Engineer working on a laptop connected to a device test rig, surrounded by wiring harnesses and test hardware

Firmware vulnerabilities & fixes

ENISA Good Practices for Security of IoT: Secure Software Development Lifecycle, explained

Last updated:

Good Practices for Security of IoT — Secure Software Development Lifecycle, published by the European Union Agency for Cybersecurity (ENISA) in November 2019 (ISBN 978-92-9204-316-2, DOI 10.2824/742784), is a free, 132-page report that turns "secure by design" from a slogan into a phase-by-phase checklist for building IoT software. Its executive summary states the goal directly: establishing "secure development guidelines across the IoT ecosystem" is "a fundamental building block for IoT security," and the report "tackles one aspect for achieving security by design" that ENISA's earlier, more horizontal IoT work had flagged as a gap. The report carries ENISA's standard copyright notice — "reproduction is authorised provided the source is acknowledged" — and states it "must be accessible free of charge."

Key takeaways

  • The 2019 report predates both the current consumer IoT baseline and the EU Cyber Resilience Act, so it cites neither.
  • Synthetic example; the product and findings are invented for illustration.
  • Treating the report as a certificate.
  • Use the six-phase table as a gap-analysis checklist first, not Annex A's full practice list — identify which phase currently has the thinnest security evidence before diving into all 16 domains at once.

Six SDLC phases, plus a cross-cutting security layer

Section 1 of the report notes that when ENISA's consulted experts were asked how many phases the SDLC comprises, "the vast majority of them considered that the SDLC comprises up to six phases." Section 2 addresses each in turn:

Phase Core security focus, per the report
Requirements Identify security requirements alongside business/functional ones from the start; use threat modelling (commonly STRIDE) and quality gateways/bug bars to keep security requirements testable and traceable; consider hardware implications (e.g. a secure-boot requirement implying a need for a hardware security module)
Software Design Translate requirements into architecture using a risk-based approach; refine threat models and run attack-surface analysis; apply the CIA triad plus, for IoT specifically, safety implications from the cyber-physical boundary (sensors/actuators)
Development/Implementation Use trusted security libraries and frameworks, OS/protocol built-in security functions, secure coding guidelines, and checks on external dependencies before and during use
Testing and Acceptance Combine automated and manual testing, static and dynamic analysis, fuzzing and penetration testing; code review catches vulnerabilities (e.g. logic bombs) other techniques miss
Deployment and Integration Secure the production environment and integration of all software elements after successful testing and release approval
Maintenance and Disposal Ongoing incident management, vulnerability assessment and secure remote (over-the-air) update delivery; secure end-of-life disposal with data erasure to prevent leakage of cached sensitive data (health records, Wi-Fi credentials, etc.)

Section 2.7, "Security in SDLC," layers a cross-cutting discipline on top of all six phases: security maturity assessment (the report names Common Criteria, CMMI, BSIMM, IEC 62443-4-1 and OpenSAMM as reference models), security gates with defined metrics between phases, and continuous documentation management — the report calls documentation "a cross-cutting activity during the SDLC process that is often not considered as essential" despite its role in making a complex IoT build "understandable, traceable, and subject to monitoring and auditing."

How does the report organize assets and threats?

Chapter 3 opens with an asset taxonomy (Table 1), grouping what needs protecting into categories including Human Factor (Business Owners/BI analysts, Software Team roles such as Testers/Integrators/Developers/Architects/Designers, Security Team roles such as the CISO/security engineers/penetration testers/incident responders, Operators, End Users, Contractors, and Decision Makers), Software Design assets (business and security requirements, metrics, quality gateways, use cases, specifications, design tools), and Software Development assets (underlying hardware components, firmware, operating systems, compilers and simulation environments). The report is explicit that "the lowest level of the taxonomy is indicative and not exhaustive" — it is a working reference, not a closed list.

Section 3.2 builds a corresponding threat taxonomy, and Section 3.3 grounds both taxonomies in three worked attack scenarios: Insecure Credentials in Embedded Devices, Lack of Flexibility to Secure Communications, and Insecure Software Dependencies in Cloud Services — each walked through to show which assets and SDLC phases a realistic attack chain touches.

