Mike Sosa
Back to portfolio

One attack chain, four labs

How my pinned labs fit together, from the phishing link to the compromised account to the controls that stop it from happening again, plus where AI agents add a new path in.

Side by side, these labs follow one common cloud-identity pattern: a phishing URL gets past a filter, an attacker signs in as the user, responders have to contain it in the right order, and identity controls make the next attempt harder. A fourth lab covers a newer entry point - instructions hidden in content an AI agent reads, not a link a person clicks.

How to read this: the arrows are thematic, not one incident I traced end to end. Labs don't share data or code. Each one is a lab (synthetic or public data); status and limits live in the repo and in the node captions below.

Labs in the chain

  1. 1. Phishing gets through

    URL-only phishing detectors attacked with cheap, realistic edits, and what defenses buy and cost. Public data; no URL is visited.

    Repo (opens in new tab) Case study

  2. 2. Account is compromised

    Draft IR runbook for a compromised Entra ID account, with KQL queries and a tabletop. Queries untested against a tenant.

    Repo (opens in new tab) Post: Order of Operations

    • Thematic: lessons from the response inform which controls to enforce → Harden identity

  3. 3. Harden identity

    Entra ID controls in Terraform: Conditional Access in report-only first, break-glass exclusions, PIM. Static checks pass; not yet applied to a tenant.

    Repo (opens in new tab) Post: Report-Only First

  4. 4. AI agent is manipulated (parallel path)

    Fictional Fabrikam helpdesk agent tested for prompt injection that leads to tool misuse, and how much simple prompt-level defenses help. Synthetic data only.

    Repo (opens in new tab) Portfolio

    • Parallel: agents act with identities too, so least privilege still applies → Harden identity
Attack chain map of four labs Four labs in a thematic attack chain: phishing leads to a compromised account, response informs identity hardening, and a parallel path covers prompt injection of AI agents. 1 Phishing gets through phishing-detector… 2 Account is compromised entra-account-compromise-ir 3 Harden identity azure-entra-identity-hardening 4 AI agent is manipulated prompt-injection-lab
1 Phishing gets through

Thematic: a phish that gets through can lead to a stolen sign-in

2 Account is compromised

Thematic: lessons from the response inform which controls to enforce

3 Harden identity
Solid arrows show a common attack sequence. Dashed arrows show a related idea. No arrow means one lab's results prove anything about another lab.