Skip to main content Scroll Top

Secure Software Development: The Practices, Standards and Evidence Behind Safer Code

Updated November 2026 · Written and maintained by the Progression Agency strategy team

Secure software development is the set of habits that keeps vulnerabilities out of released code and gets the ones that slip through fixed quickly: written security requirements, threat modeling, reviewed and tested code, protected builds and a plan for reports after launch. NIST’s framework for this work, SP 800-218, names 19 such practices, and OWASP adds a maturity model, a requirement standard and a list of the risks that matter most. This page takes each one in turn and shows the record a team should be able to hand a customer who asks. Progression’s developers are based in New York and build web, mobile and SaaS software for companies across the US.

On this page · 17 sections
  1. A security questionnaire the team could not answer
  2. What secure software development covers
  3. SP 800-218 practice by practice, with the record each leaves
  4. Threat modeling before the first line of code
  5. OWASP SAMM: measuring how mature the program is
  6. OWASP ASVS: the requirement list for one application
  7. The OWASP Top 10:2025, read as design decisions
  8. Software bills of materials and the supply chain
  9. Release integrity: hashes, signing and archived builds
  10. Safe defaults and protected build environments
  11. Code review, testing and the tools behind them
  12. After release: disclosure, triage and root causes
  13. Securing SaaS products: tenants, sign in and data
  14. Software development security best practices, as one checklist
  15. What secure development work costs
  16. How Progression’s team builds security into client software
  17. Related services

Looking for the life cycle?This page is about the practices and the proof. For where security work sits inside planning, design, build, test and release, see our software development life cycle guide.

Examples marked “Illustrative example.” are hypothetical, and no real company, product or customer is shown.

A security questionnaire the team could not answer

Illustrative example. A hypothetical SaaS company is close to its largest contract when the buyer sends a security questionnaire. It asks for a software bill of materials, the vulnerability disclosure policy, a way to check that a release has not been altered and the record of the last design review.

Two boxed panels. The first lists four items a hypothetical enterprise buyer asked a SaaS company for: an SBOM, a disclosure policy, release integrity checks and a design review record. The second shows the team had scanners running but none of the four records.Two boxed panels. The first lists four items a hypothetical enterprise buyer asked a SaaS company for: an SBOM, a disclosure policy, release integrity checks and a design review record. The second shows the team had scanners running but none of the four records.
Illustrative example. The work was partly done; the records were not.

The team had scanners running and reviewed every pull request, but none of the four items existed on paper. Two weeks went to writing down work that was partly done. That is the gap this page is about: security practices that leave a record a customer can read. No real company, buyer or contract is shown here.

What secure software development covers

Secure software development means building security into how software is specified, designed, written, tested, released and maintained, rather than auditing it once at the end. NIST’s SSDF project page describes the framework’s practices as outcome based: they say what a team should achieve and leave the tools and methods to the team.

Security in software development rests on three kinds of reference. A framework says what to do across the whole operation (NIST SP 800-218). A maturity model measures how well a team does it (OWASP SAMM). A requirement standard says what one application must do (OWASP ASVS). The OWASP Top 10 sits beside all three as the shortlist of risks worth designing out.

19: SSDF practices. Spread over four categories in NIST SP 800-218, version 1.1.
15: SAMM practices. Three in each of OWASP SAMM's five business functions.
3: ASVS levels. From a first layer of defense to high assurance.
~350: ASVS requirements. Across 17 chapters in version 5.0.0.
7: SBOM data fields. The minimum NTIA set out for each component.
4: Threat model questions. From what we are working on to whether we did enough.

Is software development and security one job or two?

One. Software development and security share the same backlog, the same pull requests and the same release, so the developers who write a feature also write its security requirements and tests, with specialists reviewing the riskiest parts.

Who asks to see this work?

Enterprise buyers, insurers, auditors and regulators in fields such as health and finance. SP 800-218 itself notes that purchasers can use its common vocabulary when talking to suppliers.

SP 800-218 practice by practice, with the record each leaves

Version 1.1 of the framework, published by NIST in February 2022 as SP 800-218, sorts 19 practices into four categories. Each practice has tasks and notional examples; NIST says no example is required and the listed ones are not the only options. The table gives each practice its NIST name, a plain reading and the record that proves it happened.

