Security Weekly Podcast Network

A popular NPM package was turned into a credential-stealing backdoor. Malicious versions of Node IPC attempted to exfiltrate AWS, Azure, Kubernetes, GitHub, Terraform, and other developer secrets through heavily obfuscated code disguised as legitimate traffic. This is the growing reality of software supply chain attacks: trusted packages becoming attack vectors overnight. How much do you actually trust your dependencies? #CyberSecurity #SupplyChainSecurity #NPM

Robot lawnmowers can become real-world attack surfaces fast. This conversation explores what happens when connected devices reuse credentials, expose firmware access, and give attackers control over physical systems with moving blades. If compromising one device gives access to every identical device, why are manufacturers still shipping products this way? #Cybersecurity #IoTSecurity #Hacking

AI isn’t just writing code anymore. In this clip, Paul describes an AI agent autonomously executing a voltage fault injection attack against an ESP32 secure boot process — including reverse engineering, scripting, and hardware control. The weird part isn’t that AI helped. It’s how much of the attack chain it handled on its own. How close are we to fully autonomous offensive security? #Cybersecurity #AI #HardwareHacking

AI agents don’t magically understand your environment. This clip explains why cybersecurity AI systems fail when they lack context about infrastructure, networks, applications, and business operations. Throwing alerts at an AI agent without context is like hiring a junior analyst and giving them zero onboarding. Are companies training AI systems — or just overwhelming them? Now booking interviews at Black Hat 2026. Early access pricing is open. Message us for details! #AI #CyberSecurity #SecurityOperations

One of the biggest security risks isn’t technical at all. It’s the assumption that “somebody else already reported it.” The same psychology behind people avoiding involvement in emergencies shows up in cybersecurity environments every day. In IAM and identity systems, ignoring unusual behavior because “someone else will handle it” can quietly create massive blind spots. How many security incidents get missed because everyone assumes someone else already acted? #Cybersecurity #IAM #SecurityCulture

A breach affecting millions of students wasn’t just about names and email addresses. The bigger concern may have been billions of private conversations between students, teachers, and classmates inside a major learning platform used across higher education. The clip also takes aim at the idea that avoiding the worst possible data exposure should count as a success. How should schools evaluate the real impact of breaches like this? #cybersecurity #databreach #privacy

“AI is really smart” might be the biggest misconception in tech right now. Ben Carr explains why LLMs are powerful at optimizing existing processes — but still struggle to invent entirely new ones. Human creativity, intuition, and process innovation still matter. The companies that win may not replace humans with AI… but combine them correctly. Where should AI stop and human judgment begin? #artificialintelligence #llm #innovation

Most companies think the breach is the crisis. Walter Wilkens explains why the harder part often comes afterward — identifying exactly what sensitive data was exposed and who must be notified. That means digging through compromised systems for PII and PHI while legal teams, insurers, and forensics teams race against the clock. How prepared is your organization for the post-breach phase? #cybersecurity #databreach #incidentresponse

Active Directory security tools have evolved, but recovery is still painfully slow. Guy Teverovsky explains how organizations often rely on highly customized disaster recovery processes that must be built and executed per environment — sometimes taking days or even weeks. Despite years of tooling improvements, the core operational problem hasn’t disappeared. Why is recovering critical identity infrastructure still so manual? #cybersecurity #activedirectory #identitysecurity

The warning signs may not be theoretical anymore. This clip breaks down a reported case where attackers allegedly used an LLM to help create a zero-day exploit targeting two-factor authentication through an open-source admin tool. The speaker pushes past the AI hype and focuses on the real concern: automation lowering the barrier for cybercrime. Are security teams prepared for AI-assisted attacks at scale? #cybersecurity #ai #zeroday

More CVEs doesn’t automatically mean more risk. In this clip, the speaker explains how Mozilla responded to Firefox sandbox escape and JavaScript prototype vulnerabilities — not by patching each issue individually, but by changing the architecture itself. They froze JavaScript prototypes to eliminate entire classes of exploits. Sometimes the real security win is not fixing bugs — it’s redesigning the system. Are security teams still over-optimizing for patching instead of architecture? #CyberSecurity #AppSec #SecureByDesign

AI systems don’t understand intent — only prompts. In this clip, the Rob Allen explains how the same underlying request can be rejected or approved depending on how it’s phrased, even when the outcome is effectively the same. That creates a hidden risk in AI-assisted security workflows. If intent is ambiguous, can AI ever reliably distinguish safe from dangerous use? #AI #CyberSecurity #LLM

Security alerts alone are useless. This clip explains why cybersecurity analysts need threat intelligence — context about malware, threat actors, and attacker behavior — before they can actually investigate anything. Without that context, alerts are just noise. How much security tooling is producing data without meaning? Now booking interviews at Black Hat 2026. Early access pricing is open. Message us for details! #CyberSecurity #ThreatIntelligence #SOC

AI in cybersecurity isn’t replacing humans — it’s amplifying them. This discussion argues that experienced practitioners using AI see major productivity gains, but the idea of full replacement is overstated. There’s even a surprising twist: AI-driven cost expectations may be reversing some layoffs. Is the industry underestimating how human cybersecurity still is? #ai #cybersecurity #infosec

A ransomware negotiator pleaded guilty to helping deploy ransomware. The discussion dives into the uncomfortable overlap between people negotiating with attackers and the defenders paying them. When someone operates on both sides of cyber extortion, where should trust end? #cybersecurity #ransomware #infosec

VPNs feel secure—but they create a major blind spot. Once inside, users often get broad, flat access across the network. That’s exactly how many modern attacks spread. This clip explains why teams are moving away from VPNs to ZTNA, and where to start: your noisiest gateway. Because security isn’t just about access— it’s about continuous validation. Are you still trusting your VPN too much? Now booking interviews at Black Hat 2026. Early access pricing is open. Message us for details! #cybersecurity #zerotrust #networksecurity

AI coding agents are being treated too casually. This clip explains why they should be handled like build systems with shell access: isolated environments, restricted networking, heavy logging, and strict review of config overrides. One bad repo configuration could quietly expand what an agent is allowed to do. Are most teams securing AI agents aggressively enough yet? #ai #cybersecurity #devsecops

Hack the developer, hack the software. This clip breaks down how a Linux RAT called Quasar steals developer credentials like Git, NPM, and PyPI tokens to compromise software repositories upstream. Instead of attacking users directly, attackers go after the people building the software. How do companies realistically defend against this kind of supply chain attack? #cybersecurity #supplychain #developers

Passwords get stolen. MFA prompts get approved. Rob Allen explains why some organizations are now locking SaaS apps like Office 365, GitHub, and Salesforce to specific trusted IP paths instead of exposing login access to the entire internet. Even if an attacker has credentials, they still can’t connect unless traffic comes from the approved location. Is location-based access control becoming the next layer after MFA? #Cybersecurity #SaaS #IdentitySecurity

Opening ports to the internet is still one of the easiest ways to increase attack surface. In this clip, Rob Allen explains how ThreatLocker is using its existing endpoint agent to provide VPN-style remote access — similar to Tailscale or WireGuard — without exposing ports publicly. Granular access rules mean users only reach exactly what they need. Is agent-based remote access becoming the new VPN standard? #Cybersecurity #VPN #ThreatLocker