Workshop

• August 9, 2026 Workshop Offensive Security Red Team Village — DEF CON 2026 Fernando Mengali BR·Thiago Cunha da Silva BR From Buffer Overflow to Blackout: Chaining Attacks Against Industrial Systems

  This presentation takes a practical, demo-driven approach to exploitation, demonstrating how a buffer overflow vulnerability can still serve as an entry point to compromise modern industrial infrastructures. The session is approximately 90% hands-on and 10% theoretical.

  The talk walks through the full attack chain, starting from initial exploitation and progressing to the development and execution of multistage shellcode. This shellcode establishes communication with a remote Command and Control (C2) server, retrieves an additional payload, and executes it directly in memory, bypassing traditional detection mechanisms.

  Once the environment is compromised, the session demonstrates how an attacker can gain full remote control, enabling command execution, lateral movement, and exfiltration of sensitive data, including critical assets such as licenses and proprietary industrial information.

  The talk also includes interaction with real-world devices such as RTUs (Remote Terminal Units), showing how malicious actions can directly impact industrial operations. In critical scenarios, this can lead to disruptions in essential services, including failures in energy systems and potential blackouts.

  The content is based on real-world pentesting experience conducted in an energy sector company in Brazil, providing a practical and applied perspective on risks in ICS/SCADA environments.

  The goal is to demonstrate how modern attacks combine vulnerabilities with advanced techniques, reinforcing the need for robust security strategies to protect critical infrastructure.

  ResearchersFernando Mengali BR·Thiago Cunha da Silva BR

  Presented at Red Team Village — DEF CON 2026

  View official session
• August 8, 2026 Workshop AI Security Recon Village — DEF CON 2026 Lenin Alevski MX Vibe Coding Your Way to a Fully Functional Open-Source Intelligence Platform

  Everyone is talking about vibe coding, but most examples stop at TODO apps, landing pages, or simple CRUD applications. This workshop goes far beyond that.

  Participants will use modern AI coding agents to build a real-world Open-Source Intelligence (OSINT) platform from scratch designed to aggregate and visualize intelligence data on an interactive 3D globe.

  Rather than simply watching a demo, attendees will actively build the application alongside the instructor while learning the mindset, prompting techniques, and engineering workflow required to successfully collaborate with AI coding assistants.

  The workshop begins with the fundamentals of vibe coding: what it actually is, why it works, where it fails, and how to communicate effectively with AI agents. Participants will learn practical prompting strategies, iterative development techniques, and methods for breaking large software systems into manageable tasks that AI can successfully implement.

  From there, we'll progressively build an intelligence dashboard capable of consuming and visualizing multiple real-time OSINT sources, including flight tracking, satellite positions, earthquake feeds, maritime traffic, and other publicly available intelligence signals. Along the way, participants will generate modern user interfaces, integrate external APIs, debug AI-generated code, and learn when to trust the model and when not to.

  Rather than treating AI as a code generator, this workshop teaches attendees how to use AI as an engineering partner capable of dramatically increasing development velocity while still maintaining high-quality software.

  By the end of the workshop, every participant will have:

  1. A working open-source intelligence platform running locally

  2. A practical workflow for building complex software using AI coding agents

  3. A reusable prompting framework applicable to future projects

  4. Experience integrating multiple public OSINT data sources

  5. An understanding of the strengths and limitations of AI-assisted software development

  6. A roadmap for continuing to extend the platform after the workshop

  The workshop is based on the open-source project: https://github.com/Alevsk/respondent-community

  Workshop Outline (150 Minutes)

  Part 1 — Introduction to Vibe Coding (20 min)

  • What vibe coding actually means
  • Mental models for collaborating with AI
  • Selecting the right AI model for the task
  • Understanding agentic coding workflows
  • Common mistakes and misconceptions

  ---

  Part 2 — Prompt Engineering for Software Development (25 min)

  • Structuring prompts that produce maintainable code
  • Breaking large systems into incremental tasks
  • Designing software through conversation
  • Iterating instead of regenerating
  • Knowing when to intervene manually