Every SSDF practice and the record a team keeps
IDThe practice, in NIST’s wordsWhat it asks for in plain termsThe record a team keeps
PO.1Define Security Requirements for Software DevelopmentKnow the security rules that apply, including what third party component suppliers must meetA requirements list, and the clauses sent to suppliers
PO.2Implement Roles and ResponsibilitiesEveryone involved, inside or outside the company, knows their security dutiesNamed roles and training records
PO.3Implement Supporting ToolchainsAutomate security checks so they are repeatable and leave evidencePipeline configuration and tool logs
PO.4Define and Use Criteria for Software Security ChecksDecide what a pass looks like before checkingWritten release criteria
PO.5Implement and Maintain Secure Environments for Software DevelopmentSeparate and protect development, build, test and distribution systemsAccess rules, sign in policy, network separation
PS.1Protect All Forms of Code from Unauthorized Access and TamperingLock down who can change codeRepository permissions and branch rules
PS.2Provide a Mechanism for Verifying Software Release IntegrityLet customers confirm a release is genuinePublished hashes or code signatures
PS.3Archive and Protect Each Software ReleaseKeep every release and its provenance dataArchived builds and an SBOM per release
PW.1Design Software to Meet Security Requirements and Mitigate Security RisksDesign with the risks in viewThreat model and design notes
PW.2Review the Software Design to Verify Compliance with Security Requirements and Risk InformationHave someone outside the design, or automation, check itDesign review record
PW.4Reuse Existing, Well-Secured Software When Feasible Instead of Duplicating FunctionalityPrefer proven, maintained components over new codeApproved component list
PW.5Create Source Code by Adhering to Secure Coding PracticesWrite code to a known secure standardCoding standard and review notes
PW.6Configure the Compilation, Interpreter, and Build Processes to Improve Executable SecurityTurn on the build and compiler protectionsBuild settings under version control
PW.7Review and/or Analyze Human-Readable Code to Identify Vulnerabilities and Verify Compliance with Security RequirementsCode review, code analysis or bothReview approvals and analysis results
PW.8Test Executable Code to Identify Vulnerabilities and Verify Compliance with Security RequirementsTest the running software, record and triage findingsTest results in the issue tracker
PW.9Configure Software to Have Secure Settings by DefaultShip with safe settings switched onA documented secure baseline
RV.1Identify and Confirm Vulnerabilities on an Ongoing BasisWatch for new issues and have a disclosure policyDisclosure policy and intake log
RV.2Assess, Prioritize, and Remediate VulnerabilitiesFix by risk, not by arrival orderRisk notes and fix plan per issue
RV.3Analyze Vulnerabilities to Identify Their Root CausesFix the pattern, not only the bugRoot cause and lessons learned notes

Before any code: people, tools and environments

The PO practices are about the company rather than one product. PO.1.3 asks that security requirements be sent to third parties who supply components, and PO.5.1 asks that each environment be separated and protected, with multi-factor sign in and network separation among NIST’s examples.

Guarding the code and every release

The PS practices protect what has been built. PS.2.1 suggests posting cryptographic hashes for release files on a well secured website and code signing through an established certificate authority; PS.3.2 asks for provenance data on every component, for example in an SBOM.

Writing and checking the software itself

The PW practices cover design, components, code, build settings, review, testing and defaults. PW.2.1 wants the design checked by a qualified person who was not involved in it, by automation in the toolchain, or both.

After release: finding and fixing

The RV practices start with RV.1.3, a policy for vulnerability disclosure and remediation, and end with RV.3.1, finding the root cause of each issue so the same mistake stops recurring.

Why is there no PW.3 in the current version?

It was merged. Version 1.1 moved PW.3, on third party software, into PW.4 and PO.1.3, which is why the numbering jumps from PW.2 to PW.4.

Threat modeling before the first line of code

Threat modeling is analyzing a picture of the system to find security and privacy concerns. The Threat Modeling Manifesto frames it as four questions, and its answer to who should do it is anyone concerned about the privacy, safety and security of the system.

A four step vertical flow of the threat modeling questions, each with the record it produces: a system drawing, a threat list, decisions with owners and a review.A four step vertical flow of the threat modeling questions, each with the record it produces: a system drawing, a threat list, decisions with owners and a review.
The four threat modeling questions and what each one leaves behind.

