Latest / Connectivity Without Chaos - The Podcast / **Your AI Agents Have Access. But Who's Watching Them? 6 Rules for Agentic AI Security**

Transcript

**Your AI Agents Have Access. But Who's Watching Them? 6 Rules for Agentic AI Security** · Connectivity Without Chaos - The Podcast

• 0:00
  SPEAKER_00: So get this, hackers are lurking in your network for an average of 24 days, right?
• 0:05
  SPEAKER_00: Just mapping out every single vulnerability.
• 0:08
  SPEAKER_00: But the moment they actually pull the trigger on a ransomware attack...
• 0:11
  SPEAKER_01: That's the wild part, yeah.
• 0:13
  SPEAKER_00: It takes just 22 seconds now.
• 0:16
  SPEAKER_00: Welcome to the Deep Dive.
• 0:17
  SPEAKER_00: Today, we are tearing into this new report from SACR, that's Software Analysts Cyber Research.
• 0:23
  SPEAKER_00: And it's all about ephemeral ransomware resilience, or ERR.
• 0:28
  SPEAKER_00: The mission here is
• 0:28
  SPEAKER_00: to figure out why traditional backups are just completely failing
• 0:32
  SPEAKER_00: against modern ransomware and how you can actually survive an attack even
• 0:36
  SPEAKER_00: if your admin credentials get stolen.
• 0:38
  SPEAKER_01: Yeah, and that 22-second execution window is what makes this so terrifying.
• 0:42
  SPEAKER_01: I mean, a human-led response is mathematically impossible at that speed.
• 0:46
  SPEAKER_00: Oh, for sure.
• 0:47
  SPEAKER_01: By the time your little monitoring alert fires off, your data's already locked.
• 0:51
  SPEAKER_01: We're officially in what they call phase three of ransomware evolution.
• 0:55
  SPEAKER_01: Phase three.
• 0:56
  SPEAKER_01: Right.
• 0:56
  SPEAKER_01: Because the playbook isn't just about sneaking in to encrypt production data anymore.
• 1:00
  SPEAKER_01: It's about hunting down admin credentials.
• 1:02
  SPEAKER_01: They want to log into your management console and wipe out your backups first.
• 1:06
  SPEAKER_00: Before they even drop the ransom note.
• 1:08
  SPEAKER_01: Exactly.
• 1:09
  SPEAKER_01: All before you even know they're there.
• 1:11
  SPEAKER_00: See, this completely breaks the old logic of like storage immutability.
• 1:15
  SPEAKER_00: It's like building this massive titanium bank vault with 10-foot steel walls
• 1:20
  SPEAKER_00: to protect your assets, right?
• 1:21
  SPEAKER_00: But then you just leave the manager's master override switch sitting out in the public lobby.
• 1:26
  SPEAKER_01: That is a great way to put it.
• 1:28
  SPEAKER_00: If an attacker already has the admin credentials,
• 1:31
  SPEAKER_00: can't they just use your own API to override the lock and delete the backups anyway?
• 1:37
  SPEAKER_01: Yes, that is the fatal flaw in what Seeker calls the adversarial recovery test.
• 1:42
  SPEAKER_01: Legacy defenses like air gap tapes or offsite replication or cloud immunability.
• 1:48
  SPEAKER_00: The stuff everyone uses.
• 1:49
  SPEAKER_01: Right, they do a fantastic job of securing the storage medium itself,
• 1:53
  SPEAKER_01: but they leave the access path entirely wide open.
• 1:57
  SPEAKER_01: If that management console is reachable on your network and the attacker holds
• 2:00
  SPEAKER_01: the keys...
• 2:01
  SPEAKER_00: Game over.
• 2:02
  SPEAKER_01: Your safety net is gone.
• 2:03
  SPEAKER_00: Okay, wait, I'm confused though.
• 2:05
  SPEAKER_00: If the very tool we use to manage the backups is the vulnerability,
• 2:10
  SPEAKER_00: how on earth do you build a network where even the admins don't have access?
• 2:15
  SPEAKER_00: How do you configure a backup if there's literally no console or API?
• 2:18
  SPEAKER_01: It sounds super counterintuitive,
• 2:20
  SPEAKER_01: but that paradigm shift is exactly what ephemeral ransomware resilience,
• 2:25
  SPEAKER_01: ERR, is built to solve.
• 2:26
  SPEAKER_00: Okay.
• 2:27
  SPEAKER_01: And it starts with its first pillar, which is management plane isolation.
• 2:31
  SPEAKER_01: So the architecture fundamentally changes to remove that attack surface.
• 2:35
  SPEAKER_00: How so?
• 2:36
  SPEAKER_01: Well, instead of the primary network pushing data to a backup target via an API,
• 2:41
  SPEAKER_01: the backup environment uses a pull-based mechanism.
• 2:44
  SPEAKER_01: It wakes up, reaches out to grab the data, and then immediately severs the connection.
• 2:49