  ---

  Part 3 — Building the Intelligence Platform (75 min)

  Participants will build the platform together while learning how to:

  • Generate a modern frontend
  • Create a real-time 3D globe visualization
  • Integrate multiple public OSINT APIs
  • Display flights, satellites, earthquakes, ships, and additional intelligence feeds
  • Build reusable UI components with AI
  • Debug AI-generated code
  • Refactor and improve generated implementations
  • Continue extending the application using AI as a development partner

  ---

  Part 4 — Lessons Learned & Advanced Techniques (20 min)

  • What AI does well
  • Where AI still struggles
  • Managing technical debt
  • Building larger projects with AI
  • Cost optimization strategies
  • Recommended workflows and tooling
  • Future directions for AI-assisted engineering

  ResearchersLenin Alevski MX

  Presented at Recon Village — DEF CON 2026

  View official session
• August 8, 2026 Workshop AI Security Red Team Village — DEF CON 2026 Mario Lobo CO AIMARU C2: The New Era of LotL (MCP AI-Driven C2)

  AImaru C2 is the evolution of a concept sparked by an unsettling question: What happens when an attacker can execute advanced Living off the Land (LotL) attacks without being an expert? While traditional LotL requires a specialist to navigate complex environments, AImaru C2 changes the game. Evolving from a PoC into a sophisticated Red Team framework, it automates expert-level tradecraft by subverting Anthropic's Model Context Protocol (MCP). By weaponizing MCP clients, it transforms a standard context protocol into a functional, autonomous multi-level RAT that natively "speaks" PowerShell.

  AImaru C2 was born out of a critical necessity to ground cybersecurity in reality. While the industry often distracts itself with over-engineered narratives and hype-driven threats that ignore actual ecosystem data, a far more pragmatic danger has been hiding in plain sight. This framework addresses the unsettling evolution of the "Expert Bottleneck": What happens when the sophisticated LotL tactics, once reserved for elite nation-state actors, are democratized through automation? Named after the mythical serpent that glides unseen between the underworld and the land of the living, AImaru C2 represents the shift from manual expertise to autonomous execution. It is no longer just a proof of concept; it is a full-scale Red Team framework designed to expose how easily an unskilled attacker can now weaponize authorized system protocols to orchestrate advanced, data-driven breaches that traditional defenses simply aren't looking for.

  To understand its impact, context is key. For years, LotL attacks have been the dominant tactic in ransomware campaigns. Instead of deploying traditional malware, threat actors abuse legitimate OS tools—such as PowerShell, WMI, or Certutil—to operate under the radar. In 2025, 82% of successful breaches were fileless or malwareless, with PowerShell ranking as the second most utilized TTP by groups like Black Basta, Royal, and LockBit.

  This robust framework moves beyond experimental concepts, consolidating itself as a production-ready offensive platform. AImaru C2 is engineered with a modular architecture that prioritizes cryptographic integrity, operational security, and intelligent automation.

  Core Framework Capabilities:

  • Model Context Protocol (MCP) RAT Engine: A paradigm shift in command and control that subverts Anthropic's MCP from a standard context-sharing tool into a functional, undetectable Remote Access Trojan (RAT), tunneling all C2 traffic through legitimate LLM API communication.
  • PowerShell Client Builder: An automated obfuscation engine featuring deep variable randomization and structural mutation to break static signatures.
  • AMSI Bypass Generator: Dynamic generation of multi-stage bypass scripts, utilizing diverse techniques to neutralize the Antimalware Scan Interface in real-time.
  • Real-time Monitoring: Advanced telemetry for live command execution tracking and granular session management.
  • Cryptographic Isolation: Implementation of per-client encryption keys derived via HKDF-SHA256, ensuring that the compromise of one beacon does not jeopardize the entire fleet.
  • RBAC: Native Role-Based Access Control, strictly segregating permissions between Admins, Users, and Viewers.
  • Complete Audit Logging: Forensic-grade command history with millisecond-accurate timestamps for post-operation deconfliction.