OWASP’s threat modeling page lists what a model usually holds: a description of the subject, assumptions that can be challenged later, potential threats, actions to reduce each one and a way to validate the model. The output feeds design, development, testing and work after deployment, which is why SSDF task PW.1.1 names threat modeling among the ways to assess risk.

  • Draw the system, including third party services and where data crosses a trust boundary.
  • List threats per component and per data flow.
  • Decide fix, control or accept for each, and name an owner.
  • Turn each fix into a ticket and each control into a test.
  • Repeat when the design changes, not once a year.

How small can a first threat model be?

One feature, one diagram and the people who designed it. Start with the feature that handles sign in, payments or personal data, then work outward one area at a time.

OWASP SAMM: measuring how mature the program is

OWASP’s Software Assurance Maturity Model measures the program rather than one application. The SAMM model defines five business functions with three security practices each, fifteen in all, and each practice holds activities in two streams across three maturity levels.

OWASP SAMM business functions and practices
Business functionIts three security practices
GovernanceStrategy and Metrics; Policy and Compliance; Education and Guidance
DesignThreat Assessment; Security Requirements; Secure Architecture
ImplementationSecure Build; Secure Deployment; Defect Management
VerificationArchitecture Assessment; Requirements-driven Testing; Security Testing
OperationsIncident Management; Environment Management; Operational Management

SAMM is built to be measurable and does not expect every company to reach the top level everywhere. Each organization picks a target maturity for each practice, assesses where it stands and closes the gaps in planned rounds.

Which SAMM level should a small team aim for?

The one that matches its risk, practice by practice. A small product team often starts at level one across the board and moves to level two first in Threat Assessment, Security Requirements and Security Testing, where the payoff shows soonest.

A security questionnaire on your desk?Send the questions your buyer asked. You get a free proposal covering the practices, records and code changes behind each answer.

Get a free proposalTell us about a project you’d like help with.

OWASP ASVS: the requirement list for one application

The Application Security Verification Standard gives developers a list of requirements for secure development and testers a basis for checking them. Version 5.0.0 was published on May 30, 2025, and the ASVS project page links the full standard in PDF, Word and spreadsheet form.

A bar chart of ASVS 5.0 levels: about 20% of requirements at Level 1, about 50% at Level 2 and about 30% at Level 3.A bar chart of ASVS 5.0 levels: about 20% of requirements at Level 1, about 50% at Level 2 and about 30% at Level 3.
OWASP ASVS 5.0 levels by share of requirements.

The standard holds about 350 requirements across 17 chapters, from Encoding and Sanitization through Authentication, Authorization, OAuth and OIDC, Cryptography and Configuration to Security Logging and Error Handling. OWASP’s own example: an early stage startup holding little sensitive data might aim for Level 1, while a bank would struggle to justify less than Level 3 for online banking.

  • Pick a level per application, based on the data it holds and what users expect.
  • Drop chapters that do not apply, such as WebRTC when the product has no calls.
  • Cite requirements with the version, for example v5.0.0-1.2.5, because numbers change between releases.
  • Write the chosen level into contracts; OWASP names procurement as one of its uses.

Which ASVS level fits a SaaS product?

Level 2 for most. ASVS says most applications should be striving for it, and it covers about 70% of the requirements once Level 1 is included.

The OWASP Top 10:2025, read as design decisions

The OWASP Top 10:2025 is an awareness document that reflects broad agreement on the most critical risks to web applications. Each risk maps to a decision a team makes in design or code.

OWASP Top 10:2025 risks and the decisions that answer them
OWASP Top 10:2025 riskThe design decision that answers it
A01 Broken Access ControlCheck permissions on the server for every request, deny by default
A02 Security MisconfigurationA documented secure baseline, applied the same way in every environment
A03 Software Supply Chain FailuresApproved components, pinned versions and an SBOM per release
A04 Cryptographic FailuresProven libraries, current protocols and keys kept out of code
A05 InjectionParameterized queries and encoding at every output
A06 Insecure DesignA threat model before building and a design review before release
A07 Authentication FailuresMulti-factor sign in and rate limits on login and reset
A08 Software or Data Integrity FailuresSigned releases and checks on updates and serialized data
A09 Security Logging and Alerting FailuresLogs for sign in, permission and admin events, with alerts someone reads
A10 Mishandling of Exceptional ConditionsErrors that fail safely and never leak internals

The list is a floor, not a test plan. ASVS turns these risks into checkable requirements, and the threat model shows which ones matter most for a given product.