  SPEAKER_00: Oh, wow.
• 2:50
  SPEAKER_00: So there's no inbound API?
• 2:52
  SPEAKER_01: No inbound API,
• 2:53
  SPEAKER_01: no console,
• 2:54
  SPEAKER_01: no command line interface for an attacker to exploit,
• 2:58
  SPEAKER_01: even if they log in with perfectly valid admin credentials.
• 3:00
  SPEAKER_00: Okay, so it's basically a one-way street.
• 3:03
  SPEAKER_00: That makes sense mechanically.
• 3:05
  SPEAKER_00: What's the second pillar then?
• 3:06
  SPEAKER_01: Automated continuous protection.
• 3:09
  SPEAKER_01: Because, you know, human intervention introduces risk during a crisis.
• 3:12
  SPEAKER_00: Oh, yeah.
• 3:13
  SPEAKER_00: People panic.
• 3:14
  SPEAKER_01: Exactly.
• 3:14
  SPEAKER_01: So ERR systems generate backup copies hourly, completely automatically.
• 3:20
  SPEAKER_01: No operator needs to trigger it, which removes the chance for manual failure.
• 3:24
  SPEAKER_00: Which leads right into the third pillar, ephemeral compute continuity.
• 3:29
  SPEAKER_00: And if you're managing an IT environment today, this is your ultimate failsafe.
• 3:33
  SPEAKER_00: It basically means having dedicated standby cloud environments ready to spin up
• 3:37
  SPEAKER_00: and run your business operations for like 30 days or more,
• 3:41
  SPEAKER_00: even if your primary infrastructure is just totally destroyed.
• 3:45
  SPEAKER_00: It completely decouples your ability to do business from the health
• 3:48
  SPEAKER_00: of your infected primary network.
• 3:51
  SPEAKER_01: Right.
• 3:51
  SPEAKER_01: But that introduces a massive operational hazard.
• 3:54
  SPEAKER_00: Wait, really?
• 3:55
  SPEAKER_00: What do you mean?
• 3:56
  SPEAKER_01: Well,
• 3:56
  SPEAKER_01: having an isolated backup and a standby cloud environment is totally useless if
• 4:01
  SPEAKER_01: the backup you restore from is already infected.
• 4:03
  SPEAKER_00: Oh.
• 4:04
  SPEAKER_01: Remember those 24 days the attacker was lurking?
• 4:07
  SPEAKER_01: They probably dropped dormant malware into those backups.
• 4:10
  SPEAKER_00: Ah, so that's the whole automated recovery testing versus clean room recovery debate?
• 4:15
  SPEAKER_00: Because most IT teams just run ART, right?
• 4:18
  SPEAKER_00: Automated recovery testing, which,
• 4:20
  SPEAKER_00: correct me if I'm wrong, that just proves the backup image can successfully boot up.
• 4:23
  SPEAKER_01: Yes.
• 4:24
  SPEAKER_01: But booting a compromised backup in your fresh cloud environment just re-triggers the attack.
• 4:29
  SPEAKER_00: You're just reinfecting yourself.
• 4:31
  SPEAKER_01: Exactly.
• 4:31
  SPEAKER_01: Which is why clean room recovery, or CRR, is absolutely mandatory here.
• 4:36
  SPEAKER_01: It doesn't just boot the backup, it restores it into a forensically sterile environment.
• 4:40
  SPEAKER_01: Okay.
• 4:41
  SPEAKER_01: It actively scans the file systems,
• 4:43
  SPEAKER_01: compares hashes against threat databases,
• 4:45
  SPEAKER_01: and systematically scrubs any of those dormant payloads.
• 4:48
  SPEAKER_00: So it sanitizes the data layer before it ever touches your production networks again.
• 4:51
  SPEAKER_01: 100%.
• 4:54
  SPEAKER_00: Understanding this shift from reactive storage to proactive architectural isolation,
• 5:00
  SPEAKER_00: it's absolutely vital for surviving today.
• 5:03
  SPEAKER_00: You really just have to assume attackers already have the keys to your castle.
• 5:06
  SPEAKER_01: You do.
• 5:07
  SPEAKER_01: The perimeter will inevitably fall.
• 5:09
  SPEAKER_01: True resilience is about surviving the breach, not just trying to prevent it.
• 5:14
  SPEAKER_00: Which leaves us with a pretty provocative thought to ponder on your own.
• 5:18
  SPEAKER_00: If ERR requires completely severing the management plane from the backup storage,
• 5:23
  SPEAKER_00: How does the day-to-day job of a traditional IT administrator change
• 5:26
  SPEAKER_00: when they are physically locked out of their own safety nets?
• 5:30
  SPEAKER_01: It's a scary thought.
• 5:31
  SPEAKER_00: It really is.
• 5:33
  SPEAKER_00: It's like staring at that flawless bank vault,
• 5:35
  SPEAKER_00: knowing your business is perfectly secure inside,
• 5:37
  SPEAKER_00: but realizing you don't even have the master override switch anymore.
• 5:41
  SPEAKER_00: Catch you on the next Deep Dive.