Ransomware attack chain: exposed service to encryption

Reporting on ransomware usually covers only the last scene: the extortion note on the screen, the line that stopped, the sum demanded. But the decisive part of the ransomware attack chain happens one to three weeks earlier, and it is far more mundane than most people expect — no zero-day, no particular sophistication.

This article breaks the chain into stages and maps each one to a MITRE ATT&CK tactic, so you can walk your own environment through it stage by stage. By the end you should be able to answer a concrete question: if someone started on us today, where would they get in, and how much work would it take them to reach the encryption stage.

The ransomware chain on a timeline, with detectable signalSix stages in order along a time axis: enumeration, initial access, internal discovery, lateral movement and privilege escalation, stopping backups and exfiltrating data, then encryption. The bars underneath show the detectable signal at each stage: almost none for the first three, concentrated in the last two.Enumerationongoing, beforeInitial accessday 0Internaldiscoverydays 1–3Lateral movement& escalationdays 3–14Backups killed,data exfiltrateddays 14–20Encryptionhourstime →Below: strength of detectable signalThe attacker is inside for two to three weeks; nearly all the detectable signal falls in the last day or two
  • Almost no detectable signal
  • Medium signal
  • Unmissable signal
FIG 01 The map for the whole article. Above: six stages and how long each one lasts. Below: the signal each stage produces — nearly all of it in the last two bars, by which point the backups are already gone.
Text version of this diagram

The six stages of the ransomware chain on one time axis, with the detectable signal at each stage drawn underneath.

The six stages

  • Enumeration: ongoing, before the attack.
  • Initial access: day 0.
  • Internal discovery: days 1–3.
  • Lateral movement and privilege escalation: days 3–14.
  • Backups killed and data exfiltrated: days 14–20.
  • Encryption: complete within hours.

The signal bars below

  • The first three stages have the shortest bars: almost no detectable signal.
  • Lateral movement and escalation rises to a medium signal.
  • Killing backups and encryption are the two tallest bars — unmissable signal.
  • The conclusion: the attacker is inside for two to three weeks, and nearly all the detectable signal falls in the last day or two.

The chain does not start with encryption, it starts with finding an address that answers

The chain opens with initial access (TA0001), and the work preceding initial access is enumeration: sweep public address ranges, record which addresses and ports respond, and fingerprint what the response looks like — which product, which version.

That work is automated, continuous and extremely cheap. It is not aimed at you; it is aimed at every address that answers. The size of your attack surface is simply how many times you appear in that internet-wide list.

The list is then joined against vulnerability data. The CISA Known Exploited Vulnerabilities catalog is the most practical version of that join: every entry has been confirmed as exploited in the wild, and remote access appliances have always been heavily represented in it.

One sentence worth keeping: an address that does not answer never enters the list. Everything below rests on that.

The three entrances that account for most intrusions

The ways in do not vary much. Broadly, there are three.

EntranceWhat the attacker needsTechnique
A known flaw in a public serviceA reachable endpoint and an exploitT1190
Working credentialsBought or guessedT1078, T1110
A person reading emailA convincing enough messageT1566

Exploiting a public-facing application (T1190) takes the least effort. VPN gateways, remote desktop portals, file transfer systems, management consoles — anything open to the whole internet is simultaneously a service entrance and an attack entrance.

Valid accounts (T1078) is the hardest to detect, because the account is genuine. Credentials may come from brute force (T1110), from a breach on some unrelated site, or straight from an initial access broker — there is now an entire market in working VPN credentials for named companies.

Phishing (T1566) and social engineering bypass technical controls and deal with people directly. Training lowers the rate but never to zero, so no architecture should assume nobody clicks.

All three paths share an endpoint: the attacker holds a foothold inside the network, and in most cases the credentials they hold are legitimate.