16 good-practice domains: People, Processes, Technologies

Section 4.2 explains the methodology behind the report's central deliverable: "extensive desktop research" plus stakeholder interviews produced "a list of 16 domains that provide a comprehensive view of the Secure IoT SDLC landscape," organized into three groups:

Group Domains (6, 7 or 3 each)
People (3) Training and Awareness; Roles and Privileges; Security Culture
Processes (6) Third-Party Management; Operations Management; SDLC Methodology; Secure Deployment; Security Design; Internal Policies
Technologies (7) Access Control; Third-Party Software; Secure Communication; Secure Code; Security Reviews; Security of SDLC Infrastructure; Secure Implementation

Each domain contains individually numbered practices with a short description — for example, under Training and Awareness: PE-01 "define a corporate strategy for specific security training," PE-02 "promote security awareness at all organisation levels," PE-03 assess whether security skills need updating, and PE-04 allocate resources to track emerging security topics. Annex A maps every numbered practice in full detail to the threats and external references it addresses, which is where a team building an actual compliance or audit checklist from this report should start.

Where this report sits relative to other secure-SDLC and IoT standards

ENISA's report is explicitly positioned as one building block among several, not a standalone framework: it names OWASP IoT Top 10, the OWASP Application Security Verification Standard, and the SAFECode Fundamental Practices for Secure Software Development paper among its own supporting references, and Section 2.7 points to IEC 62443-4-1 as one of several maturity models a team can use to benchmark its own secure-development process. A device maker following OWASP's firmware security testing methodology or working toward IEC 62443-4-1 secure-development-lifecycle conformance will recognize most of this report's 16 domains already — ENISA's contribution is organizing them specifically around the IoT asset and threat taxonomy in Chapter 3, rather than introducing a competing scheme. ENISA's 2020 Guidelines for Securing the Internet of Things report, in turn, explicitly builds on this one, extending the same good-practice logic from the software process out to the full physical-and-logical IoT supply chain (semiconductor fabrication, component manufacturing, distribution), and a 2017 ENISA baseline study on IoT security recommendations set the horizontal context both narrower studies refine.

How do the six phases map to EN 303 645 and the CRA?

The 2019 report predates both the current consumer IoT baseline and the EU Cyber Resilience Act, so it cites neither. A device team can still use its phases as the place where each later requirement gets produced. The table below is our mapping, not ENISA's; the provision names come from ETSI EN 303 645 V3.1.3 and Regulation (EU) 2024/2847.

ENISA SDLC phase EN 303 645 provisions it feeds CRA hook Evidence a firmware team keeps
Requirements 5.1 no universal default passwords; 5.8 personal data Annex I Part I: risk-based requirements; Article 13(2) cybersecurity risk assessment Threat model, security requirement list with IDs
Software Design 5.5 communicate securely; 5.6 minimise exposed attack surfaces Annex I Part I(2): limit attack surfaces, secure by default configuration Architecture and interface inventory, debug-port decision record
Development/Implementation 5.4 securely store sensitive security parameters; 5.13 validate input data Annex I Part II(1): identify and document components, including an SBOM Build log, pinned dependency list, CycloneDX or SPDX SBOM per release
Testing and Acceptance 5.7 ensure software integrity Annex I Part II(3): effective and regular tests and reviews Static analysis, fuzzing and pen-test reports tied to the release tag
Deployment and Integration 5.3 keep software updated; 5.12 easy installation and maintenance Annex I Part I(2): security updates, automatic where applicable Signed update images, secure-boot key custody records
Maintenance and Disposal 5.2 manage reports of vulnerabilities; 5.11 easy deletion of user data Annex I Part II; Article 13(8) support period; Article 14 reporting from 11 September 2026 Disclosure policy, advisory archive, end-of-support date on the product page

Two limits matter. EN 303 645 is a consumer IoT standard and is not, by itself, a CRA harmonised standard, so no row in this table grants a presumption of conformity under the Regulation. And the CRA reaches further than the 2019 report: the report treats vulnerability handling as part of Maintenance, while the Regulation turns it into a set of duties that run for the whole support period.

