SN 1103: Exposing Bad Consumer Crypto - AI vs. Encrypted Drives
Nov 4, 2026 · 2h 48m
More Technology podcasts
Summary
Steve Gibson discusses a critical Linux kernel vulnerability discovered by AI, a 16-year-old UEFI firmware flaw, and the record-breaking $6.3 billion Binance crypto heist attributed to North Korean hackers. The episode also covers Cisco’s controversial decision to have a 27-year-old rewrite 98% of its legacy code using AI, alongside a smaller Polymarket hack and an analysis of AI’s growing role in finding obscure security flaws.
Topics discussed
Intro: Cisco, crypto heist, and AI kill switch
Listener feedback on strategic plug pull
Cisco's 98% code rewrite by Gen Z
robots.txt for AI and Google's orbital haiku
Linux kernel MPTCP race condition exploit
Sponsor: Hawk Hunt security alerts
Listener sign from Polk County, Iowa
MPTCP exploit severity and difficulty
Raw TCP kernel vulnerability details
Cloud Linux updates and AI discovery
Mitigating MPTCP via IP options blocking
Reflections on finding high-level bugs
Cisco's Gen Z founder and code refactoring
Unlimited tokens for code refactoring
Founder buried in Cisco org structure
Hiring early-career workers for fresh ideas
Reverse mentoring interns and unreasonable goals
Cisco security policy about-face
Placing customers at risk with security policy
Embargoed UEFI vulnerability disclosure
Botched code execution vulnerability response
TianoCore UEFI reference code usage
UEFI rootkit installation via boot process
Patching UEFI vulnerabilities in devices
CVE 2026-976-31: 16-year-old flaw
Local network PXE boot time vulnerability
Buffer overflow in PXE receive path
Ring zero execution and attestation bypass
Subverting UEFI chain of trust
Forever bug in ubiquitous UEFI flaw
Annoyances of PXE booting
Enterprise server booting responsibilities
Microsoft's AI-driven code auditing
Microsoft performance and legacy code fixes
Sponsor: ThreatLocker endpoint protection
Binance $6.3 billion crypto hack
Binance investigation: compromised laptop
Attribution to trader-traitor by TRM
Fake crypto on Ethereum test network
Movement of stolen crypto assets
Polymarket hack and reimbursement pledge
Polymarket refund details and forms
DebugBear hack and Polymarket compromise
Malicious code via trusted DebugBear conduit
Ransomware insurance and crypto rates
Sponsor: Trusted Tech Microsoft support
Early AI developers and web scraping ethics
Book recommendation and AI context
LLM training on World Wide Web data
Publisher opposition to AI training
SEO craze and Google indexing
Major AI companies and robots.txt
Voluntary nature of robots.txt
Perplexity stealth crawling controversy
Microsoft granular AI training controls
New meta tag for AI training opt-out
Evasion by sleazy companies and AI trouble
Google's AI chip satellite response
TPU tests in orbit and 2027 milestone
Space-based warfare frontier
Sponsor: Bitwarden satisfaction score
Bitwarden self-hosting for federal agencies
Listener feedback on stealth updates
Challenges of fixing widespread flaws
Cloudflare WAF and tunnel architecture
Stateful packet filtering in tunnels
Protection from PHP flaws via WAF
Consolidating web traffic through Cloudflare
Role of consolidation in security incidents
Running AI agents for home monitoring
Knowledge vs data in books and LLMs
Encoding knowledge in neural networks
Books imparting knowledge and emotion
Small org AI-assisted threat landscape
Cloudflare WAF as part of solution
AI for infrastructure security management
Economics of security for startups
Lowered interest rates and startup costs
Better to find issues while small
Agentic AI for defensive security
Election day and AI regulation timing
Listener preference for AI coverage type
Club TWiT reminder
Sponsor: Carvana car selling
Fake encrypted USB drive scam
Doubts about Lenovo's integrity
Marketing AES-256 without hardware
Reverse engineering the Think Plus FU100
Requirements for true encrypted drives
PBKDF and master key derivation
Unspoofable encryption system design
Complexity of fingerprint recognition
Chinese reverse engineer's findings
AI reproducing research in minutes
Time travel debugging for RE
Software hooks in PDF reader
Backdoor login buffer discovery
Undocumented admin credentials in routers
Opaque crypto products and backdoors
NETAC FileLock analysis
Non-chained AES encryption blocks
File types supported and mixed feelings
Publishing findings for user awareness
AI bypassing insecure crypto designs
Significance of consumer crypto flaws
Recommendation for open-source encryption
Show outro and subscription info
Sponsor: Wayfair Christmas trees
More from Security Now (Audio)
Listen ad-free on Castria