SAST and SCA in a bank: the first 30 days
A week-by-week plan: from a list of systems to a rule that keeps critical vulnerabilities out of a release. With the ARDFM No. 53 clauses this plan covers and the figures worth showing on day 30.
Requirements: ARDFM No. 53 review · official text on adilet.zan.kz
Checklist
- There is a list of remote service systems, their repositories, languages and owners
- All repositories from the list are connected to SAST and SCA, and the scanner handles all languages (clauses 148–149)
- The baseline of the first scan is saved
- Secrets found have been replaced, not just removed from the code
- Findings are triaged by priority: exploited and externally reachable first
- Scanning runs on every merge request and release, findings go to the tracker as tasks
- New critical vulnerabilities do not pass into a release (clause 151), release is approved by security (clause 152)
- Check results for every version are stored and available for 3 years (clause 153)
Why, and where to start counting the 30 days
For remote service software (internet bank, mobile app, their server side), SAST and library analysis are mandatory under ARDFM No. 53. The goal of the first month is to launch the process, not to “close all vulnerabilities”: all needed repositories are checked automatically, new critical vulnerabilities do not reach a release, and the results are kept for audit.
clause 148Regardless of the approach to development and (or) modification of remote service software adopted by the bank or organization, security testing is a mandatory stage, during which at least the following activities are carried out: 1) static analysis of source code; 2) analysis of components and third-party libraries.editors' translation
Before the start: a list of systems and owners
One day spent on inventory saves a week of arguments later. Without a list it is unclear what counts as “coverage” and there is no one to hand findings to.
- A list of remote service systems and their repositories: web, mobile apps, backend, integrations.
- Languages and build tools for each repository, to check that the scanner supports the whole stack (clause 149).
- An owner from development and a security lead for each system; contractor development is also on the list (clause 146).
- Where the code is stored and who gives the scanner access: the internal GitLab or Bitbucket inside the bank's perimeter (clause 147).
Days 1–7: connect and scan “as is”
The first scan is a snapshot of the current state, not an exam. We block nothing: the task of the week is to connect all repositories from the list and see the real picture.
- Connect the repositories to SAST and SCA and run the first scan of the main branch.
- Check that all languages were parsed: if the scanner skipped part of the code, it shows as empty modules; better to learn it now than at the audit.
- Save the result as the baseline: new vulnerabilities and progress are counted from it.
- Pull out the secrets in code right away (passwords, tokens, keys): this is the only class of findings fixed not by editing code but by replacing the key.
Days 8–14: triage findings, starting with the main ones
After the first scan there are usually hundreds of findings. Going through them in order makes no sense; first take what is really exploited and reachable from outside.
- SCA: first the libraries with vulnerabilities from the exploited catalog (KEV) and with a critical score in services reachable from the internet.
- SAST: injections, authentication bypass, insecure deserialization in code that handles customer data.
- Mark false positives once, with a reason, so they do not return in later scans.
- Legacy (old non-critical findings) goes into a work plan with an owner and a deadline, not into the daily flow.
Days 15–21: build into development
A check once a quarter does not meet the requirement “before commissioning of each version”. Scanning must run by itself on every change, and the result must reach where the developer works.
- Scanning on every merge request and on the release build, first in observation mode, without blocking.
- Findings go as tasks to the system owner's tracker, not as a PDF report by email.
- Agree on fix deadlines by severity: critical before release, high in the next sprint.
- Assign in each team a person who looks at findings first (a security champion).
Days 22–30: turn on release control
The last step is a rule that keeps new critical vulnerabilities out of a release, and a procedure for approving the release with security. We block only what is new: if you ban a release over all the legacy at once, development stops and the tool gets switched off.
clause 151The bank or organization ensures corrective measures to eliminate the identified vulnerabilities in the manner defined by an internal document approved by the executive body. Critical vulnerabilities are eliminated before commissioning of the remote service software and (or) its new versions.editors' translation
clause 152The bank or organization commissions remote service software and (or) its new versions after approval by the information security unit.editors' translation
clause 153The bank or organization ensures storage of, and prompt access to, all versions of the source code of remote service software and the security testing results that were commissioned during the last 3 (three) years.editors' translation
How to tell the month was not wasted
Five figures worth showing management and the auditor on day 30.
- Coverage: the share of remote service systems whose repositories are scanned automatically; target 100%.
- New critical vulnerabilities in releases after control is turned on; target 0.
- Time to fix critical ones: from discovery to closure, in days.
- The share of false positives among triaged findings; if it is above a third, developers will stop looking.
- Check results for every released version are stored and found in minutes, not in a week.
Common mistakes
Almost all adoption failures are organizational, not technical.
- Blocking releases over all the legacy from day one.
- Scanning only the main branch once a quarter: there is a report for the audit, but the vulnerabilities are already in production.
- No owner for findings: the report goes to security, security forwards it to development, development does not know what to do with it.
- Prioritizing only by CVSS and drowning in “critical” findings nobody exploits.
- Forgetting contractors: their code is also yours under clause 146.
Questions and answers
How long does adopting SAST and SCA really take?
Connecting and the first scan take days. The month goes on triaging the first findings, building into the development process and the release control rule. Fully fixing the accumulated legacy is a separate plan for a quarter or longer.
Do you need to block releases right away?
No. First scanning in observation mode, then blocking only new critical vulnerabilities. The requirement of clause 151, that critical vulnerabilities be eliminated before commissioning, is met this way without stopping development.
What to do with hundreds of findings after the first scan?
Do not go through them in order. First vulnerable libraries from the exploited catalog (KEV) in services reachable from the internet, and dangerous classes in code that handles customer data. Old non-critical items go into a work plan with an owner and a deadline.
What to show the auditor?
The list of systems and coverage, the release control rule, approvals by security, check results for every released version (clause 153) and the history of fixes.
The plan is a guideline for a typical bank with in-house and contracted development. Timing depends on the number of systems and CI/CD maturity.