Vulnerability scanning is the automated inspection of a target — a network, a host, a running application, source code, open source dependencies or a cloud configuration — to find weaknesses that match known patterns. The scanner is the tool. The scanning is the process that decides what to point it at, when to run it and what to do with the output.
In the systems Fluid Attacks assessed in 2025, automated tools identified 88.4% of all vulnerabilities but only 55.8% of total risk exposure, and 87% of critical-severity findings came from manual review. Vulnerability scanning covers ground. It does not cover the ground that hurts most.
This page covers what a vulnerability scanner actually looks for, how vulnerability scanning is classified, how often to run it, where the method stops, and what has to happen after the report lands.
What is vulnerability scanning?
Vulnerability scanning is a detection method: software compares the components, versions, configurations and code of a defined target against rules and vulnerability databases, then reports every match. The term covers two different things, and the distinction matters when you write a policy around it.
Vulnerability scanner: the tool. A program with a rule set, a database and an engine that probes or parses the target.
Vulnerability scanning: the process. Defining scope, picking credentials, running the scanner, reading the output and deciding what happens next.
A scanner does not understand your application. It recognizes shapes it has been taught: a library version with a published CVE, a header that is missing, a query built by string concatenation, a storage bucket that answers to anyone. Fluid Attacks' analysis of web application scanning describes the mechanic plainly — a scan reviews the components and configurations of the predefined targets of evaluation and compares them with characteristics of vulnerable code.
That one sentence holds both the value and the ceiling of vulnerability scanning. Pattern matching scales to thousands of assets. It does not reason.
Scanning, assessment and penetration testing are not the same thing
Vulnerability scanning is a method. A vulnerability assessment is the process that may contain it, and penetration testing answers a different question altogether. Teams that treat the three as synonyms end up buying vulnerability scanning and reporting it as something else.
Practice | Question it answers | Who or what runs it | Output |
Vulnerability scanning | What known weaknesses match this target? | Automated tool | A list of findings, with severity and location |
Vulnerability assessment | How exposed is this system, and what matters first? | Tools plus analysts | Classified, prioritized and reported weaknesses |
Penetration testing | Can an attacker actually exploit this, and how far do they get? | Pentesters, assisted by tools | Proof of exploitation, chained attack paths and impact |
Are we closing faster than we are opening? | The organization | Remediation, verification and metrics over time |
A vulnerability scanning report is evidence for an assessment. It is not the assessment, and it is never the pentest — nor is it red teaming or a breach and attack simulation, both of which test detection and response rather than enumerate weaknesses.
Vulnerability scanning at Fluid Attacks
Fluid Attacks runs vulnerability scanning as one layer of continuous security testing, alongside AI and pentesters, and reports every finding with the technique that produced it. That attribution is the part most scan reports leave out, and the part that lets you judge your own coverage.
What we scan, and with which technique
Every vulnerability in the platform carries the technique that detected it, so a team can tell at a glance whether a finding came from a rule or from a person.
Technique | What it inspects |
SAST and AI SAST | Source code |
SCA | Open source dependencies and their known vulnerabilities |
Web DAST and API DAST | Running applications and APIs |
MAST | Mobile applications |
CSPM | Cloud configurations |
SS | Secrets committed to repositories |
SCR, PTaaS and RE | Manual review, pentesting and reverse engineering |

Every finding shows the technique that detected it: SAST, DAST, SCA, PTaaS and the rest.
The first six techniques are vulnerability scanning. The last three are not, and the platform never blurs the line.
How a scan runs in a real pipeline
Static Application Security Testing (SAST), Software Composition Analysis (SCA), Dynamic Application Security Testing (DAST) and an APK scanner ship as standalone CLIs and Docker containers. They report results in the CLI or export them as CSV or SARIF, and the --strict flag makes the execution fail with exit code 1 when it finds anything. That flag is what turns vulnerability scanning into a build gate instead of a report nobody opens. Fluid Attacks' SAST scanner scored 100 on the OWASP Benchmark version 1.2 in 2021, and in 2022 the App Defense Alliance added its static scanning CLI as an approved tool for validating CASA tier 2 requirements.
What tools found and what people found in 2025
The numbers below come from the systems Fluid Attacks assessed during 2025. Pentester figures come exclusively from the Advanced plan; tool figures include both Advanced and Essential plans.
Measure | Automated tools | Manual techniques |
Share of total risk exposure identified | 55.8% | 44.2% |
Share of the total number of vulnerabilities identified | 88.4% | 11.6% |
Share of critical-severity findings identified | 13% | 87% |
Read the first two rows together. Vulnerability scanning found almost nine out of ten findings and barely more than half of the risk, because a scanner is good at volume and weak at the severe, context-dependent flaws that carry most of the exposure. The report's own conclusion is that human expertise remains essential for thorough security assessments, even after implementing multiple tools and investing in AI.
