Latest / Expert Insights Podcast / #90. AI Agents, Zero Trust & Quantum Threats: Zenith Live Takeaways (Part 1)
Transcript
- 0:08SPEAKER_00: Hello and welcome to the Expert Insights Podcast.
- 0:10SPEAKER_00: I'm Mary McDade,
- 0:11SPEAKER_00: senior reporter here,
- 0:12SPEAKER_00: and over the last week I've been
- 0:13SPEAKER_00: in Vienna attending Zenith Live Zscaler's annual conference.
- 0:17SPEAKER_00: This week has included a series of keynotes,
- 0:20SPEAKER_00: interviews of cybersecurity professionals,
- 0:22SPEAKER_00: and connecting with innovators and thought leaders from across the world.
- 0:27SPEAKER_00: In this episode, we're featuring two of the interviews from the conference.
- 0:30SPEAKER_00: The first is from Swami Kocha Lakota, Executive Vice President, Agentic AI Startup at Zscaler.
- 0:37SPEAKER_00: And the other from Yaroslav Rosomako, Chief Scientist at Zscaler.
- 0:42SPEAKER_00: Swami talks us through AI security incidents that enterprises are dealing
- 0:46SPEAKER_00: with right now and why Zero Trust is the right architectural response
- 0:50SPEAKER_00: and how AI coding tools are opening new attack vectors for developers.
- 0:54SPEAKER_00: Let's get into it.
- 0:57SPEAKER_00: From Zscaler's vantage point and from your own conversations that you've had
- 1:01SPEAKER_00: with CISOs,
- 1:03SPEAKER_00: what are the most common AI related security incidents that enterprises are dealing
- 1:07SPEAKER_00: with right now?
- 1:09SPEAKER_03: They range from,
- 1:11SPEAKER_03: you know, I'm worried about what type of a data that people are putting in there.
- 1:18SPEAKER_03: Am I going to be compliant with my regulatory obligations all the way
- 1:23SPEAKER_03: to what my employees are using AI so that some code will run on it.
- 1:29SPEAKER_03: So all of those things our CISOs are concerned.
- 1:34SPEAKER_03: I mean, if you look at the internet,
- 1:36SPEAKER_03: there's so many episodes that have where Cloud went and deleted things,
- 1:41SPEAKER_03: open AI,
- 1:42SPEAKER_03: you know,
- 1:43SPEAKER_03: leaked utensils,
- 1:44SPEAKER_03: you know,
- 1:45SPEAKER_03: just those are the type of attacks that CISOs are trying
- 1:48SPEAKER_03: to protect their enterprises from proactively.
- 1:51SPEAKER_00: How do you draw the line between AI that's a productivity tool and AI that's a security risk?
- 1:58SPEAKER_00: And where does Zero Trust fit into making that distinction?
- 2:02SPEAKER_03: See, with technology, security can be like an afterthought.
- 2:09SPEAKER_03: You have to do secure AI from the beginning.
- 2:14SPEAKER_03: You cannot think about productivity without really thinking about security.
- 2:18SPEAKER_03: You have to do both.
- 2:20SPEAKER_03: and that's why having a good architecture like the zero trust everywhere
- 2:25SPEAKER_03: which is actually zero trust is a philosophy it's an architecture and it's
- 2:30SPEAKER_03: an implementation all together otherwise there are lots of philosophies
- 2:34SPEAKER_03: and principles but they can never be implemented in enterprises like principle
- 2:40SPEAKER_03: of least privilege or backs these are all things existed in enterprises before
- 2:45SPEAKER_03: but zero trust is the only one
- 2:48SPEAKER_03: that you can actually have a philosophy, principle, and an architecture and implement it.
- 2:53SPEAKER_03: And Zero Trust Everywhere for agents,
- 2:55SPEAKER_03: which is what we launched, allows you to say for AI, who are you?
- 3:00SPEAKER_03: What are you trying to do?
- 3:01SPEAKER_03: What is your intent?
- 3:03SPEAKER_03: Does the policy allow you to do it or not is part of the Zero Trust Architecture for agents.
- 3:09SPEAKER_00: That sounds good.
- 3:10SPEAKER_00: A lot of CISOs that I've spoken to have said that is a concern.
- 3:13SPEAKER_00: Unchecked AI is a concern for them.
- 3:16SPEAKER_03: Right.
