Nexus Void Research
AppSec Backlog, Security Debt, Vulnerability Remediation, Risk-Based Vulnerability Management, Vulnerability Prioritization, EPSS, CISA KEV, CVSS, VEX, Original Research
AppSec Backlog 2026: Only 4.4% of CVEs Need Urgent Action
An AppSec backlog is every unfixed security finding you know about. Our 2026 data: 4.4% of 408,395 CVEs are in KEV or EPSS 10%+, vs 55% High/Critical by CVSS.
An AppSec backlog is the pile of known, unfixed security findings in your code and dependencies, and in 2026 most of it does not need urgent work. Of 408,395 CVEs with a FIRST EPSS score on November 18, 2026, only 4.4% are either in the CISA Known Exploited Vulnerabilities catalog or have at least a 10% predicted chance of exploitation in the next 30 days. By CVSS alone, 55% look urgent.
That is our analysis of the full FIRST EPSS score file, the CISA Known Exploited Vulnerabilities (KEV) catalog (version 2026.11.17, 1,782 CVEs) and CVSS v3 severity counts from the National Vulnerability Database (NVD). The gap between those two numbers is why backlogs keep growing: teams are told to fix far more than actually matters, so they fall behind on the part that does. Below we show how to sort the backlog, how to shrink it, and how ARGUS ranks it by exploitability and opens the fix pull requests for you to review and merge.
What is an AppSec backlog?
An AppSec backlog is every open security finding an engineering team knows about but has not fixed: SAST findings in first-party code, vulnerable open-source dependencies found by software composition analysis (SCA), leaked secrets, misconfigurations and pentest findings. It grows when findings arrive faster than fixes ship. Unlike a feature backlog, it carries live risk, because attackers can use what sits in it.
The backlog is not one list. It is a flow with four parts:
Part of the flow | What it means | Typical cause of growth |
|---|---|---|
Inflow | New findings from every scan and every pull request | More code, more dependencies, AI-generated code |
Stock | Findings open today | Old findings never triaged or closed |
Noise | Findings that do not apply to you | Unreachable code paths, false positives, duplicate alerts |
Outflow | Findings fixed, or closed with evidence | Limited developer time, unclear ownership |
Most backlog advice only talks about outflow ("fix faster"). In our experience the cheaper wins are cutting inflow and removing noise, because they shrink the work instead of speeding it up.
How big is the AppSec backlog problem in 2026?
It is now the normal state of software. Veracode's 2026 State of Software Security report, which analysed 1.6 million applications, found that 82% of organizations carry security debt (flaws left open for more than a year), up from 74% in 2025. Critical security debt, meaning long-lived flaws with both high severity and high exploitability, reached 60% of organizations, up from 50%. The median half-life of a flaw across all scan types was 243 days, as reported by Help Net Security.
Supply volume keeps climbing too. The NVD held 427,777 CVE records when we queried it on November 18, 2026, and EPSS scored 408,395 of them. Every one of those can show up in a dependency scan.
How much of an AppSec backlog actually matters? Our data
We sorted every EPSS-scored CVE by the two signals that best predict real-world attacks: confirmed exploitation (CISA KEV) and predicted exploitation (EPSS, the probability of exploitation activity in the next 30 days).
Slice of all EPSS-scored CVEs (n = 408,395) | CVEs | Share |
|---|---|---|
EPSS 1% or higher | 149,213 | 36.5% |
EPSS 5% or higher | 31,055 | 7.6% |
EPSS 10% or higher | 17,558 | 4.3% |
EPSS 50% or higher | 4,388 | 1.1% |
In CISA KEV (confirmed exploited) | 1,782 | 0.44% |
In KEV or EPSS 10% or higher | 17,995 | 4.4% |
Now compare that with severity. Of the 196,640 NVD CVEs with a CVSS v3 severity, 108,555 (55.2%) are rated High or Critical: 30,882 Critical and 77,673 High. A "fix every High and Critical" policy therefore puts about six times more CVEs in the urgent queue than a "KEV or EPSS 10%+" policy does.