Worked example: a gap review for a smart thermostat

Synthetic example; the product and findings are invented for illustration. A five-person firmware team ships a Wi-Fi thermostat on an RTOS with a TLS library, an MQTT client and a vendor SDK. Walking the six phases against the report's domains takes about one working day:

  1. RequirementsNo written threat model exists. The team writes a one-page STRIDE pass on the cloud link, the local setup access point and the UART header, which the Security Design domain asks for.
  2. DevelopmentThe vendor SDK is copied into the tree without a version record. That is a Third-Party Software gap; the fix is a pinned manifest and an SBOM produced at build time.
  3. TestingOnly functional tests run in CI. The team adds a static-analysis job and a short fuzzing target for the MQTT payload parser (Security Reviews, Secure Code).
  4. DeploymentUpdates are signed, but the signing key sits on a build laptop. Moving it to a hardware token closes a Security of SDLC Infrastructure gap.
  5. MaintenanceThere is no published contact for vulnerability reports. A security.txt file and a short disclosure page close the most visible gap and line up with EN 303 645 provision 5.2.

What are the common pitfalls?

  • Treating the report as a certificate. There is no ENISA conformity scheme for it; claims such as "ENISA SDLC compliant" have nothing behind them.
  • Stopping at People and Processes. Training plans are easy to document; the Technologies domains are where shipped firmware actually differs.
  • One SBOM at launch. Components drift with every release; an SBOM that is not regenerated per build stops matching the device in the field.
  • Forgetting disposal. Factory reset that leaves Wi-Fi credentials in flash fails both the report's disposal practice and EN 303 645 provision 5.11.

Why this matters for firmware and connected-device teams

No regulation known to us names this specific ENISA report as a required control for any legal duty — no claim that following it satisfies a specific regulatory requirement (such as the EU Cyber Resilience Act's secure-development obligations) is confirmed by the actual regulatory text itself. Where the report earns its keep is as a free, structured gap-analysis tool: a team can walk its own six-phase SDLC against the report's 16 domains, see which of the People/Processes/Technologies practices it already has evidence for, and use the gaps to prioritize where a firmware SBOM or dependency review would close the most exposure fastest, since Chapter 3's threat taxonomy consistently ties insecure third-party and legacy dependencies to concrete attack scenarios.

A workable starting point

  1. Use the six-phase table as a gap-analysis checklist first, not Annex A's full practice list — identify which phase currently has the thinnest security evidence before diving into all 16 domains at once.
  2. Treat threat modelling as starting in Requirements, not DesignThe report is explicit that threat modelling "starts in the requirements phase with the identification of critical assets and is completed in the software design phase" — teams that defer it to design alone miss the business- and hardware-driven threats requirements analysis is meant to surface.
  3. Don't skip Maintenance and Disposal good practicesjust because a product has already shipped — the report flags secure over-the-air update delivery and end-of-life data erasure as ongoing SDLC obligations, not one-time launch tasks.
  4. Cross-reference Annex A's standards mapping before building a new checklist from scratch— many of the 16 domains already map to OWASP, SAFECode or IEC 62443-4-1 controls a team may partially satisfy through existing tooling.
  5. Pair this report with [ENISA's 2020 supply chain guidelines](/guides/enisa-guidelines-securing-iot-supply-chain/)if third-party hardware or contract manufacturing is involved — this 2019 report's scope stops at the software process; physical supply chain risks (component sourcing, assembly, distribution) are the later report's subject.

SoftWSP's firmware and SBOM review maps a shipped device's actual software composition and known vulnerabilities against the dependency- and update-related gaps this report's SDLC good practices are designed to catch before release: see device & firmware security.

Where AI fits into a secure-SDLC gap review

SoftWSP's firmware and SBOM review identifies a shipped device's actual component versions and matches them against known CVEs, KEV and EPSS data — the deterministic evidence a secure-SDLC gap review needs. AI drafts the narrative that connects those findings back to this report's 16 good-practice domains: an outdated, vulnerable third-party library maps to the Third-Party Software and Third-Party Management domains, a missing update mechanism maps to Secure Deployment, and so on, each sentence citing the specific component and CVE behind it. AI does not assess People or Processes domains like training or internal policy — those have no firmware evidence to draw on — and an engineer reviews every drafted mapping before it becomes part of a gap report.

This guide summarises the public ENISA report (© European Union Agency for Cybersecurity (ENISA), 2019; reproduction authorised provided the source is acknowledged); it does not constitute legal or compliance advice.