- 3:18SPEAKER_00: So attackers are weaponizing AI coding tools as infection vectors.
- 3:24SPEAKER_00: How should organizations secure the developer tool chain without killing productivity?
- 3:29SPEAKER_03: yeah so cyber is a war attackers are using AI and then defenders have to use AI
- 3:38SPEAKER_03: the first and foremost thing that I would say is that defenders also have to be open
- 3:44SPEAKER_03: to using AI to fight it right so now with respect to code injection techniques
- 3:51SPEAKER_03: without compromising on the productivity
- 3:54SPEAKER_03: You know, most of my developers who are using AI are spending time reading the code, right?
- 4:01SPEAKER_03: The 80% they build with AI,
- 4:03SPEAKER_03: but the 20% they're spending is a little fine-tuning,
- 4:05SPEAKER_03: but most of the time they're reading the code for AI.
- 4:08SPEAKER_03: I think the best way to protect these type of malware to come in through
- 4:13SPEAKER_03: The code generation agents is to make sure that you're testing and you have
- 4:16SPEAKER_03: the evaluation and you have that LLM as a judge where you're validating
- 4:21SPEAKER_03: the code that's written by one model with another model Excellent.
- 4:26SPEAKER_00: What does Zscaler mean by agentic AI specifically and how is the approach different
- 4:32SPEAKER_00: from say a year ago?
- 4:34SPEAKER_03: Yeah.
- 4:34SPEAKER_03: So first and foremost, agents is a spectrum.
- 4:39SPEAKER_03: So one end of that spectrum is if you go to a webpage and you have
- 4:46SPEAKER_03: a particular agent, you say, hey, I can summarize
- 4:49SPEAKER_03: Alex alex190399 Expert
- 5:04SPEAKER_03: right hand side of the spectrum is autonomous long lived agents
- 5:08SPEAKER_03: that have either your entitlements or some other unique entitlements
- 5:14SPEAKER_03: that are relevant to that particular agent right so that's what we mean by agent
- 5:18SPEAKER_03: and you have to protect
- 5:20SPEAKER_03: all forms of agents in the spectrum and zero trust everywhere.
- 5:25SPEAKER_03: The way we have built the policy definition and the way we are building
- 5:30SPEAKER_03: the policy enforcement allows you to protect it throughout the spectrum.
- 5:35SPEAKER_00: Excellent.
- 5:37SPEAKER_00: Can you walk me through a concrete example of something
- 5:42SPEAKER_00: An enterprise can defend against today with agentic controls that you've launched
- 5:47SPEAKER_00: that they had no way of handling say a year ago Yeah,
- 5:51SPEAKER_03: so basically let's say you are an enterprise and
- 5:58SPEAKER_03: to serve your customers you launched an agent like a customer service agent
- 6:03SPEAKER_03: which is one of the most popular use cases that we have
- 6:06SPEAKER_03: You build that agent,
- 6:09SPEAKER_03: but you have to make sure that agent is behaving exactly like
- 6:13SPEAKER_03: the way you would want it to be.
- 6:17SPEAKER_03: If the agent was done without AI, you don't
- 6:21SPEAKER_03: You don't need to worry about testing because it's all deterministic technology.
- 6:26SPEAKER_03: So here,
- 6:27SPEAKER_03: because AI is probabilistic,
- 6:29SPEAKER_03: sometimes you can ask the same question to the agent,
- 6:32SPEAKER_03: it may give two different answers.
- 6:34SPEAKER_03: So you got to do this continuous red teaming where making sure
- 6:38SPEAKER_03: that your agent is looking at the right security, the safety,
- 6:42SPEAKER_03: and toxicity and also meeting your business constraints that you would have
- 6:49SPEAKER_03: for example you don't want your agent if you're an electric car company like Tesla
- 6:54SPEAKER_03: to answer questions about a Lucid Motors right because agents if you take it out of
- 7:00SPEAKER_03: the box they know about Tesla they know about Lucid they know about all
- 7:03SPEAKER_03: the other electric cars as well but when you put that agent out there you would want
- 7:08SPEAKER_03: to put constraints around
- 7:09SPEAKER_03: Don't ask questions.
- 7:10SPEAKER_03: Don't ask questions about this.
- 7:12SPEAKER_03: Only ask questions about this.
- 7:14SPEAKER_03: You could do all of that.
