25:41SPEAKER_01: the most strongly emerging deepfake use cases,
- 25:44SPEAKER_01: which is the recruiter or HR use case, as we call it.
- 25:49SPEAKER_01: We are seeing globally as an industry,
- 25:53SPEAKER_01: the huge proliferation of deepfakes being used In the candidate recruitment process,
- 25:59SPEAKER_01: and often it's nation state attackers who are initiating applications for jobs.
- 26:05SPEAKER_01: They're using deepfake personas to essentially go through the interview process with
- 26:12SPEAKER_01: an organization with the ultimate goal to receive a laptop from that organization
- 26:17SPEAKER_01: as part of the onboarding.
- 26:19SPEAKER_01: And obviously,
- 26:19SPEAKER_01: if you have a laptop as part of your new job,
- 26:21SPEAKER_01: you're given credentials, you're given access to systems.
- 26:25SPEAKER_01: If you can make that attack work, it's an easy way in.
- 26:27SPEAKER_01: And with deepfakes becoming easier to achieve,
- 26:31SPEAKER_01: and again,
- 26:31SPEAKER_01: like phishing,
- 26:33SPEAKER_01: way more realistic,
- 26:35SPEAKER_01: this is a real attack vector that organizations really need to think about
- 26:39SPEAKER_01: in detail.
- 26:40SPEAKER_00: Definitely.
- 26:41SPEAKER_00: So with the recruitment deepfakes,
- 26:44SPEAKER_00: will the technology work in a similar way,
- 26:46SPEAKER_00: even if the person is external to the company?
- 26:48SPEAKER_00: Does it sort of rely on them having to have A few phone calls with the HR department
- 26:53SPEAKER_00: during that sort of interview process.
- 26:56SPEAKER_01: So again, we're building this in as least intrusive way as possible.
- 27:02SPEAKER_01: So we'll be able to integrate with ATS systems that the organizations have.
- 27:08SPEAKER_01: Within, you know,
- 27:09SPEAKER_01: within their existing infrastructure,
- 27:12SPEAKER_01: we'll be able to essentially take a look at resumes or CVs for folks in the UK
- 27:17SPEAKER_01: to start really at the point of application.
- 27:19SPEAKER_01: So, you know, are there problems with the application?
- 27:23SPEAKER_01: Again, using Doing OSINT research,
- 27:25SPEAKER_01: being able to understand if this persona is fake even before it gets to
- 27:30SPEAKER_01: the real-time element.
- 27:31SPEAKER_01: But again,
- 27:32SPEAKER_01: we'll be able to analyze video and audio content really to go defense in depth,
- 27:38SPEAKER_01: to be very belt and braces in our approach.
- 27:41SPEAKER_00: Yeah, brilliant.
- 27:41SPEAKER_00: That's really exciting.
- 27:43SPEAKER_00: I know that it's something that a lot of organizations are very concerned about at the moment.
- 27:47SPEAKER_00: So yeah, that's a fantastic product.
- 27:50SPEAKER_00: Looking forward to see it come to life.
- 27:53SPEAKER_00: So I've just got one more question for you today, Steve.
- 27:56SPEAKER_00: For the CISAs and security leaders that are listening,
- 27:59SPEAKER_00: if they take just two or three things away from this conversation to act on now,
- 28:04SPEAKER_00: what should those things be?
- 28:06SPEAKER_01: That's a great question because CISOs and security leaders are under a lot
- 28:11SPEAKER_01: of pressure as they always have been.
- 28:13SPEAKER_01: But again,
- 28:15SPEAKER_01: going back to the topic of AI,
- 28:18SPEAKER_01: the pressure that security professionals are now under is increasing exponentially.
- 28:23SPEAKER_01: So I think a couple of the most important things,
- 28:28SPEAKER_01: think differently about what you're actually measuring.
- 28:30SPEAKER_01: So as we talked about earlier on, attack volume is increasing.
- 28:35SPEAKER_01: That will continue.
- 28:36SPEAKER_01: So think about different ways to measure the success of your security programs.
- 28:41SPEAKER_01: So looking at,
- 28:42SPEAKER_01: as an example,
- 28:43SPEAKER_01: if you look at mean time to remediate in isolation, that will become misleading.
- 28:50SPEAKER_01: Reframe that as the total share of security team hours consumed by phishing.
- 28:57SPEAKER_01: Rather than focusing on very specific items.
- 29:00SPEAKER_01: So try to measure differently.
- 29:02SPEAKER_01: Again, thinking about the people in your organization,
- 29:04SPEAKER_01: I would say assume the content detection,
- 29:09SPEAKER_01: that red flag training are just no longer sufficient by themselves.
- 29:15SPEAKER_01: Consider putting in or even hardening the processes that you have within your organization.
- 29:20SPEAKER_01: So anything critical that involves the movement of money, Credentials, the movement of data.
- 29:27SPEAKER_01: Make those processes out of band.
- 29:29SPEAKER_01: Review how you're doing those things at the moment.
- 29:32SPEAKER_01: Move away from just depending on somebody spotting that something feels wrong.
- 29:38SPEAKER_01: And I think thirdly,
- 29:39SPEAKER_01: and we're We've talked about AI a lot in this conversation,
- 29:42SPEAKER_01: mainly through the lens of how attackers are using AI.
- 29:45SPEAKER_01: But AI is also being used by vendors.
- 29:49SPEAKER_01: So it's still an arms race and vendors have access to AI models and tooling
- 29:55SPEAKER_01: that attackers don't.
- 29:57SPEAKER_01: Which is, you know, which gives us a fantastic advantage.
- 30:00SPEAKER_01: So the third tip I would say is ask your vendors what AI they actually have access
- 30:06SPEAKER_01: to,
- 30:07SPEAKER_01: not just within the products that you're consuming from them,
- 30:11SPEAKER_01: but also what access they have to frontier models through things like,
- 30:16SPEAKER_01: you know, Anthropics program or Open Data.
- 30:19SPEAKER_01: OpenAI's program to ask how they are using AI and what models they have access to
- 30:25SPEAKER_01: that will benefit the detection and the protection that they provide for you.
- 30:29SPEAKER_00: Yeah, that's great.
- 30:30SPEAKER_00: That's some really great actionable advice,
- 30:32SPEAKER_00: especially that around asking your vendors what they're using.
- 30:36SPEAKER_00: I suppose with the attackers of using the AI and they're getting faster and we're on
- 30:40SPEAKER_00: the defense side,
- 30:41SPEAKER_00: it feels sometimes like we're also using it and just staying in the same place.
- 30:45SPEAKER_00: So yeah,
- 30:46SPEAKER_00: it's good to know that there are steps that we can take to try and get
- 30:50SPEAKER_00: that little bit more ahead as well.
- 30:52SPEAKER_00: Thank you so much for joining me today, Steve.
- 30:54SPEAKER_00: I've really, really enjoyed this conversation and it's been great to have you on the show.
- 30:58SPEAKER_01: You're welcome.
- 30:59SPEAKER_01: It's great to be here, Caitlin.
- 31:00SPEAKER_01: Thank you for having me.