Two more findings change how you should use these signals:
EPSS alone misses confirmed exploitation. 24.5% of KEV CVEs (437 of 1,782) have an EPSS score below 10% today, and 3.3% are below 1%. EPSS predicts; KEV confirms. Use both.
Ransomware is concentrated in a small set. 361 KEV entries (20%) are flagged by CISA as known to be used in ransomware campaigns. If you only have time for one list, start there.
The takeaway is not "ignore everything else." It is that a backlog sorted by exploitation evidence has a short, finite top, and a backlog sorted by severity does not.
How do you prioritize a vulnerability backlog?
Prioritize by exploitation evidence first, exposure second and severity third. Put anything in CISA KEV at the top, then anything with a high EPSS score, then raise or lower each item by whether the vulnerable code is reachable and whether the service is internet-facing. CVSS still matters, but as a tiebreaker, not the sort key.
A practical order that fits the data above:
KEV and reachable, on an internet-facing service. Fix now. Our time-to-exploit study found a median of 5 days from disclosure to confirmed exploitation for 2026 CVEs.
KEV or EPSS 10%+, reachable. Fix this sprint.
Reachable, high CVSS, low EPSS. Schedule it.
Unreachable or not affected. Close it with evidence (see VEX below) instead of carrying it.
We cover the scoring systems in depth in how to prioritize vulnerabilities with CVSS, EPSS and KEV and what CVSS misses.
What is risk-based vulnerability management?
Risk-based vulnerability management (RBVM) ranks vulnerabilities by the risk they pose to your specific environment rather than by severity alone. It combines exploit likelihood (EPSS, KEV, public exploit code), asset exposure (internet-facing or internal), reachability (whether your code actually calls the vulnerable function or package) and business impact (what data or service the asset touches).
The NIST Secure Software Development Framework (SP 800-218) asks teams to assess, prioritize and remediate vulnerabilities based on risk, and our numbers show why. When 55% of CVEs are High or Critical, severity cannot tell you where to start. Exploitation evidence cuts the urgent set to roughly 4%, and reachability cuts it further for the code you actually ship.
Reachability deserves its own note. Many dependency CVEs sit in functions or packages your application never loads. We explain how this works, and where it falls short, in reachability analysis for vulnerability prioritization.
What is vulnerability remediation, and how do you speed it up?
Vulnerability remediation is the work of removing a vulnerability's risk: upgrading the dependency, changing the code, rotating the secret or applying a compensating control. Closing a finding as "not affected" with documented evidence also counts, because it removes the item from the backlog honestly.
Four moves shorten remediation without adding headcount:
Fix at the pull request. A finding caught before merge costs one developer a few minutes in code they already have open. The same finding a year later needs an owner, context and a release.
Batch dependency upgrades. One upgrade often clears several CVEs. Group findings by package and version, not by CVE.
Generate the smallest fix, and keep a human on merge. Automated fix pull requests work best when they are small, re-scanned and approved by a person. See how to automatically fix vulnerable dependencies.
Close with VEX instead of ignoring. A Vulnerability Exploitability eXchange (VEX) statement records that a CVE does not affect your product and why. CISA's VEX guidance describes the minimum fields. It turns "we decided not to fix this" into an auditable record that customers and auditors can read.
What is security debt, and how is it different from an AppSec backlog?
Security debt is the part of the backlog that has aged. Veracode defines it as flaws left unfixed for more than a year; critical security debt is the subset that is both severe and likely to be exploited. Every item of security debt is in your backlog, but not every backlog item is debt yet.
The distinction matters for planning. Backlog is an operations problem: inflow versus throughput this month. Debt is a governance problem: risk you have been carrying, often without anyone formally accepting it. A useful governance rule is that nothing in KEV may become debt, and anything older than a set age needs either a fix date or a signed risk acceptance.