- 7:18SPEAKER_00: What's the biggest technical challenge in applying Zero Trust to AI workloads
- 7:22SPEAKER_00: and what makes it harder than traditional application security?
- 7:26SPEAKER_03: It's the architecture of having a single policy engine that you can apply across the board.
- 7:34SPEAKER_03: If you look at IT assets in an enterprise, you have users and you have software that the
- 7:41SPEAKER_03: users are using from the devices that a user is using and then you have workloads
- 7:47SPEAKER_03: and the workloads are either on your data center or on the cloud right and
- 7:53SPEAKER_03: then your users are located in branches
- 7:56SPEAKER_03: So all of the zero trust everywhere allows that to be guided by that one policy.
- 8:04SPEAKER_03: That's the unique thing about Z-Scale.
- 8:07SPEAKER_03: It's a single policy engine that you can enforce all your IT interactions at
- 8:12SPEAKER_03: the network level at wide speed.
- 8:14SPEAKER_00: Right.
- 8:15SPEAKER_00: Across the boards.
- 8:16SPEAKER_03: Exactly.
- 8:17SPEAKER_00: Excellent.
- 8:17SPEAKER_00: If a CISO hasn't started asking about AI security governance,
- 8:21SPEAKER_00: what is the first thing they should do tomorrow morning?
- 8:24SPEAKER_03: The first thing is they need to know whether how their employees are using AI.
- 8:32SPEAKER_03: Visibility is an important aspect of it.
- 8:35SPEAKER_03: If you are before protecting any of our personal belongings at home,
- 8:40SPEAKER_03: we need to have an inventory of what we have.
- 8:42SPEAKER_03: So the first question
- 8:43SPEAKER_03: The question that CISO should ask is that,
- 8:45SPEAKER_03: do I have a complete visibility and understanding of how AI is being used?
- 8:52SPEAKER_03: What AI components are deployed on the endpoint?
- 8:56SPEAKER_03: What MCP servers my employees are accessing?
- 8:59SPEAKER_03: what AI agents are asking so
- 9:00SPEAKER_03: that entire discovery is what we have our entire AI protect suite helps a CISO
- 9:06SPEAKER_03: through the journey number one get access to the discovery of where the AI usage is
- 9:13SPEAKER_03: then secure it by having a policy are you one of those enterprises where you want
- 9:19SPEAKER_03: to allow AI and deny where you cannot do for example don't go to deep seek
- 9:25SPEAKER_03: or are you an enterprise where say no first
- 9:28SPEAKER_03: but allow chat GPT, cloud and perplexity, right?
- 9:32SPEAKER_03: You could do that.
- 9:33SPEAKER_03: And then the third pillar is you could go deep.
- 9:35SPEAKER_03: You know,
- 9:36SPEAKER_03: not only I would want to allow this AI,
- 9:38SPEAKER_03: but I want to see what conversations my employees are having and then
- 9:43SPEAKER_03: Alex alex190 Expert Insights Podcast
- 10:03SPEAKER_03: having an understanding of where your data is who has access to it
- 10:10SPEAKER_03: which humans have access to it and which non-humans have access to it and
- 10:14SPEAKER_03: the whole access graph technology that we got not only from a static discovery
- 10:19SPEAKER_03: but runtime actual usage as well is an important element
- 10:24SPEAKER_03: because you know CISOs now can apply adaptive access to the agents at that time
- 10:29SPEAKER_00: What's the AI security problem that no one is talking about yet,
- 10:32SPEAKER_00: but there will be in 12 months?
- 10:34SPEAKER_03: I think interagent communication.
- 10:37SPEAKER_03: Today we are seeing a lot of MCP traffic.
- 10:41SPEAKER_03: We will see a lot more agent-to-agent communication using the A2A protocol.
- 10:47SPEAKER_03: And that's part of our roadmap is not only provide a broker,
- 10:52SPEAKER_03: AI broker that supports MCP protocol,
- 10:54SPEAKER_03: but that will also support A2A protocol also and govern that protocol.
- 10:59SPEAKER_00: Excellent.
- 10:59SPEAKER_00: So beyond kind of headline adoption figures,
- 11:03SPEAKER_00: is there a specific security number,
- 11:05SPEAKER_00: say threats blocked,
- 11:07SPEAKER_00: malicious agent actions stopped,
- 11:09SPEAKER_00: prompts flagged, that show that this is in production and not just roadmaps?
- 11:14SPEAKER_03: Yeah, we have lots of statistics.
