Data Security Decoded

Are your cloud security controls actually protecting your infrastructure, or are they just keeping the lights on? With host Caleb Tolin, Matt Castriotta, Field CTO for Cloud at Rubrik, breaks down the tactical gaps exposed when organizations blindly replicate data center mindsets in public cloud networks. Castriotta charts the history of high-profile incidents from the Colonial Pipeline timeline up through modern adversaries like Scattered Spider and Storm-0501. He highlights how today's attackers move laterally by exploiting over-privileged, non-human identities to trigger malwareless mass deletion rather than relying on on-prem style encryption loops.

The discussion pivots into an actionable critique of popular resilience assumptions. Castriotta details why relying on built-in features like S3 versioning and cross-region replication handles business continuity but leaves organizations entirely defenseless against automated cyber assaults. He delivers a precise operational roadmap for defining a "minimum viable business," establishing secure isolated recovery environments, and breaking the 80% ransomware reinfection cycle. This episode serves as an essential strategic guide for any enterprise trying to align the cloud shared responsibility model with predictable, audited return-to-service timelines.

Resources

⁠Rubrik Cloud Cyber Resilience Solutions


Microsoft Threat Intelligence Report on Storm-0501


Scattered Spider Threat Profile

What You’ll Learn


How to separate low-probability disaster recovery protocols from high-probability cyber attacks.


The architectural threat mechanisms behind malwareless, privilege-driven data destruction.


A blueprint for prioritizing operations based on your minimum viable business components.


Solutions to tackle non-human credential sprawl and enforce just-in-time domain separation.


The hard realities of cloud platform pricing mechanics during major recovery events.

What happens when the adversary’s dwell time is measured in years, but your defense is measured in tickets? Adam Karcher, FBI Supervisory Special Agent, Cyber Division, and a member of the Bureau’s AI Working Group, joins host Caleb Tolin to break down the "convergent evolution" of modern cyber threats. Karcher explains why defenders are often stuck in a cleanup cycle, while threat actors operate in a sophisticated, compartmentalized ecosystem that requires a fundamental shift in defensive strategy.

The conversation provides a rare look at how the FBI evaluates agentic AI technology. Karcher warns of the transition from AI that simply answers questions to agents that take independent actions, emphasizing why these systems must remain well-bounded and auditable. He also debunks the "glamorous" myth of cyber investigations, revealing why law enforcement breakthroughs almost always stem from human OPSEC mistakes rather than complex code analysis. Whether you are managing legacy mainframes or securing a modern identity stack, this episode provides a tactical roadmap for treating your security posture as an "inverted offensive campaign."

Resources


Information Sharing and Analysis Centers (ISACs)⁠


⁠Local FBI Field Offices

What You’ll Learn


Match your defensive cadence to the adversary's multi-year campaign dwell time.


Prioritize auditable AI use cases to prevent autonomous agents from acting on hallucinations.


Focus on "people mistakes" like infrastructure reuse rather than just analyzing malicious code.


Secure identity stacks to defend against AI-driven deep fakes and precision phishing.


Engage with ISACs and local field offices before a crisis occurs.

Dr. Ido Sivan Sevilla joins host Caleb Tolin⁠⁠⁠ to break down battlefield stories from a massive analysis of over 3,000 local government entities. Dr. Sivan Sevilla, who serves as an Assistant Professor at the UMD College of Information and holds joint positions at the Hebrew University School of Public Policy & Governance and the School of Computer Science and Engineering, brings a multidisciplinary lens to the alarming reality of risk clusters. Their discussion moves past theory to explore how hundreds of counties share identical IP addresses and third-party service providers, creating centralized points of failure that attackers can identify using data.

The dialogue highlights the dual-use nature of modern AI models. While these tools allow adversaries to automate exploit generation for open-source software, Dr. Sivan Sevilla, leveraging his expertise as founder of UMD's Tech Policy Hub, explains how defenders can use AI operations to map their own attack surfaces for free. By utilizing honeypots and large language models, limited-resource organizations can transition from reactive patching to a proactive posture. The episode concludes with a strategic look at identity resilience, advocating for adaptive regulations that learn from compliance data rather than static, outdated legislative mandates.

Resources


CISA KEV Catalog: https://www.cisa.gov/known-exploited-vulnerabilities-catalog


This research was conducted by Dr. Ido Sivan Sevilla, Dr. Charles Harry, and Mr. Mark McDermot, with additional support from student researcher Mr. Parthav Poudel

What You’ll Learn


How to prioritize the 3% of vulnerabilities that actually result in real-world exploitation.


The definition of attack surface diversity versus severity in measuring county level risk.


The impact of LLMs on identifying flaws in open source software for attackers and defenders.


Why risk clusters create a single point of failure for hundreds of independent county governments.


