Ransomware is malware that encrypts files or locks systems and then demands a payment to restore access. Attackers who deploy it today almost always steal the data first, so a decryption key no longer ends the incident.
In February 2024, the attack on Change Healthcare cost UnitedHealth Group $3.1 billion in that year alone and exposed data belonging to roughly 190 million people. The attackers broke no cryptography to get in. They used a Citrix portal with no multi-factor authentication.
This page covers what ransomware is, how a modern attack unfolds, which software weaknesses open the door, and how to find and fix those weaknesses before someone else finds them.
What is ransomware?
Ransomware is a family of malware built for extortion. It makes an organization’s own data unusable, then sells back the ability to use it. The encryption is the visible part; it is also the last part.
The category splits in two. Cryptor ransomware encrypts files and leaves the system running, so victims can read the ransom note but not their documents. Locker ransomware blocks the device without touching the files. Cryptors dominate corporate incidents because they scale: one payload can reach thousands of hosts through a single management console.
What changed is the business around it. The 1989 PC Cyborg trojan locked a single machine and asked the user to renew a license for $189. A ransomware incident in 2026 involves at least three parties who never meet, and the payload is the cheapest part of the operation.
Verizon’s 2025 Data Breach Investigations Report found ransomware present in 44% of the breaches it reviewed, up from 32% the year before. In the same dataset, 64% of victim organizations did not pay, and the median ransom paid fell to $115,000.
How does a ransomware attack work?
A ransomware operation is a chain, and encryption sits at the end of it. Before any file is locked, an attacker has to get in, move, find what matters and copy it. Each step is a chance to stop it while it is cheap.
Stage | What happens |
|---|---|
Access | An access broker or affiliate gets in through a stolen credential, a phishing lure or an unpatched internet-facing service |
Escalation | The intruder raises privileges and disables logging, backups and endpoint defenses |
Lateral movement | They map the network and reach domain controllers, file shares and backup servers |
Data exfiltration | Sensitive data is copied out, often over days or weeks, so the leak threat survives any restore |
Encryption | The payload runs across as many hosts as possible, usually outside business hours |
Extortion | The victim is given a deadline, a payment address and a leak site countdown |
At Change Healthcare the attackers spent nine days moving laterally before deploying anything, according to our review of the ten biggest ransomware incidents. Nine days of opportunity — but only for a defender whose first step was not trivially easy.
Single, double and triple extortion
Modern attackers stack pressure instead of relying on encryption alone. Each layer answers a reason to refuse to pay.
Level | Technique | Why it pressures the victim |
|---|---|---|
Single | Encrypt data and sell the key | Operations stop until systems are restored |
Double | Data exfiltration first, then a threat to publish | Clean backups no longer remove the risk; breach notification and regulatory exposure begin |
Triple | Add DDoS attacks and direct contact with customers or regulators | Reputational damage spreads beyond the victim’s control |
Double extortion is now the default, and it changes what “recovery” means. An organization that restores from backup in six hours has solved availability and nothing else: the stolen data is still gone. At Change Healthcare, after the first group was paid, a second extorted the victim again with the same files.

How the extortion escalated: from encryption to leaks to outside pressure. Own work.
Ransomware-as-a-service
Ransomware-as-a-service (RaaS) is a rental model: one group writes and maintains the malware, another breaks in and runs the attack, and they split the proceeds. It is why attacks that look sophisticated are often run by operators of modest skill.
Role | What they contribute |
|---|---|
Developer | Builds the payload, the leak site and the payment infrastructure; takes a cut |
Access broker | Sells working access to a target, usually a credential or an exploited service |
Affiliate | Buys access, moves laterally, exfiltrates data and launches the payload |

Who does what in the ransomware-as-a-service model. Own work.
CISA’s #StopRansomware advisory on Medusa says the developers recruit access brokers in criminal forums, and those brokers get in through phishing and the exploitation of unpatched software vulnerabilities. The division of labor matters for defenders: the broker’s job is the only part that touches your systems before the damage starts.
Ransomware at Fluid Attacks
We do not respond to ransomware incidents. We work one step earlier, on the software weaknesses that give an access broker something to sell — a narrower claim than most vendors make, and the part we can show data for.
What we assess, and under what rules
We test source code, running applications and infrastructure for organizations that build their own software, under a scope the client defines and with written authorization. Our pentesters work the way an affiliate would — chaining findings, escalating privileges, reaching data — and stop short of impact. The data below comes from our State of Attacks 2026 report, covering 1,119,396 vulnerabilities reported across client systems between January and December 2025.
What it looks like in practice
Each test maps to a link in the chain above.