A limit worth stating: our own scanner found 22.7%
In Fluid Attacks' benchmark of 36 third-party AppSec tools against a purpose-built web application containing 1,201 vulnerabilities across 105 CWE categories, the automated tool with the highest recall was Fluid Attacks' own scanner, at 22.7% of the vulnerabilities and 8.8% of total risk exposure measured in CVSSF. A pentester working the same application reached 89.6% recall at 100% precision.
That is a limit of vulnerability scanning as a method, not of a product, and it is the number to keep in mind every time a scan comes back clean. Fluid Attacks' guide to benchmarking AppSec tools puts the same finding in plainer terms: the automated scanner with the highest true positive count still missed almost 80% of the vulnerability universe.
A scan is not finished until the fix is verified
A patch can fail, miss the real cause or introduce something new, so Fluid Attacks closes the loop with reattacks: retests run by its tools and pentesters after a fix attempt. Each one ends in Verified (vulnerable), Verified (safe) or On hold. Until a finding reaches Verified (safe), the vulnerability scanning cycle around it is still open, whatever the ticket says.
What do vulnerability scanners look for?
Vulnerability scanning looks for more than CVE identifiers. Known vulnerable dependencies are the easiest category to match, but modern rule sets also cover insecure code patterns, weak configurations and exposed secrets.
The rules behind vulnerability scanning are worth inspecting before you trust a report. Fluid Attacks publishes its detection logic openly: the Security Rules catalog holds more than 1,900 automated rules across 43 targets, described as automated logic and checks across code, cloud, applications and binaries, each mapped to the standards it supports. Vulnerability scanning that only enumerates CVEs in dependencies is doing a fraction of what the method allows.

The automated rules our scanners run, with their targets and the standards they map to.
Types of vulnerability scans
Vulnerability scanning gets classified two ways: by what it points at, and by how it is run. Both classifications change what the results mean.
By target
The target decides which weaknesses vulnerability scanning can even see.
Scan type | Target | Typical findings |
Network vulnerability scanning | IP ranges, open ports, services, operating systems | Unpatched services, exposed ports, weak protocol versions |
Web application scanning | Running web applications and APIs | Injection, broken access control, misconfigured headers |
Code scanning | Source code in the repository | Insecure patterns, hardcoded secrets, unsafe data handling |
Dependency scanning | Third-party and open source components | Components with published vulnerabilities |
Cloud and container scanning | Cloud configurations and container images | Permissive policies, public storage, vulnerable base images |
By mode
The same target produces different results depending on where vulnerability scanning runs from and whether it holds credentials.
Mode | What it sees | When to use it |
External scan | The attack surface reachable from the internet | Perimeter checks and compliance evidence |
Internal scan | Assets reachable from inside the network | Lateral movement risk and segmentation gaps |
Unauthenticated scan | Only what an anonymous visitor sees: banners, responses, exposed endpoints | A first view of the attacker's perspective |
Authenticated scan | Installed package versions, local configuration, user privileges, settings that only appear after login | Accurate inventory and far fewer version-guessing errors |
Comprehensive scan | Every asset in scope, every check | Periodic baselines |
Targeted scan | One application, one branch, one asset group | Change-driven checks in a pipeline |
The authenticated and unauthenticated distinction quietly decides the false negative rate of your vulnerability scanning. An unauthenticated scanner infers the version of a service from what the service advertises; an authenticated scan reads it. OWASP warns that this inference leads to both false positives and false negatives, because vendors backport security patches without moving the version number.
What are the limitations of vulnerability scanners?
Vulnerability scanners miss what they have no rule for, and the gap is structural rather than a matter of buying a better product. That includes business logic flaws, authorization errors that depend on who the user is, vulnerabilities that only matter when chained with another, and anything nobody has published a pattern for yet.
OWASP's Web Security Testing Guide makes the same point about the automated tools that perform vulnerability scanning: they can find many common problems, but they often lack the application-specific context needed to detect deeper flaws, and the most serious issues are frequently intertwined with business logic and custom design. Fluid Attacks reaches the same conclusion from its own data — the relationship between vulnerability scanning and pentesting is one of complementation, not substitution.
Three failure modes of vulnerability scanning show up in practice:
False positives: findings that do not hold up under review. They cost analyst time and, worse, teach developers to ignore the queue.
False negatives: real vulnerabilities the scanner never reports. Nobody sees them, which is why recall deserves at least as much scrutiny as precision when you evaluate a tool.
Context blindness: vulnerability scanning returns a finding that is technically correct and practically irrelevant, or the reverse — a low-severity issue that becomes critical because of where it sits.
