Back to BOSS
BOSS Advanced Offensive Security Training

Augmented Pentesting.

LLM-Assisted 1-Day Vulnerability Research & Exploitation

An advanced, hands-on course for pentesters who want to bring focused vulnerability research and exploit development into real-world engagements. Two days investigating a real open-source product, using LLMs to support the research, not replace the researcher.

Dates25–26 February 2027
DeliveryLive remote, English/French
Duration2 days / 14 hours
SeatsMaximum 6
Request pre-registration Pre-registration by email. No payment required at this stage.
01 / About the training

Why this course?

LLMs have changed the amount of technical research an experienced pentester can realistically perform in the time available for a security assessment. Patch analysis, code review, debugging and proof-of-concept development can now be approached more efficiently, making focused 1-day research increasingly practical during ordinary engagements.

Attackers have access to these capabilities too. Our job is to assess what a capable attacker could actually achieve, including exploitation paths that automated scanning and conventional testing may miss.

Efficiency matters for two reasons: it leaves more time to investigate complex weaknesses and makes it possible to cover more of the attack surface. The value goes to the client through deeper research and broader coverage.

None of this removes the need for technical expertise. The researcher's experience, judgment and persistence still determine which leads are worth pursuing and whether a hypothesis survives testing.

02 / Format

Real software.
Real research.

The instructor will select a recent, official open-source product shortly before the training. The target will be revealed on the first day. Each participant will deploy the same version and commits in an isolated local Docker environment using a supplied deployment script.

There is no prepared CTF path and no promise of a final exploit. Participants will examine advisories, issues, commits and source code, reproduce suspicious behavior and pursue promising primitives. Findings and approaches will be discussed as a group, with room for independent investigation.

LLMs from different providers can be used through interactive Chat and Work workflows to assist with analysis, hypothesis generation and debugging. Manual testing, scripts, debuggers and technical decisions remain under the pentester's control.

03 / Course outline

Two days of technical investigation

Thursday, 25 February 2027

Day 1: Research & Primitives

Research strategy and environment

Introducing the methodology, deploying the target and understanding its architecture and attack surface.

1-day vulnerability intelligence

Reviewing advisories, GitHub issues, fixes, pull requests, crash reports and other leads.

Multi-LLM-assisted investigation

Source-code analysis, patch review, hypothesis comparison and prioritization.

Reproduction and debugging

Building experiments and determining whether suspicious behavior yields useful primitives.

Friday, 26 February 2027

Day 2: Exploitation & Validation

Research review

Evaluating evidence and deciding which leads justify further investigation.

Exploit development

Iterating on proofs of concept with manual testing, reverse engineering and debugging.

Attack-chain investigation

Exploring the security impact of individual primitives and possible combinations.

Client-context demonstration

A brief demonstration of local prompt sanitization against a separate, fictitious client Docker environment.

Technical debrief

Reviewing successes, dead ends and how to apply the methodology during authorized engagements.

04 / Audience

Who should attend?

This is an advanced course for experienced pentesters, red team operators and vulnerability researchers who are comfortable investigating technical problems without guided solutions.

Expected background

  • Practical pentesting and exploitation experience.
  • Ability to read source code and troubleshoot independently.
  • Comfort with Linux, Git, Docker and command-line tools.
  • Familiarity with application exploitation; basic memory-corruption and native-debugging concepts.

Prior LLM experience is useful, but not required. Participants use their own LLM accounts. Access to more than one model is recommended; no specific provider is mandatory.

05 / Instructor

From real engagements.

Jean-Marie Bourbon

Founder of Bourbon Offensive Security Services (BOSS), Luxembourg. Offensive security practitioner and vulnerability researcher with 18 years of field experience, OSCP and OSCE certified.

The methodology draws on authorized research and exploitation work involving software including OpenCTI, libheif and PostgreSQL. These are examples of prior work, not an indication of the product used during this course.

Read the libheif research

06 / Registration

Join the first session.

990 € excl. VAT / participant, both days

25 and 26 February 2027. Live remote training in English, with no more than six participants.

To request pre-registration, email BOSS with your name, organization (if applicable) and a short description of your offensive security experience. BOSS will follow up with the registration details and invoice.

Payment is by bank transfer and must be received by 25 January 2027. A place is confirmed once payment has been received. Cancellation and postponement terms will be provided before any binding registration.

Request pre-registration by email

Training materials, deployment scripts and a local prompt-sanitization tool are included. Personal LLM subscriptions are not included.