Three entrances, one footholdA known flaw in a public service, working credentials, and a person reading email all lead to the same internal foothold — and in most cases the attacker holds legitimate credentials. Everything after the foothold, from discovery through lateral movement to encryption, requires breaking through nothing further.Known flaw in a public serviceT1190Working credentialsT1078 · T1110A person reading emailT1566Internal footholdT1078usually holding legitimate credentialsEverything afterInternal discoveryLateral movementEncryption and leakAll three paths end in the same place, and nothing after it has to be broken through
  • Attack path
  • Holding legitimate credentials
FIG 02 Each entrance has its own defence, and all three converge on one point. Nothing inside the box to the right of that junction has to be broken through again — which is why holding the entrances cannot be the whole plan.
Text version of this diagram

Three entrances lead to the same internal foothold, and nothing after that foothold has to be broken through again.

Left — the three entrances

  • A known flaw in a public service, ATT&CK T1190.
  • Working credentials, T1078 and T1110.
  • A person reading email, T1566.
  • All three arrows converge on a single point.

Centre — the internal foothold

  • The junction is labelled "internal foothold", annotated with T1078.
  • Below it, highlighted: in most cases the attacker holds legitimate credentials.

Right — everything after

  • Inside a dashed box, in order: internal discovery, lateral movement, encryption and leak.
  • The conclusion: all three paths end in the same place, and nothing after it has to be broken through.

What happens next: from one host to the whole domain

With a foothold established, the first job is orientation — network service discovery (T1046): sweep the internal ranges, list reachable hosts and services, locate the domain controllers and the backup servers.

Traditional networks barely resist this step. A VPN issues an internal address after authentication, so what the attacker can see equals what that segment can see. They then use remote services (T1021) — RDP, SMB, WinRM, all already open internally — to move host by host, the stage known as lateral movement (TA0008).

In most real incidents, days to weeks pass between initial access and domain administrator privileges. During that time the attacker’s activity looks a great deal like a new IT hire finding their way around, which is precisely why it goes unnoticed.

The architectural dividing line sits here: is reachability determined by network location, or by identity? In the first case a stolen credential is worth an entire segment; in the second it is worth only the handful of services that identity was granted. The mechanism behind that difference is described in how a zero trust network actually works.

Why antivirus and backups still end in an extortion note

Because encryption is the last scene, and by then the attacker is an administrator.

  • Backups are dealt with first. A backup host in the same domain, using the same administrative credentials, mounting a writable share, is inside the attacker’s reach. Deleting snapshots and stopping backup services (T1489) before encryption is standard practice.
  • Antivirus only fires at the loudest step. The preceding weeks run on legitimate tools and legitimate accounts, which produce very few detectable signals.
  • Encryption is just the closing move. Data encrypted for impact (T1486) typically completes within hours, and most cases now exfiltrate data first — so even a clean restore leaves you with a data breach.

“Can we recover” and “will this hurt us” are therefore two different questions. Backups answer the first one only. For preparation, response and recovery process, NIST SP 800-61 is the reference; regional advisories and threat intelligence come from bodies such as ENISA and, in Taiwan, TWCERT/CC.

Line the stages up against the cost of defending each one and the answer is not subtle.

StageUsual defenceThe difficulty
EnumerationDo not answer unauthorised requestsRequires architectural change, once
Initial accessPatching, MFA, trainingContinuous, never reaches 100%
Internal discoverySegmentation, microsegmentationRule sprawl, high maintenance
Lateral movementDetection and responseNeeds staffed, tuned monitoring
EncryptionBackups and restore drillsLimits loss; the event happened

The higher up, the cheaper — because higher up you are defending a single variable, reachability. Lower down you are defending against attacker behaviour, and behaviour has unlimited variations.

The cost of defending each linkFive stages from top to bottom. At the top, enumeration, the defence covers one variable — reachability — and is a one-time change. Below it come vulnerabilities and credentials, segments and rules, then attacker behaviour, with the number of variables and the maintenance cost rising at each level. At the bottom, encryption, only the loss can be reduced.What you are defendingCost →EnumerationReachability: one variableone-timeInitial accessFlaws and credentials: movingcontinuousInternal discoverySegments and rules: growinghigh upkeepLateral movementAttacker behaviour: unlimitedstaffed watchEncryptionAlready happened; limit lossloss incurredHigher up there are fewer variables to defend; lower down you defend behaviour, and behaviour has unlimited variations
  • One-time cost
  • Continuous upkeep
  • Staffed watch or after-the-fact recovery
FIG 03 Bar length is cost. The top row defends one variable, reachability, and is changed once. The bottom two defend attacker behaviour, and behaviour has no upper bound on variation.
Text version of this diagram

Five stages from top to bottom, each row naming what is defended and at what cost; cost is the length of the bar, growing downward.