A clean scan is not an all-clear: Equifax
In 2017, attackers reached Equifax's online dispute portal through a critical Apache Struts vulnerability, CVE-2017-5638, which US-CERT had alerted about publicly on March 8 of that year. The U.S. Senate Permanent Subcommittee on Investigations' staff report found that none of Equifax's subsequent scans identified the vulnerable version of Apache Struts running on its network, and traced the failure to scope rather than to the tool: the company "had no inventory to determine where or if it used Apache Struts on its network", and the lack of an IT asset inventory limited the effectiveness of its scanning tools.
The report's own formulation is the lesson: an organization can only defend the assets it has identified. Vulnerability scanning inherits the quality of the asset inventory it runs against, and a negative result only covers what was in scope.
How often should you run vulnerability scans?
Run vulnerability scanning on a fixed cadence and on every significant change, because the two catch different things. The cadence catches newly published vulnerabilities in code you have not touched; the change trigger catches what you just introduced.
Trigger | What to scan | Reference |
Every three months, at minimum | Internal and external attack surface | PCI DSS requires internal and external scans at least once every three months for entities handling cardholder data |
Every three months, by an Approved Scanning Vendor | External attack surface | |
On every commit, pull request or release | Code, dependencies and the running application | Fluid Attacks' own practice; CIS Control 7 asks for ongoing identification and tracking without fixing an interval |
After each fix attempt | The specific finding | Reattack, to confirm the fix held |
Quarterly is a compliance floor, not a security target. For software that changes weekly, vulnerability scanning belongs in the pipeline, where a failing scan blocks the merge and the window between introducing a flaw and finding it shrinks from months to minutes. Running vulnerability scanning continuously during development is what keeps the inventory of findings current enough to act on.
What to do after a vulnerability scan
A vulnerability scanning report is an input, not an outcome. Four steps turn one into the other.
Validate. Confirm the finding is real and reachable in your context. This is where false positives get removed and, more importantly, where a finding gets its actual severity.
Prioritize by risk, not by count. Fluid Attacks measures risk exposure as the sum of the CVSSF scores of all open vulnerabilities, so that closing one critical finding moves the number more than closing twenty trivial ones.
Assign and remediate. Every finding needs an owner and a deadline, or vulnerability scanning just fills a dashboard nobody acts on.
Retest. Verify the fix removed the vulnerability instead of moving it.
Skip step four and your vulnerability scanning program measures activity instead of results.
Conclusions
Vulnerability scanning is the fastest way to find known weaknesses at scale, and it is the wrong tool for the flaws that carry the most risk. Fluid Attacks' own 2025 data says both things at once: vulnerability scanning found 88.4% of the vulnerabilities and 55.8% of the risk exposure, while 87% of critical findings needed a person.
The practical conclusion is not to run less vulnerability scanning. It is to run it continuously, with credentials, against an asset inventory you trust, to treat a clean report as the beginning of the question rather than the answer, and to put human testing where the scanner's rules run out.
Frequently asked questions
What is the difference between vulnerability scanning and vulnerability assessment?
Vulnerability scanning is an automated detection method; a vulnerability assessment is the broader process that may contain it. The assessment adds scoping, validation, classification, prioritization and reporting, and it can combine automated tools with manual techniques. A scan produces a list. An assessment produces a judgment about exposure.
What is the difference between an authenticated and an unauthenticated vulnerability scan?
An unauthenticated scan sees only what an anonymous visitor sees: exposed ports, service banners, public responses. An authenticated scan logs in and reads installed package versions, local configurations and privileges directly. Authenticated vulnerability scanning produces fewer false negatives, because it stops inferring versions from what a service advertises.
How often should you run vulnerability scans?
At minimum every three months, which is what the PCI Security Standards Council states for internal and external scans at entities handling cardholder data, plus a scan after every significant change. For applications under active development, that floor is far too slow: run vulnerability scanning on every commit or pull request, and rescan after each fix attempt to confirm it held.
Does passing a vulnerability scan mean my organization is PCI DSS compliant?
No. The PCI Security Standards Council states that an external scan completed by an Approved Scanning Vendor "is not an indication that any other PCI DSS requirements have been reviewed or are in place". A passing scan is evidence for one requirement, not a compliance verdict.
Does Fluid Attacks use vulnerability scanners?
Yes, and it publishes what they miss. In Fluid Attacks' benchmark against an application with 1,201 vulnerabilities, its own scanner led all automated tools with 22.7% recall, while a pentester reached 89.6% at 100% precision. That gap is why its Essential plan runs on automated tools and AI, while the Advanced plan adds manual penetration testing, secure code review and reverse engineering by its experts.
Scan continuously, then look where the rules run out
Our vulnerability scanning runs on every change and our pentesters go after what no rule can match, so you find out what a clean report was hiding before an attacker does. Start free trial · Contact us.
Get started with Fluid Attacks' RBVM solution right now