Put this guide to work

Frequently asked questions

What is ENISA's Good Practices for Security of IoT report?

"Good Practices for Security of IoT — Secure Software Development Lifecycle" is a report published by the European Union Agency for Cybersecurity (ENISA) in November 2019 (ISBN 978-92-9204-316-2, DOI 10.2824/742784), authored by ENISA with an acknowledgements list of over 40 industry and academic contributors (including representatives from ARM, Cisco, Microsoft, Siemens, Kaspersky, OWASP and INCIBE). Per its own executive summary, the study "introduces good practices for IoT security, with a particular focus on software development guidelines for secure IoT products and services throughout their lifetime," targeted mainly at "IoT software developers, integrators and platform and system engineers." It is released under ENISA's standard copyright notice: "Reproduction is authorised provided the source is acknowledged," and the publication itself states it "must be accessible free of charge."

What are the six phases of ENISA's IoT secure SDLC?

Section 1 states that during the study, "the vast majority" of consulted experts considered the SDLC to comprise six phases, which Section 2 then addresses in turn: Requirements, Software Design, Development/Implementation, Testing and Acceptance, Deployment and Integration, and Maintenance and Disposal. Section 2.7, "Security in SDLC," adds a cross-cutting layer on top of all six — security maturity assessment (citing models such as Common Criteria, CMMI, BSIMM, IEC 62443-4-1 and OpenSAMM), security gates and metrics between phases, and ongoing documentation management.

How does the report organize assets and threats?

Chapter 3 provides an asset taxonomy (Table 1) organized into groups such as Human Factor (Business Owners, Software Team roles, Security Team roles, Operators, End Users, Contractors), Software Design assets (requirements, security assessments, design tools) and Software Development assets (underlying hardware/firmware/OS components, guidelines), explicitly noting the taxonomy's lowest level is "indicative and not exhaustive." Section 3.2 then provides a threat taxonomy, and Section 3.3 walks through three worked attack scenarios: Insecure Credentials in Embedded Devices, Lack of Flexibility to Secure Communications, and Insecure Software Dependencies in Cloud Services.

What are the People/Processes/Technologies good practices, and how many are there?

Section 4.2 states that desktop research and stakeholder interviews produced "a list of 16 domains that provide a comprehensive view of the Secure IoT SDLC landscape," grouped into three categories: **People** (3 domains — Training and Awareness, Roles and Privileges, Security Culture), **Processes** (6 domains — Third-Party Management, Operations Management, SDLC Methodology, Secure Deployment, Security Design, Internal Policies), and **Technologies** (7 domains — Access Control, Third-Party Software, Secure Communication, Secure Code, Security Reviews, Security of SDLC Infrastructure, Secure Implementation). Each domain contains individually numbered practices (for example PE-01 through PE-04 under Training and Awareness), with full detail, threat mappings and references collected in Annex A.

How does this relate to ENISA's other IoT security publications?

The report builds on ENISA's 2017 **Baseline Security Recommendations for IoT**, which addressed IoT ecosystem security horizontally; this 2019 study narrows in on the software development process specifically as one building block of that broader security-by-design goal. ENISA's later (November 2020) **[Guidelines for Securing the Internet of Things](/guides/enisa-guidelines-securing-iot-supply-chain/)** report explicitly states it "builds up on" this 2019 SDLC study, extending the lens from the software process to the full physical-and-logical IoT supply chain, including semiconductor fabrication, component manufacturing and distribution stages this 2019 report does not cover.

Sources

  1. ENISA: Good Practices for Security of IoT — Secure Software Development Lifecycle — publication page
  2. ENISA: Good Practices for Security of IoT — Secure Software Development Lifecycle, full report PDF (November 2019, ISBN 978-92-9204-316-2, DOI 10.2824/742784)
  3. ENISA: Baseline Security Recommendations for IoT (2017) — the horizontal study this report builds on
  4. OWASP: OWASP Application Security Verification Standard (ASVS) Project
  5. ETSI EN 303 645 V3.1.3 (2024-09): Cyber Security for Consumer Internet of Things: Baseline Requirements
  6. Regulation (EU) 2024/2847 (Cyber Resilience Act), EUR-Lex
Was this guide helpful?

In this cluster

ENISA Good Practices for Security of IoT: Secure Software Development Lifecycle, explained