Test | Link in the ransomware chain it addresses | Evidence produced |
|---|---|---|
Command injection and other flaws that give code execution on a server | Vulnerable line, file and fix | |
Vulnerable third-party components, the Kaseya-style supply chain route | Affected dependency, version and whether the vulnerable function is reachable | |
Exposed services, weak authentication on internet-facing applications | Reproducible request and response | |
Manual penetration testing of chained paths from a low-severity finding to privileged access | Working exploitation path with evidence |
Three entries in our weakness catalog map to how ransomware operators get in and escalate: remote command execution, where a system builds OS commands from externally manipulable input; lack of multi-factor authentication on critical services, exactly what failed at Change Healthcare; and sensitive information in source code, where usernames, passwords and IP addresses sit in a repository waiting to be read.
What our own data says about the gap
Three numbers from State of Attacks 2026 describe the problem and a way out.
70% of the systems we assessed in 2025 had at least one high- or critical-severity vulnerability, up from 53.3% in 2024. Across all severities, 62.2% of reported vulnerabilities were remediated.
The gap opens around automation in the pipeline. Systems running our CI Gate reached a 72.2% remediation rate by year’s end; systems without it reached 58%. Clients who configured their own prioritization policy remediated 64.4% of findings, against 34% on the default.
What automation misses, ours included
Automated scanning alone does not close this. In Fluid Attacks’ benchmark of 36 third-party AppSec tools against a web application seeded with 1,201 vulnerabilities across 105 CWE categories, the best-performing scanner was our own, and it found 22.7% of them — only 8.8% of the application’s total risk exposure in CVSSF. A pentester on the same target reached 89.6% recall at 100% precision.
That limit is worth stating plainly because it cuts against our own product. It is also why 87% of critical-severity vulnerabilities in our 2025 data were identified through manual review, and why we sell the combination rather than the scanner.
A second limit belongs here: application security testing reduces the ways in, but it is not a ransomware control by itself. Offline backups, endpoint detection and a zero trust security model that blocks lateral movement do work no code review replaces.
The practice that decides the rest
Finding a vulnerability and fixing it are different events, and so are fixing it and having fixed it. Our documentation puts it bluntly: a fix can add new vulnerabilities, or it can fail to fix the original one. That is what reattacks are for. Our tools and pentesters retest after each remediation attempt and mark the finding Verified (safe) or Verified (vulnerable), with exploitation evidence attached when it survives. An unverified fix is a hypothesis.
Go deeper: how to prevent ransomware attacks.
How ransomware attacks exploit software vulnerabilities
Vulnerability exploitation is now one of the main ways ransomware operators get in, and the fastest-growing one. Verizon’s 2025 DBIR put it at 20% of breaches as an initial access vector, a 34% rise over the previous edition, with edge devices and VPNs going from 3% to 22% of the targets.
CISA’s advisories name the specific flaws. The #StopRansomware advisory on the Play group documents entry through public-facing applications, specifically known FortiOS vulnerabilities (CVE-2018-13379 and CVE-2020-12812) and Microsoft Exchange ProxyNotShell (CVE-2022-41040 and CVE-2022-41082). None of these were unknown at the time of use.
Two patterns recur in ransomware incidents:
Unpatched internet-facing services: VPN gateways, file transfer software and mail servers are scanned at internet scale within days of a disclosure. WannaCry spread in 2017 through a flaw Microsoft had patched months earlier.
Third-party components: in 2021 the Kaseya ransomware campaign used zero-day flaws in the VSA on-premises software to reach the clients of managed service providers at once. A vulnerable dependency inside your own build is the same problem with your name on it, and software composition analysis is what surfaces it.
The connection to application security is direct. A remote command execution flaw in your own code gives a ransomware affiliate the same foothold as an unpatched appliance, and it will never appear in a vendor patch bulletin because nobody outside your organization knows it exists.
How to prioritize the vulnerabilities ransomware groups exploit
Severity alone is a poor filter for ransomware risk. A critical CVSS score says how bad exploitation would be, not whether anyone is exploiting it or whether your application reaches the vulnerable code.
Our analysis of vulnerability severity makes the case for two more signals. The Exploit Prediction Scoring System produces a probability between 0 and 1, and most vulnerabilities — including many with high CVSS scores — score very low. CISA’s Known Exploited Vulnerabilities catalog answers the blunter question that matters most for ransomware: has this been used against someone already?
Our platform combines these rather than choosing. The Priority value is a percentile that folds CVSS together with CISA KEV listing, EPSS, dependency transitivity, estimated time to fix and reachability; risk exposure is measured in CVSSF, the sum across everything still unremediated.

The platform orders findings by Priority, which folds CVSS together with KEV, EPSS and reachability.
