Chaos Computer Club - recent audio-only feed

MCP security hot potato: how to stay secure integrating external tools to your LLM (god2025)

Model Context Protocol (MCP) is the latest hot topic in cybersecurity. Business wants it (AI is the new mantra), developers are excited (new toys, new code), and security teams are left to make it safe—often with already packed schedules. Let's treat it like just another Tuesday. Like many shiny new technologies (remember the early days of cloud?), MCP is being built with a “features first, security later” mindset. As a fresh piece of tech, it blends novel vulnerabilities with familiar, well-known ones. If you're an early adopter, it's important to accept that MCP and its current implementations are imperfect—and to be ready for that. In this talk, we'll dive into the real-world challenges companies are facing with MCP and equip you with practical remediations. We'll cover topics such as: An introduction to the MCP protocol and its security considerations, including authentication Emerging vulnerabilities like prompt injections, tool poisoning, rug pull attacks, and cross-server tool shadowing Classic vulnerabilities that may resurface around MCP, based on recent CVEs Remediation strategies and available tooling Licensed to the public under https://creativecommons.org/licenses/by-sa/4.0/ about this event: https://c3voc.de

Extract: A PHP Foot-Gun Case Study (god2025)

Do you always read the documentation before using a function in your languages' standard library? This talk explores the attack surface of a special feature in PHP which is easy to misuse with unforseen consequences. The `extract` function allows to replace the value of local variables named after the keys in an array. Calling it with user-controlled input allows the attacker to change arbitrary variables in the program. The documentation warns against the dangers of using it with untrusted data, but our large-scale analysis on 28.325 PHP projects from GitHub shows, that this warning is ignored. The talk walks through the process of identifing `extract`-based vulnerabilities and how they might have ended up the way they are by looking at the surrounding code. After introducing different levels of attacker-control guided by concrete exploits, listeners gain an intuition on what to look out for while reviewing code. Attending this talk, the audience will learn: Rich ways users have control over input in PHP. How to exploit insecure calls to `extract` given multiple real-world case-studies from the dataset of open source projects from GitHub. Tips on how to avoid this and similar threats in new and legacy code. Possible changes to PHP itself for risk reduction. Licensed to the public under https://creativecommons.org/licenses/by-sa/4.0/ about this event: https://c3voc.de

Pwn My Ride: Jailbreaking Cars with CarPlay (god2025)

Apple CarPlay is a widely known protocol that connects smartphones to car multimedia systems. Based on AirPlay, CarPlay is installed in millions of cars, as it is supported by hundreds of car models from dozens of different manufacturers across the globe. In our talk, we will share how we managed to exploit all devices running CarPlay using a single vulnerability we discovered in the AirPlay SDK. We'll take you through our entire exploit development process from identifying the vulnerability, to testing it on a custom device emulator, and finally, executing the exploit on actual devices. The session will include a demonstration of our RCE exploit on a well known third-party CarPlay device to show how an attacker can run arbitrary code while in physical proximity to a target car. We will also share how we managed to blindly exploit CarPlay without a debugger, knowing the vulnerable code is present on the system. Licensed to the public under https://creativecommons.org/licenses/by-sa/4.0/ about this event: https://c3voc.de

The Automation Illusion? What Machines Can't Do in Threat Modeling (god2025)

Threat modeling stands at a critical juncture. While essential for creating secure systems, it remains mostly manual, handcrafted, and often too slow for today's development cycles. At the same time, automation and AI offer new levels of speed and scalability— but how much can we rely on them? This talk explores the tension between automation and human expertise in threat modeling. We'll dissect the traditional threat modeling process—scoping, modeling, threat identification, risk analysis, and mitigation—and perform a step-by-step gap analysis to identify what can realistically be automated today, what cannot, and why. We'll dive into: Current tooling: Review the AI threat modeling tools that handle diagram-based automation, template-driven modeling, risk scoring, and pattern matching. Emerging AI use cases: automatically generating threat models from architecture diagrams, user stories, or use case descriptions; providing AI-assisted mitigation suggestions; and conducting NLP-driven threat analysis. Limitations and risks: False confidence, hallucinations, model bias, ethical accountability, and the challenge of modeling new or context-specific threats. We will ground this analysis with examples from organizations and academic research that aim to scale threat modeling without compromising depth or quality, drawing parallels to how other activities, such as SAST and DAST scanning, evolved. Attendees will walk away with a practical roadmap for integrating automation without undermining the human insight threat modeling still requires. This talk isn't a tool pitch. It's a candid, experience-based view of where automation can meaningfully accelerate threat modeling—and where the human must remain firmly in the loop. Licensed to the public under https://creativecommons.org/licenses/by-sa/4.0/ about this event: https://c3voc.de

Phishing for Passkeys: An Analysis of WebAuthn and CTAP (god2025)

WebAuthn was supposed to replace swords on the web: uniform, secure, manageable authentication for everyone! One of its unique selling points was supposed to be the impossibility of phishing attacks. When passkeys were introduced, some of WebAuthn's security principles were watered down in order to achieve some usability improvements and thus reach more widespread adoption. This presentation discusses the security of passkeys against phishing attacks. It explains the possibilities for an attacker to gain access to accounts secured with passkeys using spear phishing, and what conditions must be met for this to happen. It also practically demonstrates such an attack and discusses countermeasures. Participants will learn which WebAuthn security principles still apply to passkeys and which do not. They will learn why passkeys are no longer completely phishing-proof and how they can evaluate this consideration for their own use of passkeys. Licensed to the public under https://creativecommons.org/licenses/by-sa/4.0/ about this event: https://c3voc.de