- 11:17SPEAKER_03: For example,
- 11:18SPEAKER_03: we shared that there are about 750 billion transactions that we are seeing
- 11:24SPEAKER_03: and I exactly know how many of those 750 billion transactions are emceeds.
- 11:30SPEAKER_03: And I exactly know what are the top 50 widely used AI assets by our customers.
- 11:41SPEAKER_03: And we also provide custom analytics to each of our customers.
- 11:44SPEAKER_03: Alex alex190 Expert Insights Podcast
- 11:59SPEAKER_03: you're listening to the Expert Insights Podcast home
- 12:02SPEAKER_03: of leading cybersecurity conversations subscribe now to get insights
- 12:04SPEAKER_03: from top cybersecurity leaders straight to your inbox
- 12:32SPEAKER_00: Thank you to Swami for sharing those thoughts with us.
- 12:35SPEAKER_00: And now on to Yaroslav,
- 12:36SPEAKER_00: Chief Scientist at CScaler,
- 12:38SPEAKER_00: who starts off with his takeout on agentic AI before shifting
- 12:41SPEAKER_00: into post-quantum cryptography and what practical steps you can take today.
- 12:46SPEAKER_00: What's the most interesting problem that you're working on right now at
- 12:49SPEAKER_00: the intersection of AI and security?
- 12:53SPEAKER_01: It's very hard to pin that down to just an individual problem.
- 12:57SPEAKER_01: I would say that controlling of agentic AI is a
- 13:05SPEAKER_01: Alex alex190399
- 13:20SPEAKER_01: because agentic AI is actively evolving.
- 13:23SPEAKER_01: So we're seeing new kinds of protocols.
- 13:25SPEAKER_01: We're seeing new kinds of behavioral patterns.
- 13:29SPEAKER_01: Companies such as Anthropic and OpenAI and others, they release frontier models
- 13:35SPEAKER_01: every other week.
- 13:36SPEAKER_01: Well, maybe not every other week,
- 13:37SPEAKER_01: but quite frequently, and we see significant changes in behavior.
- 13:42SPEAKER_01: So making sure that the necessary guardrails controls are in place,
- 13:46SPEAKER_01: the necessary inspections help secure agentic AI, but also not get in the way.
- 13:53SPEAKER_01: Allow organizations to get the benefits of their investment into the agentic AI.
- 14:00SPEAKER_00: How do you weigh the tradeoff between AI model capability and security risk
- 14:05SPEAKER_00: and is there a point where more capable models are becoming harder to secure
- 14:09SPEAKER_00: and what does that mean for how a CISO should defend with them?
- 14:14SPEAKER_01: Yeah, that's of course an extremely hard problem for which I don't think there is
- 14:20SPEAKER_01: a single silver bullet solution that would apply
- 14:23SPEAKER_01: to all organizations Different organizations have different risk appetite They have
- 14:29SPEAKER_01: different sensitivity of data that they work with They are on a different level
- 14:34SPEAKER_01: of cybersecurity maturity So some are very mature and ready to take on
- 14:40SPEAKER_01: the next challenge Others have to do a lot of homework
- 14:43SPEAKER_01: for example, deploy Zero Trust architecture before they can fully embrace the Agenda KI.
- 14:50SPEAKER_01: So yes, there is really no one-size-fits-all,
- 14:53SPEAKER_01: but one common theme that I see across multiple organizations
- 14:58SPEAKER_01: Alex alex190399 Expert
- 15:12SPEAKER_01: Alex alex190399 Expert Insights Podcast
- 15:35SPEAKER_01: Alex alex190399 Expert
- 16:04SPEAKER_01: Okay.
- 16:05SPEAKER_01: Well, first of all,
- 16:07SPEAKER_01: what we already see,
- 16:09SPEAKER_01: adoption of agentic AI really highlights importance of Zero Trust architecture.
- 16:15SPEAKER_01: So many organizations who were spinning the wheels a little bit, being slow at
- 16:21SPEAKER_01: changing from traditional routed or firewalled architectures
- 16:27SPEAKER_01: into proper Zero Trust now accelerate their plans,
- 16:31SPEAKER_01: they're adopting Zero Trust much faster.
- 16:35SPEAKER_01: So we see
- 16:36SPEAKER_01: certainly a massive acceleration of an increase in interest in zero trust
- 16:43SPEAKER_01: that is driven by AI use cases.