How do you stop an AppSec backlog from growing?
Treat it like any queue: inflow must stay below outflow. A worked example (illustrative numbers): if a team adds 40 findings a month and fixes 25, the backlog grows by 15 a month no matter how hard they work. Cutting inflow by half with a merge gate, and closing 10 a month as not affected with evidence, turns the same team's backlog from growing to shrinking.
Lever | What it changes | Example effect on a 40-in, 25-out month |
|---|---|---|
Merge gate on new high-risk findings | Inflow | 40 in becomes 20 in |
Reachability plus VEX | Noise | 10 closed as not affected |
Exploitation-first ranking | Which 25 get fixed | The 25 fixed are the ones attackers use |
Small, verified fix pull requests | Outflow | More fixes per developer hour |
How does ARGUS clear an AppSec backlog?
ARGUS turns a backlog into a short review queue. It ranks every finding by what attackers can actually use, puts the exploitable and reachable ones on top, then writes the fix and opens a pull request. Your job shrinks to reviewing the diff and merging it.
Here is what happens to a backlog, step by step:
Scan everything in one pass. Connect a GitHub repository and ARGUS runs static analysis (SAST) on your code, software composition analysis (SCA) on your dependencies and secret scanning, so the whole backlog lands in one list instead of three tools.
Rank by exploitability, not severity. Each finding gets a 0 to 100 priority score that blends CVSS, EPSS, CISA KEV and reachability. A reachable CVE that attackers are already exploiting rises to the top. An unreachable one sinks to the bottom without disappearing.
Show the evidence. Every dependency finding is marked reachable, unreachable or unknown, with an evidence string that says why. ARGUS never silently deletes a finding, so you can always check its reasoning.
Close what does not affect you. Unreachable findings become VEX "not affected" statements you can hand to a customer or auditor, which removes them from the backlog with a record instead of a shrug.
Write the fix and open the pull request. Click Assign to Agent on a finding and ARGUS clones the repository, generates the smallest fix, verifies it with a build and a re-scan, and opens a pull request. If the re-scan finds a new problem, the fix is flagged for your review instead of passing quietly.
You review and merge. ARGUS never merges on its own. A developer reads the diff, and teams that want an extra check can require an admin to approve each fix request first.
Stop new debt at the door. The ARGUS pull request reviewer checks every new pull request and can post a required status check, so new high-risk findings get blocked before they merge and the backlog stops refilling.
Backlog problem | What ARGUS does | What is left for you |
|---|---|---|
Too many findings to triage | Ranks by CVSS, EPSS, KEV and reachability | Start at the top of the list |
Noise from unreachable CVEs | Marks reachability with evidence, exports VEX | Spot-check the evidence |
No time to write fixes | Opens a verified fix pull request | Review the diff and merge |
New findings keep arriving | Blocks risky pull requests with a required check | Fix it in the pull request you already have open |
Limits, stated plainly: reachability is import-level and covers JS/TS and Python only today, so Go, Maven, Cargo and RubyGems findings are treated as reachable. Fix pull requests take several minutes per finding. Scanning and prioritization start on the free plan; agentic fix pull requests are part of the paid plans.
Scan a repository free and see your own backlog sorted by exploitability, or read how SCA with reachability and AI code review at the pull request work.
Our read
The AppSec backlog is not a sign that teams are lazy. It is a sign that most tools still hand developers a severity-sorted list that is about six times longer than the exploitation-sorted one, and then measure them on how much of it they clear. Fix the sort order, remove the noise and hand developers a ready-made fix instead of a ticket, and the backlog becomes a short queue of pull requests waiting for review. That is the job we built ARGUS to do.