  By utilizing a multi-tier decision logic, the system dynamically selects and rewrites payloads based on the specific operational objective. This allows the C2 to scale its complexity in real-time—transitioning from basic PowerShell structures and stealthy WMI manipulation to the strategic abuse of LOLBins for high-complexity tasks. This adaptive approach ensures that every command utilizes the most effective legitimate system tool, successfully evading environment-specific defenses by blending into the target's unique operational noise.

  Tactic Description: In this interactive session, attendees will sit down with the developer to operate the AImaru C2 framework in a controlled lab environment. Participants will experience the "Natural Language Intent" paradigm shift firsthand, moving away from manual syntax to intent-based exploitation.

  Deployment: Deploy a lightweight MCP-Client on a target Windows machine and establish a beacon back to the AI-C2 controller.

  Intent-Based Recon: Issue high-level natural language prompts (e.g., "Find sensitive PDF files and identify lateral movement paths") and observe the LLM selecting and generating the appropriate PowerShell/WMI code in real-time.

  Automated Evasion: Trigger the Adaptive AMSI Bypass module, witnessing how the framework re-writes its own memory-resident payloads to evade active defenses.

  ResearchersMario Lobo CO

  Presented at Red Team Village — DEF CON 2026

  View official session Project on GitHub
• August 8, 2026 Workshop Offensive Security Red Team Village — DEF CON 2026 Arnold Morales MX Evading EDR in ATM

  ATMs in the end are a computer with windows.

  I will talk about EDRs and how parent-child process inheritance works.

  The Windows boot process, known as the boot process, is divided into several phases as we have already seen: PreBoot (initialization of BIOS/UEFI firmware and POST), Boot Manager (loading of bootmgr or bootmgfw.efi), OS Loader (execution of winload.exe to load the kernel), and Kernel Initialization (loading of ntoskrnl.exe and hal.dll, initialization of drivers and services). Finally, winlogon.exe is started to present the user interface. This process enables the transition from a powered-off state to a functional operating system through a hierarchical and verified sequence.

  Regarding kernel-based EDRs (Endpoint Detection and Response), they operate in kernel mode, the most privileged level of the system (some also operate at the UEFI level), which allows them to monitor low-level activities such as system calls, memory manipulation, driver loading, and file access. By running in this mode, EDRs can detect and block advanced threats such as rootkits or persistent malware that attempt to evade user-mode protections. Their deep integration with the kernel enables real-time monitoring, behavioral analysis, and immediate response to suspicious activities during and after the boot process.

  EDRs are typically located in the kernel, but as mentioned, some also operate at the UEFI level. If we recall how Windows works, beneath UEFI lies SMM (System Management Mode), making this an even more complex layer to analyze and operate within.

  This talk is intended for individuals interested not only in ATMs and simple cash dispensing, but in going beyond that—focusing on deeper analysis and the full experience gained when confronting these systems.

  ResearchersArnold Morales MX

  Presented at Red Team Village — DEF CON 2026

  View official session DMA Introduction (write-up) Telemetry Bypass (write-up)
• August 8, 2026 Workshop AI Security La Villa Hacker — DEF CON 2026 Luis Pazmiño EC Before the SIEM Blinks: How AI Memory Turns Alerts into Intelligence

  La mayoría de los SOCs son reactivos por diseño: los eventos llegan, se almacenan, se correlacionan horas después, y en el mejor de los casos un analista los revisa cuando puede. Este workshop rompe ese modelo. Presentamos una arquitectura SOC-less impulsada por IA donde la detección ocurre en línea, antes de que los datos lleguen al SIEM, usando reglas Sigma traducidas a lógica ejecutable en Groovy. Cuando una regla hace match, el evento se enriquece al instante con contexto de usuario, criticidad del activo, estructura de procesos y técnicas asociadas a MITRE ATT&CK.