- 16:47SPEAKER_01: Organizations see that their agentic AI can be compromised,
- 16:50SPEAKER_01: it can be tricked,
- 16:53SPEAKER_01: it sometimes could be misbehaving, so applying necessary guardrails is important.
- 16:58SPEAKER_01: The second area where AI really helps Zero Trust is improving controls.
- 17:06SPEAKER_01: It's an additional level of controls that we can add on top
- 17:09SPEAKER_01: of what we have been already building.
- 17:11SPEAKER_01: So, of course, the
- 17:13SPEAKER_01: AI security capabilities,
- 17:15SPEAKER_01: they do not invalidate antivirus and IPS and sandboxes,
- 17:20SPEAKER_01: access controls, all the things that the industry has been building for decades.
- 17:25SPEAKER_01: They add an additional opportunity to detect malware,
- 17:29SPEAKER_01: detect wrong behavior, detect any kind of data exfiltration attempts.
- 17:34SPEAKER_01: So they're really adding
- 17:37SPEAKER_01: Consistency to additional level of consistency to what we have been doing in
- 17:42SPEAKER_01: the past and then finally agentic AI when it comes
- 17:47SPEAKER_01: to zero trust adds new opportunities to bring new services to customers
- 17:54SPEAKER_01: that make them more secure so for example red teaming traditionally
- 18:00SPEAKER_01: Only most sophisticated organizations could afford hiring a team
- 18:05SPEAKER_01: that would prevent vulnerability scanning,
- 18:07SPEAKER_01: that would try to break their defenses, outsource that or do that in-house.
- 18:12SPEAKER_01: That requires significant budget,
- 18:14SPEAKER_01: that requires very sophisticated cybersecurity professionals who would be able to do
- 18:19SPEAKER_01: that.
- 18:19SPEAKER_01: But with advancements in AI technology,
- 18:23SPEAKER_01: this allows to reduce the barrier so to speak so to help automate many of the things
- 18:31SPEAKER_01: that red teaming is doing so this is one of the red teaming addition to zero trust
- 18:36SPEAKER_00: that we have brought recently nice yeah makes a lot of sense
- 18:40SPEAKER_00: so your Wednesday breakout covers quantum threats for CISOs
- 18:44SPEAKER_00: and IT directors how urgent is this a five-year program or
- 18:48SPEAKER_01: a two-year program well there are two
- 18:53SPEAKER_01: areas when it comes to quantum threat.
- 18:57SPEAKER_01: One that is so-called harvest now decrypt later.
- 19:00SPEAKER_01: That is communication that we use today could potentially be intercepted
- 19:05SPEAKER_01: by bad actors passively and they would be able in the future to replay
- 19:13SPEAKER_01: that communication through quantum computer
- 19:15SPEAKER_01: and get whatever confidential data was stored there.
- 19:19SPEAKER_01: This part is imminent.
- 19:22SPEAKER_01: So organizations need to migrate to post-quantum cryptography when it comes
- 19:29SPEAKER_01: to key exchange as soon as possible.
- 19:32SPEAKER_01: The good news about that element
- 19:36SPEAKER_01: Standards are already in place.
- 19:38SPEAKER_01: They're relatively mature.
- 19:40SPEAKER_01: And we already have, for example, modern browsers, Chrome, Firefox, Safari already support them.
- 19:45SPEAKER_01: This is something that we're rolling out now on our Zero Trust Exchange platform.
- 19:51SPEAKER_01: So I would highly encourage CISOs and all the cybersecurity practitioners not
- 19:58SPEAKER_01: to delay this migration to migrate today or as soon as possible.
- 20:03SPEAKER_01: This is not a matter of two or five
- 20:06SPEAKER_01: A year program, this is a matter of urgency that we need to begin as soon as possible.
- 20:10SPEAKER_01: The other element is digital signatures.
- 20:14SPEAKER_01: So certificates or any kind of tokens, YubiKeys, TPMs, it's a very wide range of fields.
- 20:21SPEAKER_01: So digital identity is typically confirmed with a cryptographic signature such
- 20:27SPEAKER_01: as RSA, ECDSA or EDDSA.
- 20:30SPEAKER_01: And those signatures are also vulnerable to quantum computers.
- 20:33SPEAKER_01: The good news is
- 20:35SPEAKER_01: There is no harvest now decrypt later attack factor here.