Software bills of materials and the supply chain

An SBOM is a formal record of the components used to build software and how they relate, as NTIA’s minimum elements report of July 12, 2021 defines it. Software supply chain failures sit third on the OWASP Top 10:2025, and SSDF task PS.3.2 asks for provenance data on every component of a release.

  • Supplier name.
  • Component name.
  • Version of the component.
  • Other unique identifiers.
  • Dependency relationship.
  • Author of the SBOM data.
  • Timestamp.

NTIA sorts the minimum elements into three areas: those data fields, automation support so SBOMs are generated and read by machines, and practices covering how often they are made, how deep they go, known unknowns, delivery, access control and correcting mistakes.

Which SBOM format should we use?

SPDX or CycloneDX, both named by NTIA along with SWID tags. SPDX is recognized as the international standard ISO/IEC 5962:2021; CycloneDX, developed by OWASP and Ecma International, is at version 1.7 and standardized as ECMA-424. Pick the one your buyers and tools read, and generate it in the build.

Release integrity: hashes, signing and archived builds

A customer should be able to tell that the file they installed is the one you built. SSDF practice PS.2 exists for that, and its examples are concrete: publish cryptographic hashes for release files on a well secured website, and sign code through an established certificate authority so operating systems can check signatures.

  • Build releases in the pipeline, never on a laptop.
  • Publish a hash for every release file.
  • Sign installers, packages and container images.
  • Review the signing process and protect the signing keys.
  • Archive each release with its build data and SBOM, as PS.3 asks.

Why archive old releases?

When a vulnerability turns up after launch, the team needs the exact files and components that shipped to judge who is affected. PS.3 asks for that archive for this reason.

Safe defaults and protected build environments

SSDF practice PW.9.1 asks for a secure baseline: every setting that affects security is configured so the defaults are safe and do not weaken the protections the platform provides. Testing the defaults belongs in the release checks.

PW.6 turns on the protections compilers, interpreters and build tools offer, and PO.5 protects the environments where software is built. NIST’s examples include multi-factor, risk based sign in for each environment and network separation between development, build, test and production.

  • Ship with the most protective settings on, and let customers relax them knowingly.
  • Keep build settings in version control.
  • Separate production credentials from every other environment.
  • Require multi-factor sign in for repositories, pipelines and cloud consoles.

Code review, testing and the tools behind them

SSDF practice PW.7 lets each organization decide when to use code review, a person reading the code, and code analysis, tools finding issues alone or with a person. PW.8 asks that testing be scoped and its results recorded and triaged in the team’s issue tracker.

Security checks, what they look at and when they run
Kind of checkWhat it looks atWhen it runsSSDF practice
Peer code reviewLogic, permissions and anything a tool cannot judgeEvery pull requestPW.7
Static analysisSource code patterns that lead to vulnerabilitiesEvery commit or pull requestPW.7
Dependency scanningKnown issues in third party componentsEvery build, and daily on released codePW.4, RV.1
Secret scanningKeys and passwords committed by mistakeEvery commitPS.1, PO.5
Dynamic testingThe running application, from the outsideBefore release, against a test environmentPW.8
Penetration testingWhat a skilled attacker could chain togetherBefore major releases and on a schedulePW.8

Can automated scanning replace a code review?

It cannot judge intent. Scanners are fast and consistent on known patterns; a reviewer catches broken permission logic and business rules a tool has no way to read, so most teams need both. Our quality assurance testing services cover the functional side.

Scanning belongs in the pipeline so every change is checked the same way; our DevOps services set that up, and website security services cover audits and penetration tests of what is already live.

After release: disclosure, triage and root causes

No release is free of flaws, so the response matters as much as the build. SSDF task RV.1.3 asks for a policy on vulnerability disclosure and remediation, with a program that makes it easy for security researchers to report, and, as a further example, a product security incident response team.