The five rows

  • Enumeration: defends reachability — one variable — at a one-time cost.
  • Initial access: defends flaws and credentials, which keep moving, at a continuous cost.
  • Internal discovery: defends segments and rules, which keep growing, at high upkeep.
  • Lateral movement: defends against attacker behaviour, unlimited in variation, needing a staffed watch.
  • Encryption: already happened; only the loss can be limited.

What to take from it

  • The top row has the shortest bar in the safe colour; the bottom two are longest and in the risk colour.
  • Higher up there are fewer variables to defend; lower down you defend behaviour, and behaviour has unlimited variations.

That is the concrete relevance of zero trust architecture to ransomware. It does not claim to stop phishing; it collapses the first and third stages at once. Services do not answer unauthorised requests, so they never enter the enumeration list; authentication grants a single service rather than a segment, so the lateral path is not there to begin with. How the architectures differ on those two points is set out in comparing traditional VPN, gateway-based ZTNA and SDP.

Five things you can check today

No tooling and no project budget required.

  1. List every port your organisation exposes publicly and ask of each one whether it truly has to be open to the world. The list is usually longer than IT remembers.
  2. Confirm MFA is enforced on all remote access, vendor and contractor accounts included.
  3. Verify at least one backup copy is offline or immutable, and reachable without day-to-day administrative credentials.
  4. Run an actual restore, rather than reading a successful-backup log.
  5. Take an ordinary employee account and ask how many hosts it can reach. If the answer is “the whole segment”, the lateral path is already paved.

If items 1 and 5 both come back badly, the priority is architecture rather than procurement — which is what choosing a VPN replacement and planning the migration covers. Establishing how far an attacker could actually get is the scope of red team versus penetration testing.

Terms and further reading

TermIn one sentenceSource
RansomwareMalware that encrypts data and demands paymentRansomware
Initial accessThe stage of gaining the first footholdMITRE ATT&CK TA0001
Attack surfaceThe sum of externally reachable entry pointsAttack surface
Known Exploited VulnerabilitiesVulnerabilities confirmed exploited in the wildCISA KEV
Exploit public-facing applicationBreaking in via a flaw in an exposed serviceMITRE ATT&CK T1190
Valid accountsLogging in with real credentialsMITRE ATT&CK T1078
Brute forceRepeated credential guessing until one worksMITRE ATT&CK T1110
Initial access brokerA market intermediary selling network accessInitial access broker
PhishingForged messages that harvest credentialsMITRE ATT&CK T1566
Social engineeringManipulating people rather than systemsSocial engineering
Network service discoveryScanning internally for reachable servicesMITRE ATT&CK T1046
Remote servicesMoving via remote protocols already in useMITRE ATT&CK T1021
Lateral movementSpreading from one internal host to the nextMITRE ATT&CK TA0008
Service stopHalting backup or security services firstMITRE ATT&CK T1489
Data encrypted for impactEncrypting victim data to force paymentMITRE ATT&CK T1486
Incident handling guidePreparation, response and recovery processNIST SP 800-61
Zero trust architectureTrust not derived from network locationWhat zero trust architecture is

The technique identifiers above map directly onto detection rules, which makes them a quick way to find the stages where you currently have no signal at all. If you would like this chain assessed for feasibility against a specific environment, get in touch.

Back to the blog

What SDP v3.0 Changed: SPA Is No Longer the Definition

SDP v3.0 demotes single packet authorisation from definition to option. What the new guide actually changed, why SPA struggles on IoT and cloud functions, where the line between SDP and microsegmentation really falls, and what to do first.