- 20:38SPEAKER_01: So if you compromise an identity associated with certificate that's already expired,
- 20:45SPEAKER_01: there is no retrospective attack factor here.
- 20:47SPEAKER_01: But the bad news is the migration is much more complex.
- 20:51SPEAKER_01: So even though we don't expect cryptographically relevant quantum computers
- 20:55SPEAKER_01: to arrive for another three years maybe organizations need
- 20:59SPEAKER_01: to start planning migration to post-quantum digital signatures today because
- 21:04SPEAKER_01: that is much more complex and sophisticated process and you need to be prepared for
- 21:09SPEAKER_01: that migration not leave it to the very last minute Makes a lot of sense
- 21:14SPEAKER_00: So what are the practical steps that organizations should take to prepare
- 21:18SPEAKER_00: for post-quantum cryptography even if the threat feels distant?
- 21:22SPEAKER_01: Right, right.
- 21:22SPEAKER_01: So first of all, visibility.
- 21:25SPEAKER_01: I would say that is the foundational building block.
- 21:29SPEAKER_01: So understand what's your exposure.
- 21:31SPEAKER_01: So that is where you have legacy applications,
- 21:36SPEAKER_01: legacy appliances, where you communicate with some
- 21:40SPEAKER_01: content providers over protocols or cyber or cryptographic stacks
- 21:49SPEAKER_01: that do not support PQC and prioritize the migration so figure out
- 21:55SPEAKER_01: over which communication channels you have your most critical data,
- 21:59SPEAKER_01: the one that you really care about in terms of confidentiality,
- 22:03SPEAKER_01: the one that is likely to stay sensitive in three,
- 22:06SPEAKER_01: five years from now, so the one that you cannot lose under any circumstances.
- 22:10SPEAKER_01: And prioritize migration on those channels and do protection
- 22:14SPEAKER_01: against harvest now decrypt later attacks first.
- 22:17SPEAKER_01: It's a relatively straightforward thing to do.
- 22:20SPEAKER_01: And start planning migration for post-quantum digital signatures
- 22:25SPEAKER_01: Alex alex190399
- 22:44SPEAKER_01: Oh, it is very practical.
- 22:45SPEAKER_01: Yes, we see many organizations investing into crypto agility.
- 22:52SPEAKER_01: So that is making sure that they don't just treat it as a one-time exercise moving
- 22:58SPEAKER_01: from an old classical algorithm to new shiny post-quantum algorithms.
- 23:04SPEAKER_01: In cryptographic world,
- 23:06SPEAKER_01: Exciting is a very bad word.
- 23:08SPEAKER_01: You really want boring because exciting means that there is a chance
- 23:12SPEAKER_01: that there will be a kind of flaw that we don't yet know about discovered
- 23:16SPEAKER_01: in those algorithms.
- 23:18SPEAKER_01: So there is a chance that those algorithms that we migrate to today maybe
- 23:23SPEAKER_01: in future somebody will discover vulnerability in them or
- 23:27SPEAKER_01: Alex alex190399 Expert Insights Podcast
- 23:57SPEAKER_00: great excellent um I have a few more optional questions you never know exactly how
- 24:03SPEAKER_00: the time will go I'm just being cautious of your time I'm fine no no fine
- 24:07SPEAKER_02: with me yeah sure yeah it looks like we have about eight minutes great probably time
- 24:11SPEAKER_00: for at least one or two um let's see okay um
- 24:18SPEAKER_00: You explain this stuff for a living.
- 24:21SPEAKER_00: What's the analogy you would use to make quantum risk land with, say, a non-technical executive?
- 24:35SPEAKER_00: What happens to an organization that chooses not to do anything about this for
- 24:39SPEAKER_00: the next three years?
- 24:42SPEAKER_01: Well, there are a couple of risks that organization will expose itself to.
- 24:47SPEAKER_01: So, first of all, as I mentioned, harvest now, decrypt later risk.
- 24:50SPEAKER_01: Whatever you transmit today over the internet or over any other channel
- 24:55SPEAKER_01: that could potentially be
- 24:57SPEAKER_01: passively monitored.
- 24:58SPEAKER_01: So it could be something as simple as you're sitting in a cafe, right, on a public Wi-Fi.
