Learn
Worked intrusions: what the evidence said, in what order, and what turned out to be wrong on the way. Every one of ours has a discussion under it, because the disagreements are usually worth more than the write-up.
Some of these describe invented organisations and invented incidents, written so there is something to practise on. Those are marked Constructed on the card and again at the top of the piece. The techniques, artifact locations and reasoning in them are real; only the company and the people are not.
A full DFIR report on a constructed ransomware incident, written the way one gets handed to a board. Three hours twenty from first execution to encryption, four separate points where it could have been stopped, and the write-up that follows.
A business email compromise where the obvious finding is a decoy. Ninety minutes on Microsoft 365 audit logs, OAuth consent grants, and why resetting the password does not end this incident.
A constructed intrusion written as a DFIR report. You get the same artifacts an analyst gets on day one, in the order they arrive, and you build the timeline yourself. The company is fictional; the techniques, event IDs and analytic reasoning are not.
Every other lab here asks for a virtual machine. This one asks for a terminal. Pre-recorded event logs, a cross-platform parser, and the same reasoning, on whatever laptop you already have.
The first lab for a machine of your own. You create the evidence yourself, delete a file from it, and then recover the file you deleted, which is the only way to be certain the recovery worked.
Windows services run as SYSTEM and start before anyone logs in, which makes them the most valuable persistence on the box and the hardest to spot among the two hundred already there. Part three of the persistence series.
Plant a scheduled task with Atomic Red Team, then find it four ways and work out which one survives an attacker who names their task properly. Closes the loop on the persistence you found on the domain controller in the Northwind case.
WMI event subscriptions run code on a condition you choose, survive reboots, live nowhere in the filesystem, and log nothing by default. The last of the persistence series, and the one that changes how you think about the other three.
A first investigation for anyone who has read about persistence but never gone looking for it. You will plant a Run key with Atomic Red Team, then find it three different ways and work out which of the three you would actually trust in production.
A beginner's guide to MITRE ATT&CK that does not open with the matrix. What it is, what it is not, why the giant grid is the worst way to meet it, and how to use it on a Tuesday afternoon rather than in a strategy deck.
CVE-2026-73570 needs no credentials and no user interaction. A crafted SMTP request executes commands as the zimbra user. Mail servers accept connections from strangers by definition, which is what makes pre-auth bugs in them a different category of problem.
CVE-2026-60004 turns repository write access into shell execution as the Gitea service account. The interesting part is not the injection. It is that Git hooks are executable files sitting inside a directory your developers write to all day.
Between 25 and 27 August, KEV gained CVEs from 2015, 2021 and 2022, most of them local and none of them remote entry points. A catalogue of what attackers are actively using is telling you about the second stage, and about how much end-of-life software is still running.
A lab reconstruction of a user-level persistence case, and why the Run key timestamp is the least useful thing in the artifact set.
Four variants of the same PowerShell technique, three of which walked past a rule that looked fine on paper. Notes from a purple team afternoon.
Entropy, subdomain cardinality and query-rate shape — three statistics that find a covert channel faster than any rule matching a known tool.
Six months of a detection lab that was impressive to look at and useless to work in, and the four changes that fixed it.
Volunteer analysts publishing real intrusions
The closest thing the field has to a case-law library. Each report walks an intrusion from initial access to impact with the actual artifacts, and most carry Sigma rules at the end. Read one a week and your instincts change.
US Cybersecurity and Infrastructure Security Agency
Joint advisories carry TTPs and detection guidance for named actors, and being US government work they are public domain, so you can reuse the content in your own runbooks without asking anybody.
A managed detection provider, published annually
Ranks the techniques they actually saw across thousands of environments, with detection guidance for each. The value is the ranking: it tells you what to build detection for first, which is a different question from what is interesting.
Cisco Talos; Palo Alto Unit 42 publishes comparably
Malware analysis with the reverse engineering shown rather than summarised. Useful when you have a sample and need to know what a family does, rather than when you have a host and need to know what happened.
Incident responders publishing from engagements
Long-form intrusion analysis, usually of the more capable end of the spectrum. Their M-Trends report each year is the best single read on how dwell time and initial access are actually trending.
Two teams that publish quickly during live campaigns
Where to look during an unfolding event. Both tend to publish artifacts and indicators within a day or two of a mass-exploitation campaign starting, which is when you need them rather than a month later.
