The question Goran Svetlecic says he hears most in post-incident briefings is the same one every time: “How could we have stopped this?” What usually follows, he writes, is the hard realisation that the paths were there long before the compromise, walkable the whole time, but were never surfaced in time to act on.
That sentence reframes the whole review. The failure was not detection. The failure was that nobody walked the map while walking it was still cheap.
One map, two cost functions
Attack path management maps the environment exactly as an adversary sees it: a continuous graph connecting any initial foothold to the crown jewels. The paths here are identity paths: Active Directory and Entra, extended by certificate templates, group policy links, and the graph of accounts, groups, sessions, and rights that bind them. Identity is how an attacker actually moves. Every implant, every foothold that outlasts the first one, is reached by walking that graph.
An adversary does not need a million attack paths. They need one. And most of the graph is uninteresting. What matters is the small set of edges that collapse distance to Tier 0, the choke points through which a disproportionate number of paths run. Cut one of those and you do not remove a single attack. You remove the entire fan of attacks that depended on traversing it.
Peacetime work is not insurance. It is the only phase in which you hold every advantage at once.
Peacetime: the work is cheap
In peacetime you close paths in priority order, on your own schedule. The cost function is completeness: find every path that matters, rank it by choke-point value, remediate systematically. Tempo is yours, so you can test a change, validate that nothing legitimate broke, document, roll back, and fix again. Mistakes are cheap: a misjudged remediation in peacetime is a ticket and a short outage in a maintenance window. It is not a breach.
And the work compounds. Every path cut in peacetime is a permanent contraction of the surface the next intruder inherits. The graph the attacker eventually loads is smaller, flatter, and poorer in choke points than the one you started with. The surveyor who walks the graph a hundred times in peacetime is the person you want holding the map when it turns into wartime.
Wartime: the dilemma
Wartime begins the moment the graph stops being yours alone. The adversary is on it, moving along edges you may not have mapped, and watching what you do. Now the cost function inverts: triage under observation, contested tempo, expensive and visible mistakes, and a graph mutating while you work, as their lateral movement creates new edges and your remediation closes others.
Here is the dilemma at the center. Day three of an intrusion: the team has a known backdoor and a path from the foothold, through a server admin group, across one misconfigured ACL, into Tier 0. Cut the known backdoor now and you close the channel you can see, but a backdoor you have identified is rarely the only one, and severing it tells a capable adversary their primary channel is burned. A financially motivated operator may monetize destructively before losing access: the thief becomes an arsonist. A patient espionage actor may go dormant and return in six months through a door you still have not found: the ghost. Leave the persistence in place and watch it, and the adversary keeps moving toward the Tier 0 path anyway. Cut and you may detonate them or lose them; wait and they arrive regardless.
What determines the outcome is almost entirely what was done before the intrusion began. If that ACL had been found and cut in peacetime, the path never reaches Tier 0 and the worst case is a contained incident in the user tier, no matter what the adversary does with their backdoors.
The five path questions
So the blocking control is not a product. It is a peacetime cut, made while it is free. Five questions, asked about the map before anything is on fire, decide which cuts matter:
- Which edges collapse the most distance to Tier 0, carrying the largest fan of distinct routes?
- Which of those cuts are cheap right now, in effort, breakage risk, and documentation?
- What legitimate work depends on each edge, and what breaks if it is cut?
- Who owns the edge, and who can approve the cut without a committee?
- What evidence proves the cut held, after the next change window?
Answer those, cut the highest-fan choke points first, and repeat. You do not need the whole map perfect. You need the few edges that collapse the most distance, closed while closing them is still free.
What to do Monday
Enumerate the paths to your crown jewels as they exist right now. Rank them by how many distinct routes each edge carries. Cut the highest-fan choke points first. Document what broke and what was unknown while you still control the conditions. Then walk the graph again, because the first pass will surface paths you did not know were there.
Peacetime buys optionality. Wartime spends it, almost always at a loss. The incident in the briefing was not unstoppable. It was unpriced. The paths were there, walkable, waiting. The only question that mattered was whether anyone cut them while cutting was still free.
This is the same discipline CSI applies to agent authority: map the authority graph, cut the choke points, decide before the clock starts.
Source: Goran Svetlecic, “Attack Path Management in Peacetime and Wartime,” LinkedIn, October 27, 2026: https://www.linkedin.com/pulse/attack-path-management-peacetime-wartime-goran-svetlecic-cissp-eqeif/