  El diferenciado de nuestra propuesta es la memoria. Cada alerta, ticket, resolución, triaje de analista y falso positivo alimenta un índice recurrente. los agentes consultan ese historial para determinar si una detección es nueva, recurrente, vinculada a un activo crítico o parte de un patrón conocido, luego generan el ticket, sugieren severidad, recomiendan la contención y, en escenarios controlados la ejecutan. El resultado del flujo es la detección y respuesta inicial medidas en segundos, no en turnos de analistas.

  ResearchersLuis Pazmiño EC

  Presented at La Villa Hacker — DEF CON 2026

  View official session
• August 8, 2026 Workshop Offensive Security Red Team Village — DEF CON 2026 Filipi Pires BR DPAPI Is Not a Boundary: A Full Infostealer Kill Chain Operators Can Replicate

  Complete infostealer kill chain hands-on, executed against Windows 11 21H2 on default UAC using only open-source tooling. No zero-days, no commercial C2. Participants run every command live: HTA delivery, Meterpreter in-memory session, Chrome and Edge credential extraction via DPAPI-locked SQLite, keylogging in explorer.exe memory, UAC bypass via fodhelper, SYSTEM via Named Pipe Impersonation, and LSASS dump via Kiwi. The DPAPI_SYSTEM key recovered from LSA secrets decrypts the browser credentials captured earlier, closing the loop.

  OUTLINE

  • Stage 1: generate HTA payload, deliver via Python HTTP server, open Meterpreter session
  • Stage 2: extract Chrome and Edge credentials via post module, inspect DPAPI-encrypted loot
  • Stage 3: migrate to explorer.exe, start keyscan, capture live keystrokes
  • Stage 4: UAC bypass via fodhelper registry hijack, elevate to SYSTEM via Named Pipe Impersonation
  • Stage 5: load Kiwi, dump NTLM hashes, SAM, LSA secrets, extract DPAPI_SYSTEM key, decrypt browser loot
  • Detection debrief: Sysmon Event ID 1 process tree and PowerShell Script Block log review per stage

  TAKEAWAYS

  1. A complete credential theft playbook from HTA delivery to LSASS dump, validated against Windows 11 21H2 with Chrome 147 and Edge 147 on default UAC, ready to adapt for engagements.
  2. The DPAPI close-loop: how DPAPI_SYSTEM from LSA secrets decrypts browser credentials collected earlier, giving operators a concrete credential pivot from a single user-context session.
  3. A detection gap map calibrated against real Sysmon telemetry showing which events fire on default Windows 11 and which require EDR behavioral rules most organizations do not have deployed.

  ResearchersFilipi Pires BR

  Presented at Red Team Village — DEF CON 2026

  View official session
• August 8, 2026 Workshop Offensive Security Adversary Village — DEF CON 2026 Filipi Pires BR Infostealer: Replicating Commodity Threat Actor Credential Theft TTPs and Validating Detection Gaps

  Commodity infostealers account for a significant share of initial access in enterprise breaches. Their TTPs are documented, their tooling is public, and most detection stacks still miss them. This workshop emulates the full infostealer playbook against a live target and measures exactly where the gaps are.

  Attendees execute an eight-stage adversary emulation scenario replicating documented commodity threat actor TTPs: HTA phishing delivery, in-memory C2 via Meterpreter, browser credential theft from Chrome and Edge using DPAPI and the SQLite Login Data file, in-memory keylogging via process migration to explorer.exe, UAC bypass using the fodhelper registry hijack documented across multiple threat actor profiles, and LSASS credential dumping via Kiwi/Mimikatz recovering NTLM hashes, SAM contents, and the DPAPI_SYSTEM master key. Every technique is drawn from documented threat actor behavior and mapped to MITRE ATT&CK before execution begins.

  The second half of the workshop shifts to validation. Attendees enable Sysmon64 telemetry and replay the emulation, stage by stage, identifying which techniques generated detectable events and which did not. Each gap maps to a concrete control recommendation. The keylogging stage writes nothing to disk. The Meterpreter session exists only in memory. The DPAPI_SYSTEM extraction is not anomalous to most SIEMs by default. Attendees document all three as validated findings, not assumptions.