Methods for conducting ethical passive reconnaissance to map organizational security postures from the outside.


How adaptive regulations can improve compliance by learning from real-time security data and metrics.


The strategic benefit of using honeypots to monitor targeted threats against limited-resource digital infrastructure.

In this episode, host⁠ ⁠Caleb Tolin⁠⁠ explores the battlefield of enterprise defense, which has moved from simple data theft to ultra heinous crimes that put patient outcomes at risk. Guest⁠ ⁠Cynthia Kaiser⁠⁠ shares Battlefield Stories from her time at the FBI and her current work as SVP of the Ransomware Research Center at⁠ ⁠Halcyon⁠⁠, illustrating how the industrialization of cybercrime has reached a tipping point. They dive into the alarming reality of modern dwell times, specifically looking at how groups like Akira move from initial access to full encryption in as little as one hour.

The conversation challenges the industry to face the inconvenient truths of cybercrime and ransomware. Kaiser shares case studies of how modern cybercriminals are adopting multilateral techniques to gain access to and exploit your network. By adopting an Assume Breach mindset, elite defenders can build the defense in depth required to combat malicious threat actors who follow their own rules to cause disruption and destruction.

Resources

House Homeland Security Committee Testimony: ⁠Online Scams, Crypto Fraud, and Digital Extortion⁠

Halcyon Analysis: ⁠Akira Ransomware Attacks in Under an Hour⁠

Halcyon: ⁠Sicarii Ransomware Encryption Key Handling Defect⁠

Previous Episode Referenced: ⁠Downtime in Healthcare is Fatal: Achieving Resilience in Health & Life Sciences

What You’ll Learn

Why designating ransomware as terrorism helps influence adversary target selection.

The impact of Akira's accelerated dwell time on traditional incident response.

How AI enables clumsy amateur "wannabes" to conduct messy attacks.

The critical role of phishing resistant MFA in securing the identity perimeter.

Why Assume Breach necessitates deep defense in depth strategies.

The overestimation of readiness among CISOs compared to actual red team performance

Episode Highlights

[00:00] - The Case for Designating Ransomware as Terrorism

[04:20] - Modern Extortion and the Shortening of Dwell Time

[08:30] - Ransomware Recovery in Interconnected Cloud Environments

[11:45] - The Impact of AI on the "Wannabe" Attacker

[17:45] - Three Actionable Steps for Modern Defenders

[21:30] - Inconvenient Truths for Government and Private Sector

The race for AI dominance has created a dangerous imbalance between business velocity and cyber resilience. In this episode, host Caleb Tolin is joined by Joe Hladik, Head of Rubrik Zero Labs, and Staff Security Researcher Amit Malik to break down the findings of their latest report on agentic adoption. The discussion centers on the Agentic Paradox. This is the technical reality that tools designed to automate high-level tasks are inherently built to find the most efficient path around obstacles, including existing security policies.

A primary focus is implementing a three-layer framework for AI Operations. This model targets the Tool Layer, where agents interact with databases; the Cognitive Layer, which serves as the LLM brain; and the critical Identity Layer. The conversation explores stories in which agents, without malicious intent, have caused catastrophic data loss simply by following an optimized logic path. These instances prove that agents need not be sentient to be destructive when they lack proper human-in-the-loop checkpoints.

Technical hurdles of Identity Resilience are also addressed, specifically the explosion of non-human identities that spin up and down like elastic cloud infrastructure. The episode examines the fear index regarding job security, noting that 92% of leaders fear for their roles post-breach. Joe and Amit join Caleb to explore the evolution of personal liability for CISOs and the urgent need to move from basic visibility to deep observability. This is a forward-looking briefing for leaders who recognize that, in an era of autonomous routines, the human must remain the ultimate command-and-control center.

What You’ll Learn


Define the agentic paradox to understand why AI efficiency naturally compromises traditional security guardrails.


Implement a three-layer framework to secure the tool, cognitive, and identity components of AI.


Transition from basic visibility to deep observability to track autonomous decision-making in real time.


Mitigate prompt injection risks by auditing the input and output flows of the cognitive layer.


Utilize ephemeral containers to sandbox agentic tools and prevent unauthorized database alterations.


Manage the elasticity of non-human identities to maintain control over rapidly spinning AI agents.


Anchor AI operations with human-in-the-loop checkpoints to ensure integrity during high-stakes executions.

Episode Highlights


Defining the Agentic Identity and Autonomous Routines


Revenue vs. Resilience: The Drivers of AI Urgency


The Three-Layer Framework for Agentic Defense


Shadow AI and the Rise of Invisible Insider Threats


The Context Gap: Why Rolling Back AI Actions is Hard


The CISO Fear Index and Personal Liability Post-Breach


Visibility vs. Observability in Elastic Identity Environments