Code security

Your most dangerous vulnerability sits in your own code. And no one sees it.

CAVRIX continuously scans your repositories for hard-coded credentials, vulnerable dependencies and unsafe code patterns. On every commit, not once a year. If something critical turns up, you get more than an alert: you get the fix already done.

A firewall protects the network, an antivirus tool protects the machine. But the gap that most attacks start from today sits in the software itself: a forgotten API key in the repository, an outdated library with a known flaw, a line of code that passes input on without checking it.

The numbers

Why code has become the number one point of entry

31%

of all attacks now start by exploiting a software vulnerability. For the first time ahead of stolen passwords.

23.8M

new hard-coded credentials were found in public GitHub repositories in 2024 alone, a quarter more than the year before.

45%

of AI-generated code contained a security-relevant vulnerability in controlled tests.

87%

of all codebases examined contain at least one known vulnerability in their open-source dependencies.

Sources: Verizon Data Breach Investigations Report 2026, GitGuardian State of Secrets Sprawl 2025, Veracode GenAI Code Security Report 2025, Black Duck OSSRA 2026. Industry data, no guarantee for individual cases.

The threat landscape

The attacker does not need to break in. They read your code.

Hard-coded secrets01

A password in the code is a public password.

API keys, database passwords and tokens end up in a repository faster than you would think. A single commit to the wrong repository, and the key is out in circulation. Attackers scan public and leaked repositories for them automatically. In 2024, almost 24 million such credentials were found on public GitHub alone. Once a key is published, it stays valid until someone revokes it.

Vulnerable dependencies02

Third-party code you never wrote, with flaws you do not know about.

Modern software is largely made up of open-source packages. 87 percent of all codebases carry at least one known vulnerability in exactly these dependencies. When a flaw in a widely used library becomes public, attackers have the addresses of the vulnerable within hours. Software supply chain is newly among the top three of the OWASP Top 10 in 2025.

Unsafe patterns and AI code03

Code written quickly is rarely secure code.

Input that flows into database queries without being checked. Access controls that get forgotten. Broken Access Control ranks number one in the OWASP Top 10 in 2025. The pressure is rising further from AI assistants: in controlled tests, AI-generated code came with a vulnerability in 45 percent of cases, and the newer models were no safer.

The solution

Scan. Assess. Fix.

A report full of findings that no one works through is not protection. CAVRIX checks at three depths and takes on the fix rather than passing it back to you.

Instant scan on every commit.

  • Hard-coded passwords, keys and tokens
  • Known unsafe functions and patterns
  • Dangerous calls such as unchecked system commands
  • Result in seconds, before the code moves on
  • No waiting for the next review date

Every change reviewed in context.

  • Every change assessed in the context of the codebase
  • Access control and input validation in view
  • Findings rated by real risk, not by volume
  • False positives filtered out before they reach you
  • Clear priority instead of an endless list of hits

Deep review of the entire codebase.

  • All dependencies against known vulnerabilities
  • Repository history searched for old secrets
  • Configuration and unsafe default settings
  • Fix carried out, secret rotated, library updated
  • Finding and action documented with a date
How it works

From connection to ongoing protection. In days, not months.

Day 0

Connection

We connect your repositories and run a first scan of the existing code for secrets, vulnerable dependencies and unsafe patterns.

48 hours

First findings

You get a clear report: which findings are real and critical, which can wait, and what needs to happen first.

Week 1

Cleanup

Critical secrets are rotated, vulnerable libraries updated, dangerous patterns defused. Prioritised by real risk.

Ongoing

Monitoring

From now on, on every commit. We handle new critical findings straight away and report back to you what is already done.

Compliance

What standards and regulators actually require for secure development.

Others claim across the board that a code scanner is mandatory. We show you the passages in their exact wording and say openly where there is no explicit obligation.

NIS2, Art. 21(2)(e)

Security in development and maintenance

The directive explicitly requires security in the acquisition, development and maintenance of systems, including vulnerability handling and disclosure. In Germany this is transposed in Section 30 BSIG. Vulnerabilities in your own code are exactly what this requirement addresses.

ISO 27001:2022, A.8.25 to A.8.28

Secure development lifecycle

The 2022 edition introduces dedicated controls for secure development: A.8.25 secure development lifecycle, A.8.26 application security requirements, A.8.28 secure coding. In addition, A.8.8 requires the technical management of vulnerabilities. All four can be evidenced directly with an ongoing scan.

Cyber Resilience Act, (EU) 2024/2847

Obligations for products with digital elements

Anyone placing software or connected products on the market must actively handle vulnerabilities across the lifecycle. The CRA entered into force on 10 December 2024. Reporting obligations for actively exploited vulnerabilities apply from 11 September 2026, the main obligations from 11 December 2027.

OWASP Top 10:2025

The technical reference framework

Not a law, but the globally recognised standard for the most common risks in applications. Broken Access Control ranks number one, software supply chain is newly among the top three. Our scan is aligned to exactly this catalogue.

None of these standards mandates a specific scanner product. NIS2 and ISO 27001 require secure development and vulnerability management, not a particular tool. The Cyber Resilience Act binds manufacturers of products with digital elements, not everyone who merely uses software internally. We therefore make no claim of a duty to buy, but show which requirements an ongoing scan measurably supports. This overview is a technical assessment, not legal advice.

What you gain
On every commit

Review of new changes, not once a year at the audit.

< 15 min

Response to critical findings, proactively in your channels.

0

Effort for you, we fix the finding and document it.

FAQ

Questions about code security

  • What exactly do you scan?

    Three things. First, hard-coded credentials such as passwords, API keys and tokens, including in the repository history. Second, vulnerable dependencies, meaning open-source packages with known flaws. Third, unsafe code patterns such as unchecked input, dangerous system calls and missing access controls.

  • Does this replace a penetration test?

    No, and we do not claim it does. A scanner finds patterns and known flaws continuously and cheaply. A pentest examines your system as a whole from an attacker's perspective. One is the daily hygiene, the other the periodic deep review. They belong together, neither replaces the other.

  • Does a scanner really find every flaw?

    No. No scanner finds every vulnerability. Business logic flaws, such as a mistaken authorisation logic, have no fixed pattern and escape automatic analysis. And every scanner also reports findings that are not exploitable in the specific case. That is precisely why, with us, a person assesses every finding rather than leaving you alone with a raw list.

  • We have code written by an AI. Is that a problem?

    It is a reason to look more closely. In controlled tests, AI-generated code came with a vulnerability in 45 percent of cases, and the newer models did no better. AI speeds up development but pushes the review to the back. An ongoing scan catches exactly that.

  • Do we have to do this for compliance reasons?

    No law mandates a specific scanner product. NIS2 requires security in development and maintenance including vulnerability management, ISO 27001 calls for a secure development lifecycle. An ongoing scan is one of the few measures that produces dated, verifiable evidence of this. That is how we argue it in the audit as well.

  • Do you get access to our source code?

    For the analysis, yes, with a data processing agreement and clearly regulated access. Processing takes place in the EU. We check the code for vulnerabilities and do not pass it on. On request, the scan runs in your environment, so the code never leaves it.

  • We do not develop in-house at all. Is this relevant for us?

    In part. Anyone who writes no software of their own has no hard-coded secrets. But even purchased and self-operated applications come with open-source dependencies and configurations that age and develop flaws. If you really only use finished software from third parties, our darknet monitoring is the more fitting place to start.

What is in your code that should not be there?

Free code analysis, a clear result, no obligation.