Reachability deserves its own note for third-party code. Our scans verify whether your software actually calls the vulnerable function in a dependency, which separates the alerts that describe your risk from the ones that merely describe the library’s.
This is what turns a list of findings into vulnerability management that reduces ransomware exposure instead of generating tickets.
Famous ransomware attacks
Five incidents explain most of what changed between 1989 and now, and our review of the biggest ransomware incidents documents each in detail.
Attack | Year | Entry point or method | What it demonstrated |
|---|---|---|---|
WannaCry | 2017 | SMB flaw patched months earlier | Patch latency is a ransomware vector |
NotPetya | 2017 | Exploited unpatched devices, spread across networks | Data was irreversibly encrypted; paying recovered nothing |
Colonial Pipeline | 2021 | Exposed VPN account password | ~$4.4 million paid; fuel supply disrupted across the US East Coast |
Kaseya | 2021 | Zero-days in VSA on-premises software | One vendor flaw reached hundreds of downstream clients |
Change Healthcare | 2024 | Citrix portal without MFA | $3.1 billion cost, ~190 million people affected, two extortion attempts |
Norsk Hydro’s 2019 LockerGoga incident belongs beside them for the opposite reason: it said it would not pay any ransom, ran its plants manually and published what happened. It put the cost at $45.6 to $51.3 million.
What to do after a ransomware attack
Isolate first, then identify, then report — and treat the stolen data as a separate incident from the encrypted data. Disconnect affected systems from the network and from shared storage first, including backup drives still mounted.
Identify the ransomware strain from the ransom note and file extensions. It determines whether recovery without payment is possible.
Report it. The FBI asks ransomware victims to report through IC3, and CISA offers technical assistance.
Check No More Ransom before considering payment. The Europol and Dutch police project hosts free decryptors for many ransomware families, and says victims should never pay.
Restore from offline backups, or wipe and rebuild. Verify backups are clean before reconnecting them.
Assume the data is gone. When the data was stolen first, notification obligations do not wait for a decryption key.
Paying is discouraged by the FBI, and the data supports the position: 64% of victims in Verizon’s 2025 dataset did not pay. NotPetya is the clearest argument. It presented a ransom note, but the encryption was irreversible by design and no payment recovered anything.
Conclusions
Ransomware stopped being an encryption problem years ago. It is an access problem with an encryption step at the end, run by specialists who buy their way in through credentials and unpatched software instead of breaking cryptography.
That reframing puts ransomware prevention inside work you already do. Unpatched internet-facing services, missing MFA, vulnerable dependencies and command execution flaws in your own code are all findable and fixable before an affiliate finds them. Organizations that build those checks into their pipeline fix more of them: a 72.2% remediation rate with our CI Gate in the pipeline, against 58% without it.
Frequently asked questions
Is ransomware still just file encryption?
No. Double extortion — stealing the data before encrypting it, then threatening to publish — is now standard, and some groups add DDoS attacks and direct pressure on customers and regulators. Clean backups solve availability and leave the disclosure problem untouched.
How do ransomware groups get initial access?
Mainly through stolen or weak credentials, phishing and the exploitation of unpatched internet-facing software. Verizon’s 2025 DBIR put vulnerability exploitation at 20% of breaches as an entry vector, up 34% year over year, with edge devices and VPNs rising from 3% to 22% of exploited targets.
Does paying the ransom recover the data?
Not reliably. 64% of victim organizations in Verizon’s 2025 dataset did not pay. NotPetya displayed a ransom note while encrypting data irreversibly, so payment recovered nothing. At Change Healthcare a second group extorted the victim again with the same files after the first was paid.
Which vulnerabilities should be patched first to reduce ransomware risk?
Start with anything in CISA’s Known Exploited Vulnerabilities catalog on an internet-facing system, then weight by EPSS probability and by whether your application reaches the vulnerable code. CVSS severity on its own over-selects: most high-scoring vulnerabilities have very low exploitation probability, so ransomware prevention budgets get spent on the wrong findings.
Does Fluid Attacks prevent ransomware?
Not on its own, and we would not claim otherwise. We reduce the ways in, through continuous testing and penetration testing of the weaknesses that give ransomware operators their first foothold. In Fluid Attacks’ benchmark of 36 third-party AppSec tools against an application with 1,201 vulnerabilities, a pentester found 89.6% of them at 100% precision, while the best automated scanner — ours — reached 22.7%. That gap is why we combine AI, scanners and pentesters.
Close the door before the encryption
Our scanners and pentesters find the exploitable weaknesses that give ransomware operators their first foothold, and verify the fixes actually hold. Start free trial · Contact us.
Get started with Fluid Attacks' ASPM solution right now