A five point timeline: receive, confirm, rank, fix and learn.A five point timeline: receive, confirm, rank, fix and learn.
Vulnerability response as a sequence of five steps.

RV.2.1 asks that each vulnerability be analyzed for risk before planning the fix, and RV.3.1 that its root cause be recorded, so the team fixes the pattern across the code rather than one instance.

Does a small company need a disclosure policy?

Yes. A short page saying where to report, what to include and how quickly you reply costs little, and it gives researchers a better option than posting a flaw in public.

Securing SaaS products: tenants, sign in and data

Securing SaaS adds problems a single customer app never has. Many customers share one system, so a single missing check can expose one tenant’s data to another, which is why broken access control leads the OWASP list.

  • Scope every query and every file path to the tenant on the server.
  • Offer single sign on and multi-factor sign in, and test both against ASVS chapters on authentication, sessions, authorization and OAuth.
  • Keep audit logs customers can read for sign in, permission and admin events.
  • Encrypt data in transit and at rest, and keep keys out of code.
  • Write the security requirements enterprise buyers ask about into the backlog early.

Our SaaS development work covers these features, and products holding health data follow the safeguards on our HIPAA compliant software development page.

What changes for app development security on phones?

The app runs on a device you do not control. Secure app development keeps secrets off the device, checks every request on the server, stores tokens in the platform’s protected storage and treats anything the app sends as untrusted input.

Software development security best practices, as one checklist

These are the best practices for secure software development we would set up first on any product, each tied to the standard it comes from.

A drawn card with ten rows: security requirements, threat model, design review, review and scan results, SBOM, release integrity, archived releases, secure defaults, disclosure policy and root cause notes.A drawn card with ten rows: security requirements, threat model, design review, review and scan results, SBOM, release integrity, archived releases, secure defaults, disclosure policy and root cause notes.
A drawn checklist card of ten records a software buyer can ask for.
  1. Write security requirements and choose an ASVS level (PO.1).
  2. Threat model each major feature before building it (PW.1).
  3. Have the design reviewed by someone who did not design it (PW.2).
  4. Prefer maintained components and scan them on every build (PW.4).
  5. Follow a secure coding standard and review every change (PW.5, PW.7).
  6. Test the running application and triage findings in the tracker (PW.8).
  7. Ship with safe defaults (PW.9).
  8. Protect repositories, pipelines and environments with multi-factor sign in (PS.1, PO.5).
  9. Publish hashes or sign releases, and archive them with an SBOM (PS.2, PS.3).
  10. Publish a disclosure policy and record root causes (RV.1, RV.3).

Building something new?Tell us what the product does and whose data it holds. The free proposal sets out the threat model, the ASVS level and the build.

Get a free proposalTell us about a project you’d like help with.

What secure development work costs

Pricing depends on scope, is subject to change and is set in the written proposal. The bands below are market figures, typical low to published high, for the security work most often added to a build.

published US price bands (market ranges only)
Type of workTypical low to published high
Web application security review$5,000 to $25,000 per project
CI/CD pipeline setup$5,000 to $20,000 per project
Enterprise SaaS features (SSO, audit, SOC 2 prep)$50,000 to $150,000 per project
US development rate$100 to $200 an hour
Horizontal range bars for three secure development jobs, each from typical low to published high.Horizontal range bars for three secure development jobs, each from typical low to published high.
Published US price bands (market ranges only) for a web application security review, CI/CD pipeline setup and enterprise SaaS security features.

The cost of security for software development is mostly time inside the build: requirements, reviews and tests sit in the same sprints as the features. For wider project figures, see custom software development and app development cost.

Pricing depends on scope and is confirmed in a written proposal. Published prices are subject to change.

How Progression’s team builds security into client software

Our developers, designers and strategists are based in New York and build web applications, mobile apps, APIs and SaaS products for companies across the US. The security work our team can take on includes:

  • Security requirements and an ASVS level agreed at the start.
  • Threat models for the product and for each major feature.
  • Design and code review, with scanning in the pipeline.
  • SBOMs, release hashes or signing, and archived builds.
  • A disclosure policy and response plan ready for launch.
  • The records buyers ask for, ready before the questionnaire arrives.

Can you take over security work on software another team built?

Yes. We start from the code, the pipeline and any past findings, run a threat model and a review, and set out the fixes in order of risk.