Methodology
On November 18, 2026 we downloaded the current FIRST EPSS score file (model v2026.06.15, 408,395 CVEs) and the CISA KEV catalog (version 2026.11.17, 1,782 CVEs). Every KEV CVE had an EPSS score. We counted CVEs at each EPSS threshold, the overlap with KEV, and KEV entries flagged for known ransomware use. CVSS v3 severity counts come from the NVD CVE API 2.0 total-results counts for each cvssV3Severity value on the same day; CVEs without a v3 score are excluded from the 55.2% figure. EPSS scores change daily, so the exact counts will drift. Security debt figures are from Veracode's 2026 State of Software Security report as reported by Help Net Security.
Frequently asked questions
What is an AppSec backlog?
An AppSec backlog is the set of known security findings in an organization's applications that have not been fixed or closed: code flaws, vulnerable dependencies, leaked secrets and misconfigurations. It grows whenever new findings arrive faster than teams fix or close them, and it carries live risk because attackers can exploit what sits in it.
How much of a vulnerability backlog needs urgent action?
Far less than severity suggests. In our November 2026 analysis, 4.4% of 408,395 EPSS-scored CVEs were in CISA KEV or had an EPSS score of 10% or higher, while 55.2% of CVSS v3-scored CVEs were rated High or Critical. Reachability and exposure narrow the urgent set further for your own code.
How do you prioritize a vulnerability backlog?
Rank by exploitation evidence first: CISA KEV entries, then high EPSS scores. Then adjust by reachability and whether the asset is internet-facing, and use CVSS as a tiebreaker. Close unreachable or not-affected findings with documented evidence, such as a VEX statement, instead of leaving them open.
What is the difference between security debt and an AppSec backlog?
The backlog is every open finding. Security debt is the aged part of it: Veracode defines it as flaws left unfixed for more than a year. In 2026, Veracode found 82% of organizations carry security debt and 60% carry critical security debt, meaning old flaws that are both severe and likely to be exploited.
What is risk-based vulnerability management?
Risk-based vulnerability management ranks vulnerabilities by actual risk to your environment, combining exploit likelihood (EPSS, CISA KEV), exposure, reachability and business impact, instead of fixing by CVSS severity alone. It exists because severity marks over half of all CVEs as High or Critical, which is too many to act on first.
Should I only fix vulnerabilities in CISA KEV?
No. KEV is the right place to start, but it lists only confirmed exploitation, so it lags real attacks. Pair it with EPSS for predicted exploitation and keep fixing reachable high-severity issues on a schedule. In our data, 24.5% of KEV CVEs had an EPSS score under 10%, so neither signal is enough alone.
Is EPSS enough to prioritize vulnerabilities?
Not on its own. EPSS predicts the probability of exploitation in the next 30 days, which makes it useful for ranking, but 437 of the 1,782 CVEs in CISA KEV scored below 10% in our November 2026 snapshot. Use EPSS together with KEV, reachability and exposure.
How do you reduce an AppSec backlog without more engineers?
Cut inflow and noise before chasing outflow. Gate new high-risk findings at the pull request, close unreachable or not-affected CVEs with VEX evidence, batch dependency upgrades by package, and use small, re-scanned fix pull requests that a developer approves. Each removes work instead of speeding up the same work.
Can a tool fix my AppSec backlog automatically?
It can do most of the work, but a person should still merge. ARGUS ranks findings by exploitability and reachability, then its Assign to Agent feature generates the smallest fix, verifies it with a build and re-scan, and opens a pull request. It never merges, so a developer reviews the diff and merges it.
What is VEX in vulnerability management?
VEX, short for Vulnerability Exploitability eXchange, is a machine-readable statement that says whether a product is affected by a specific CVE and why. It lets teams close not-affected findings with evidence and share that decision with customers and auditors. CISA publishes minimum requirements for VEX documents.
DATA SOURCES
FIRST EPSS scores (model v2026.06.15, 2026-11-18): https://www.first.org/epss/ ; CISA KEV catalog v2026.11.17: https://www.cisa.gov/known-exploited-vulnerabilities-catalog ; NVD API 2.0 severity counts: https://nvd.nist.gov/developers/vulnerabilities ; Veracode 2026 State of Software Security via Help Net Security