- 25:03SPEAKER_01: Of course,
- 25:04SPEAKER_01: you're secured by ZTNA or something,
- 25:06SPEAKER_01: but browsing over HTTPS websites,
- 25:10SPEAKER_01: but still somebody sitting next
- 25:11SPEAKER_01: to you could potentially passively monitor those packets and in the future,
- 25:15SPEAKER_01: once Quantum Glitter Selva will decrypt that and get
- 25:18SPEAKER_01: potentially to confidential information that would be available it could be somebody
- 25:24SPEAKER_01: in the service provider it could be
- 25:25SPEAKER_01: a nation state actor potentially redirecting traffic rerouting traffic in a way
- 25:29SPEAKER_01: that they would be able to see encrypted packets today
- 25:33SPEAKER_01: so confidentiality future breakage of confidentiality is one issue the second one is
- 25:41SPEAKER_01: when the time comes
- 25:43SPEAKER_01: They, instead of carefully planning, they might end up with a fire drill.
- 25:47SPEAKER_01: So we need to immediately replace all our digital signatures,
- 25:50SPEAKER_01: replace PKI, replace lots of equipment.
- 25:53SPEAKER_01: So that will be very costly, extremely disruptive for organization.
- 25:57SPEAKER_01: And third one
- 25:58SPEAKER_01: is regulatory compliance.
- 26:00SPEAKER_01: Many industries are regulated.
- 26:05SPEAKER_01: In some cases, regulators provide guidance and recommendations.
- 26:09SPEAKER_01: In other, they provide strict requirements for organizations to stay compliant.
- 26:14SPEAKER_01: And of course,
- 26:16SPEAKER_01: Risk of miscompliance carries a significant financial burden so organizations
- 26:21SPEAKER_01: that might choose not to migrate to post-quantum cryptography
- 26:25SPEAKER_01: because they might not believe in quantum computers they might expose themselves
- 26:30SPEAKER_01: in addition to concrete cybersecurity risks to also potentially legal risks
- 26:36SPEAKER_01: or associated with miscompliance
- 26:39SPEAKER_00: that's probably the one in your quantum breakout tomorrow you're talking
- 26:45SPEAKER_00: to you're going to be talking to a room full of security leaders what is
- 26:48SPEAKER_00: the biggest misconception about quantum risk that you're hoping
- 26:51SPEAKER_01: to correct well quite a few actually first one is that PQC migration is
- 27:03SPEAKER_01: a singular transformation so first of all
- 27:09SPEAKER_01: As I explained previously,
- 27:11SPEAKER_01: we really should be separating key exchange and digital signature problem.
- 27:15SPEAKER_01: They are very separate.
- 27:18SPEAKER_01: They're both affected by quantum computing.
- 27:21SPEAKER_01: One is more urgent than another.
- 27:23SPEAKER_01: One is simpler than the other.
- 27:26SPEAKER_01: When cybersecurity professionals are planning migration,
- 27:29SPEAKER_01: they really should be structuring it and separating them instead of trying
- 27:33SPEAKER_01: and doing it all at once as one heroic effort.
- 27:39SPEAKER_01: Secondly, one common misconception is, again, that that's a one-time thing.
- 27:44SPEAKER_01: So we migrate from old legacy mechanisms to new shiny,
- 27:48SPEAKER_01: exciting mechanisms, and that's it, the job done.
- 27:50SPEAKER_01: So embracing
- 27:52SPEAKER_01: Cryptographic Agility,
- 27:53SPEAKER_01: so be able in the future to migrate to something else if we have to,
- 27:57SPEAKER_01: is another important part of the project.
- 28:00SPEAKER_01: And then another misconception that I keep hearing sometimes is
- 28:04SPEAKER_01: that we cannot protect from quantum threats without quantum computers.
- 28:08SPEAKER_01: That's just incorrect.
- 28:10SPEAKER_01: So post-quantum cryptography,
- 28:12SPEAKER_01: so that is those new algorithms that are equally secure from classical computers
- 28:17SPEAKER_01: and quantum computers, they are
- 28:19SPEAKER_01: Thank you to Jaroslav for speaking with us.
- 28:47SPEAKER_00: that's all for this episode but it's not all from
- 28:49SPEAKER_00: the conference we've got another episode coming featuring exclusive interviews
- 28:53SPEAKER_00: from Zenith Live so please subscribe so you don't miss out you can read all
- 28:58SPEAKER_00: the articles I wrote whilst at the conference
- 28:59SPEAKER_00: at expertinsights.com see you next time