Each page below covers one part of building and running software safely.

What this page leaves out: the Progression Method, proprietary and reserved for clients. Reach out to learn more

Frequently asked questions

Which habits make up secure software development?
Written security requirements, threat modeling, reviewed and tested code, protected builds and releases, and a plan for vulnerability reports, so fewer flaws ship and the rest are fixed quickly.
What is the NIST Secure Software Development Framework?
It is NIST SP 800-218, version 1.1, published in February 2022: a core set of high level practices meant to be added to any development process to reduce vulnerabilities in released software.
How many practices does the SSDF contain?
Nineteen, in four categories: five that prepare the organization, three that protect the software, eight that produce well secured software and three that respond to vulnerabilities.
Does the SSDF require particular tools?
It leaves tools to the team. NIST describes the practices as outcome based and says no implementation example is required.
What are the four questions of threat modeling?
What are we working on? What can go wrong? What are we going to do about it? Did we do a good enough job?
What is OWASP SAMM?
It is the Software Assurance Maturity Model, which measures a security program across five business functions and fifteen practices, each with three maturity levels.
How is OWASP SAMM different from OWASP ASVS?
SAMM measures how mature the whole program is; ASVS lists the security requirements one application should meet and gives testers a basis for checking them.
Which ASVS level should our application meet?
Most applications should aim for Level 2. Level 1 is the starting point, and Level 3 is for applications that must show the highest assurance, such as online banking.
How many requirements does ASVS 5.0 have?
About 350, in 17 chapters. Version 5.0.0 was published on May 30, 2025.
What is on the OWASP Top 10:2025?
Broken Access Control, Security Misconfiguration, Software Supply Chain Failures, Cryptographic Failures, Injection, Insecure Design, Authentication Failures, Software or Data Integrity Failures, Security Logging and Alerting Failures, and Mishandling of Exceptional Conditions.
What is a software bill of materials?
A formal record of the components used to build software and their supply chain relationships, in NTIA’s definition.
What must an SBOM include at minimum?
Supplier name, component name, version, other unique identifiers, dependency relationship, author of the SBOM data and a timestamp, per NTIA’s minimum elements.
Which formats are used for SBOMs?
SPDX, recognized as ISO/IEC 5962:2021, and CycloneDX, standardized as ECMA-424, are the common choices; NTIA also names SWID tags.
How can customers check that a release has not been altered?
By comparing the file against a hash published on a well secured website, or by checking a code signature from an established certificate authority, as SSDF practice PS.2 suggests.
What is a vulnerability disclosure program?
A published way for security researchers to report flaws, backed by a policy for handling and fixing them, which SSDF task RV.1.3 asks for.
Should code review be manual or automated?
Both, in most cases. The SSDF lets each organization decide where people review and where tools analyze, and each catches problems the other misses.
What are the best practices for secure software development?
Written requirements, threat modeling, independent design review, maintained components, secure coding and review, testing, safe defaults, protected environments, signed and archived releases, and a disclosure policy.
How is securing SaaS different from securing other software?
Many customers share one system, so tenant isolation on every query, strong sign in, audit logs and access checks on the server carry more weight.
Who should own software development and security in a small company?
The development team, with one named person accountable for the security practices and outside specialists brought in for design reviews and penetration tests.
How much does security work on a software project cost?
Published US bands run $5,000 to $25,000 for a web application security review and $50,000 to $150,000 for enterprise SaaS security features. Progression sets its price in the written proposal.
Can Progression build our software around the SSDF and ASVS?
Yes. Our team can set security requirements and an ASVS level, threat model the product, build review, scanning, SBOMs and signing into the pipeline, and keep the records buyers ask for.

A security questionnaire on your desk?Send the questions your buyer asked. You get a free proposal covering the practices, records and code changes behind each answer.

Get a free proposalTell us about a project you’d like help with.

Get a free marketing proposal

Tell us what you are trying to grow and we will come back with a plan, not a pitch deck. Same-day reply on weekdays.

Privacy Preferences
When you visit our website, it may store information through your browser from specific services, usually in form of cookies. Here you can change your privacy preferences. Please note that blocking some types of cookies may impact your experience on our website and the services we offer.
Contact Us