Yinkozi
Contact
services / red team

Adversarial simulation, designed for the customer who actually has adversaries.

Goal-based, multi-vector red team engagements built around how a real attacker would approach the customer — not a generic kill-chain exercise. Custom payloads, custom OSINT, custom command-and-control infrastructure. Hardware implants where the threat model includes them.

01 / methodology

Modelled on the actual adversary, not the framework.

Every engagement begins with the threat actors the customer is actually concerned about — sophisticated criminal groups for tier-1 banks, state-aligned actors for sovereign customers, insider scenarios for high-trust environments. We profile their tradecraft from open and private intelligence sources and model the engagement after them.

That is different from "an emulation of TTPs from MITRE ATT&CK". MITRE is a useful taxonomy. It is not a substitute for thinking about the specific human or group that is going to come after the customer.

We design the engagement to look the way a real attack would look — including the parts that the customer's blue team is not currently watching for.

Red team multi-vector attack — initial access converges into the customer environment, then lateral movement to goal — initial access vectors — customer perimeter — lateral & goal 01 Targeted phishing 02 Supply-chain abuse 03 Physical & insider 04 Exposed-service exploit 05 Watering-hole / OSINT PERIMETER Customer environment Identity · Network · Endpoint Custom payloads land here — assumed-breach starts here L1 Privilege escalation L2 Cross-system pivot L3 Persistence GOAL Objective defined per engagement
02 / what we don't do

The shape we don't ship.

  • No SaaS BAS dressed as red team

    Breach-and-attack-simulation platforms have a place in continuous detection validation. They are not red-team engagements. We are clear about which work is which.

  • No off-the-shelf adversary emulation

    Cobalt Strike profiles and out-of-the-box C2 frameworks are detected by every modern EDR. We build custom infrastructure and custom tooling per engagement — that is the point.

  • Local LLMs accelerate, they do not lead

    We do not deploy LLM agents to "automate red team". The decisions a real attacker makes — what to escalate, when to wait, what to leave alone — are not in the capability of a model that can hold a multi-week operation in context. We do use local LLMs we run in-house — to accelerate OSINT, scaffold infrastructure, and pattern-match across our threat library. Customer-sensitive material never goes to a commercial SaaS model.

  • No "compliance red team"

    We do not run engagements where the deliverable is a stamp for an audit. Every engagement is goal-based and has a real adversary model behind it.

03 / capabilities

What we bring across the kill chain.

Initial access

Targeted phishing with custom infrastructure, watering-hole, supply-chain abuse, exposed-service exploitation, OSINT-driven pretext development.

Custom tooling

Custom implants, custom C2 channels (HTTPS, DNS, custom-protocol), in-house EDR-bypass research, dropper development.

Physical & hardware

Physical access scenarios, hardware implants, badge cloning, USB drop variants, building OSINT, social engineering against on-site staff.

Lateral movement

AD / Entra ID attack-path analysis, Kerberos abuse, cross-tenant pivot, custom credential-harvesting, on-network discovery without ringing every alarm.

Cloud privilege escalation

IAM-driven escalation across AWS / Azure / GCP, cross-account abuse, identity-federation attacks, secrets sprawl exploitation.

Objective & impact

Goal-based — payment fraud, data exfiltration, ransomware-pre-condition, sovereign-data access, agent-banking compromise, signed-update poisoning.

04 / engagement shape

Four shapes of engagement.

Full-scope red team. Threat-model design, OSINT, custom infrastructure, multi-vector execution, post-engagement debrief and purple-team replay.

Assumed-breach simulation. Customer grants a beachhead; we exercise lateral movement, privilege escalation, and goal achievement against the existing detection stack.

Long-window persistent simulation. Low-volume, deliberately patient. Designed against blue teams that have already passed the first two engagement shapes.

Purple-team integration. Follows any engagement. We replay the operation with the customer's blue team to harden detection where the gaps were.

Duration depends on the customer's scope, footprint, and threat profile. We scope these conversations against the actual surface.

05 / start a conversation

Tell us about the adversary you're modelling against.

email