  Key Takeaways

  1. Adversary emulation with real TTPs produces validated detection gaps, not theoretical ones. Running the actual technique against your stack tells you whether the control works. Running it in this workshop tells you before an attacker does.

  2. In-memory execution breaks file-based detection entirely. Meterpreter leaves no binary on disk. The only detection path is behavioral: process injection signals, memory scanning, or network correlation between a user-context process and an external C2 session.

  3. DPAPI_SYSTEM recovery after LSASS access retroactively compromises all protected data on the machine. Most SIEMs do not alert on this extraction by default. Attendees leave with the specific Event IDs and Sysmon rule configuration that catch it.

  ResearchersFilipi Pires BR

  Presented at Adversary Village — DEF CON 2026

  View official session
• August 7, 2026 Workshop AI Security La Villa Hacker — DEF CON 2026 Matias Armándola AR PentestGPT: The Art of Hacking with Words

  PentestGPT: The Art of Hacking with Words regresa a DEF CON, esta vez como un workshop completamente práctico. Luego de presentar el año pasado cómo los LLMs comenzaban a transformar la seguridad ofensiva, esta nueva edición lleva la experiencia al siguiente nivel: no solo será escuchar de este proyecto, lo podrás crear y ejecutar en tu entorno favorito.

  Durante 100 minutos, los participantes trabajarán directamente con PentestGPT y PentestGPT 2.0 para aprender cómo integrar IA en flujos reales de pentesting. A través de ejercicios guiados y escenarios ofensivos realistas, explorarán reconocimiento asistido por IA, generación de attack paths, análisis de hallazgos, prompt engineering ofensivo y validación técnica de resultados.

  Más allá de usar una herramienta, el workshop propone un cambio de mentalidad: entender cómo pasar de ejecutar tareas manuales a orquestar inteligencia ofensiva.

  ResearchersMatias Armándola AR

  Presented at La Villa Hacker — DEF CON 2026

  View official session
• August 7, 2026 Workshop AppSec & Supply Chain Payment Village — DEF CON 2026 Dan Borgogno AR Introduction to the Wall of Wallets Challenge

  Wall of Wallets is a hands-on CTF challenge where the goal is simple: collect as many mock payment card details as you can from other participants. Using intentionally vulnerable services in a safe, isolated environment, you'll learn how common implementation mistakes expose sensitive data.

  Compete to top the Wall of Wallets leaderboard while discovering the real-world techniques attackers use and, importantly, how to defend against them.

  ResearchersDan Borgogno AR

  Presented at Payment Village — DEF CON 2026

  View official session
• August 7, 2026 Workshop Malware & Reversing Red Team Village — DEF CON 2026 Mauro Eldritch AR·Nelson Rafael Colón Merán DO Haetae: An Agent to Takedown North Korean C2 Servers

  In this talk, we introduce Haetae, an agent designed to profile, identify, and exploit C2 frameworks used by North Korean malware.

  Haetae supports both automated and interactive modes, allowing analysts to map and understand adversary infrastructure with different levels of control. It is built around a flexible rule-based system, enabling users to extend and refine detections as new patterns and frameworks emerge.

  In this first release, we will walk through real-world cases where Haetae was used to identify and take down infrastructure associated with Mach-O Man and POWerful Armadillo.

  We will also release a safe emulator of both malware C2 servers, allowing researchers and newcomers to experiment with Haetae in realistic, controlled environments without risk.

  This is a highly technical talk but can be enjoyed by both beginners and seasoned threat hunters.

  ResearchersMauro Eldritch AR·Nelson Rafael Colón Merán DO

  Presented at Red Team Village — DEF CON 2026

  View official session
• August 7, 2026 Workshop Offensive Security Red Team Village — DEF CON 2026 Rafael Felix BR Shapeshifting C2: Applying DAITA Traffic Shaping to Defeat DPI and DLP Detection

  ML-based DPI and DLP solutions have made signature evasion insufficient. Modern enterprise detection analyzes packet size distributions, inter-packet timing, and burst patterns to fingerprint C2 traffic regardless of how well headers and TLS certificates are spoofed.

