
Vulnerability Research, AI Slop, and Why Fundamentals Still Win with Stephen Sims
Phillip Wylie ShowShow Notes
AI can now hand someone a working Linux kernel privilege escalation exploit, even if that person cannot explain a single line of how it works. Stephen Sims joined the show to talk about what that shift means for offensive security, and why the fundamentals matter more now, not less.Stephen is the curriculum lead for SANS Institute's Offensive Operations program, where he has spent more than 15 years as an author and instructor, and he is a co-founder of Off by One Security. He is a longtime exploit developer and vulnerability researcher who came up through game hacking, network engineering, and years of binary exploitation, reverse engineering, and weaponizing bugs in browsers and the kernel.In this episode, Stephen and Phillip get into how AI is reshaping vulnerability research and exploit development. Stephen explains why human validation is still doing the heavy lifting behind the big vulnerability-count headlines, how AI tends to overstate or understate severity, and why so many submissions are now AI slop or duplicates. He makes the case that logic bugs remain the hardest thing for AI to find, walks through his roadmap for anyone serious about learning exploit development, and shares why he tells students to do the work manually before letting AI do it for them. Stephen also offers grounded advice for breaking into the field, from building a real technology foundation first to staying curious and paying your dues, and gives his honest read on where the opportunity is heading. ========================= Connect with Stephen Sims: LinkedIn: https://www.linkedin.com/in/stephen-sims-2788091/ X: https://x.com/Steph3nSims Off by One Security: https://offbyonesecurity.com/ YouTube: https://www.youtube.com/@OffByOneSecurity ========================= Connect with your host, Phillip Wylie: LinkedIn: https://linkedin.com/in/phillipwylie X: https://x.com/PhillipWylie Instagram: https://www.instagram.com/phillipwylie ========================= Sponsored by Suzu Labs ========================= All the ways to connect with Suzu Labs https://suzulabs.com https://x.com/suzulabs https://www.linkedin.com/company/suzu-labs/
More from Phillip Wylie Show
What happens when someone who knows exactly how applications are built decides to start breaking them? Jarrod Rizor went from writing code and running DevOps to hacking web and mobile apps for a living, and he recently joined TrustedSec as a penetration tester. His path is a real-world playbook for anyone trying to break into offensive security.
Tim MalcomVetter has spent 26 years moving between offense and defense, and his biggest takeaway is that the future of security comes down to speed - whoever has the fastest software wins. Tim got his start by offering to work for free at a hospital in the early days of HIPAA, which turned into his first security job. From there he moved into pen testing, including work on the point-of-sale system behind roughly one in three U.S. credit card transactions, and then into red teaming at Walmart, where engagements could run as long as eight months. Today he is the co-founder of Wirespeed and General Manager of Coalition Security. Recorded in person at Cybr.Sec.Con in Houston, this conversation with Phillip gets into what separates real red teaming from pen testing, what cyber insurance data reveals about which security controls actually work, and why so many in the industry misjudge risk. Tim also shares how his red team experience shaped his approach to automation, and why he sees the next era of attack and defense being decided in milliseconds. ========================= Connect with Tim MalcomVetter: LinkedIn: https://www.linkedin.com/in/malcomvetter WireSpeed: https://wirespeed.co ========================= Connect with your host, Phillip Wylie: LinkedIn: https://linkedin.com/in/phillipwylie X: https://x.com/PhillipWylie Instagram: https://www.instagram.com/phillipwylie
License plate readers, police drones and fake cell towers are spreading fast, and two hackers are proving that anyone can detect them with about ten dollars of hardware and some open source firmware. Colonel Panic is a malware researcher and hardware builder whose hacking started with AOL dial-up and stacks of 2600 Magazine. After years in restaurants and grocery retail, he went back to school during COVID, earned his certs and turned his love of rapid prototyping into privacy tools like OUI Spy. Cooper got his start on a hand-me-down 286 in fifth grade, spent his early years teaching privacy and security to activists, and has been a public interest technologist at the Electronic Frontier Foundation for almost thirteen years, where he works on Rayhunter. Building on their DEF CON talk, "ESP32 as a Counter-Surveillance Platform," this episode covers: - How OUI Spy uses an ESP32-S3 and a buzzer to alert you when specific Bluetooth or Wi-Fi devices or drones are nearby, and what changed when Flock updated its firmware - Mesh Detect, which sends alerts over Meshtastic networks to build a sensor net for tracking drones - Rayhunter, EFF's open source firmware that turns a cheap Orbic mobile hotspot into a cell-site simulator detector - Why "I have nothing to hide" falls apart, including a Flock misread that put an innocent woman in jail for weeks - The national security risk of ALPR data piling up with vendors who have a track record of vulnerabilities - The current surveillance boom and EFF's work on litigation, activism, Privacy Badger and Certbot - Why war driving is legal, and how to get involved by building tools or showing up at city council meetings DEF CON 34 - ESP32 as counter-surveillance platform: https://www.youtube.com/watch?v=PcK-TDmzshcEFF ========================= Connect with Colonel Panic: Website: https://colonelpanic.tech https://x.com/_colonel_panic https://colonelpanic.tech/ Connect with Cooper: https://bsky.app/profile/cooperq.com EFF: https://www.eff.org ========================= Connect with your host, Phillip Wylie: LinkedIn: https://linkedin.com/in/phillipwylie X: https://x.com/PhillipWylie Instagram: https://www.instagram.com/phillipwylie
Luke Canfield came up the way a lot of hardware hackers do, the geeky kid in high school everyone came to when the school's computers broke, tinkering with Windows XP and war driving with homemade Yagi antennas built out of Pringles cans. That fascination with the hardware side never left him, and today he runs Reconnaissance Drone Consulting, teaching people about a threat surface most of the security world still isn't looking at.
Zack Korman came into cybersecurity from about as far outside the field as you can get, a UK law degree and a master's in law and finance, and he now spends his days on the AI defense problems most of the industry is still catching up to. His outsider path is exactly what makes his take on the current AI moment worth hearing. Based in Norway, Zack got into code to win an argument with a professor, worked his way up to running a media company's tech department, then spent four and a half years as CTO of an Oslo cybersecurity startup building AI phishing simulations and AI insider threat detection. That work pulled him deep into the defensive side of security and into a conviction that detection is one of the most underserved areas of the field. Phillip and Zack get into why the "thousand crap alerts" model of detection is broken, the blind spots that let obvious insider threats slip past a SOC that was never even receiving the data, and Zack's work on AI agent security, including the malicious MCP server he built to prove how completely an agent can exfiltrate a company. From there the conversation turns to the thing Zack has become known for online, pushing back on inflated AI risk claims, the fallout from the Hugging Face incident, and why so many of the loudest warnings about AI in cybersecurity are coming from people who have never worked in security. They also cover frontier versus local models, why smaller models could reshape defensive work, model guardrails and offensive testing, and Zack's straightforward advice for anyone trying to fold AI into their day to day. ========================= Connect with Zack Korman: LinkedIn: https://www.linkedin.com/in/zacharyakorman/ X: https://x.com/ZackKorman ========================= Connect with your host, Phillip Wylie: LinkedIn: https://linkedin.com/in/phillipwylie X: https://x.com/PhillipWylie Instagram: https://www.instagram.com/phillipwylie ========================= Sponsored by Suzu Labs ========================= All the ways to connect with Suzu Labs https://suzulabs.com https://x.com/suzulabs https://www.linkedin.com/company/suzu-labs/