  This talk demonstrates how three techniques from the VPN privacy research space - originally developed to defeat traffic analysis against Tor and Mullvad, can be adapted for offensive C2 operations:

  1. Constant-size packet morphing: forcing all frames to a fixed MTU-aligned size removes the packet-size fingerprint that ML classifiers depend on;

  2. Cover traffic injection: dummy packets sent at Poisson-distributed or adaptive intervals make real beacon timing statistically invisible;

  3. Probabilistic pattern distortion: Maybenot-inspired state machines inject cover bursts alongside real packets, destroying recognizable C2 sequences;

  We then will see how layering these techniques with application-layer malleable profiles (traffic impersonating Microsoft Graph, Slack, or Okta) and bonus: indirect transports like cloud storage dead-drops (S3, OneDrive) creates C2 channels that defeat detection at every layer simultaneously - behavioral, flow, and application.

  Attendees leave with a concrete mental model for building evasion stacks that go beyond header manipulation.

  ResearchersRafael Felix BR

  Presented at Red Team Village — DEF CON 2026

  View official session
• August 7, 2026 Workshop AppSec & Supply Chain Red Team Village — DEF CON 2026 Natan Morette BR Quality over Quantity: How to Publish CVEs that Actually Matter

  This hands-on tactic focuses on making CVE discovery and publication practical, accessible, and repeatable. Instead of treating vulnerability research as an individual or highly specialized skill, this session introduces a structured workflow that attendees can immediately apply.

  Attendees will follow a step-by-step approach to:

  • Explore a scoped attack surface
  • Apply a curated vulnerability pattern checklist
  • Identify and validate a potential vulnerability
  • Structure the finding in a simplified CVE-style format

  All materials are prepared in advance to ensure the activity can be completed within 30–60 minutes and repeated across multiple waves of participants.

  This tactic emphasizes doing over theory, helping attendees leave not just with knowledge, but with a clear, reusable method for identifying and documenting vulnerabilities and a better understanding of how CVE publishing can be integrated into team workflows.

  ResearchersNatan Morette BR

  Presented at Red Team Village — DEF CON 2026

  View official session
• August 7, 2026 Workshop Offensive Security La Villa Hacker — DEF CON 2026 Juan Camilo Palacio CO·Andrés Restrepo CO De la nube a la corona: Uso indebido de la identidad híbrida en Azure DevOps Azure, AD, AWS y EKS

  Demostrar cómo las relaciones de confianza entre entornos híbridos de Azure DevOps, Microsoft Azure, Active Directory, Amazon Web Services y Kubernetes pueden ser abusadas por un atacante para realizar movimientos laterales, escalamiento de privilegios y compromiso de identidades utilizando herramientas y funcionalidades legítimas de administración cloud en infraestructuras hibridas corporativas.

  ResearchersJuan Camilo Palacio CO·Andrés Restrepo CO

  Presented at La Villa Hacker — DEF CON 2026

  Speaker page (Juan Camilo Palacio) Speaker page (Andrés Restrepo)
• August 3, 2026 Workshop Incident Response BSides Las Vegas 2026 Rodrigo Montoro BR AWS Principal Threat Hunting: Behavioral Baselining for Malicious Activity

  Cloud security encounters attacks that elude standard detection. In AWS, unauthorized access keys are a common cause of breaches. The vastness of AWS, over 450 services and 19,000 API actions, complicates threat visibility and exposes gaps in traditional tools. This workshop empowers participants with direct, hands-on experience using the AWS Threat Hunter tool to improve threat detection.

  Attendees will focus on building behavioral baselines and leveraging data-driven analysis to detect subtle AWS principal anomalies, enabling more precise detection than traditional event monitoring. Attendees will move beyond standard techniques by building multi-stage detection pipelines that create individualized baselines for each IAM principal and systematically flag personalized deviations, linking outliers directly to risks.

  ResearchersRodrigo Montoro BR

  Presented at BSides Las Vegas 2026

  View official session