Go back
How to Implement a SIEM: A Step-by-Step Guide for German Companies

How to Implement a SIEM: A Step-by-Step Guide for German Companies

Securevisio
17.11.2026

A SIEM implementation is a project of 8 phases, from scoping and use cases to handover into steady state, and in Germany it includes a legal track with the works council (Betriebsrat) and data protection law: the EU General Data Protection Regulation (Datenschutz-Grundverordnung, DSGVO), supplemented in Germany by the Federal Data Protection Act (Bundesdatenschutzgesetz, BDSG). Detection is the weak point that makes this project necessary: among German companies that suffered damage from cyberattacks in the last 12 months, 59% name insufficient detection of security incidents as a factor that contributed to the successful attack (Bitkom, Wirtschaftsschutz 2026, n=580). This article shows the project in the order it should run: the German rules and BSI documents, scope and use cases, the works council and DSGVO track, log source waves, parsing, correlation rules, tuning, integrations, the operating model, a planning timeline, KPIs and the most common failure modes.

What does a SIEM implementation project consist of?

A SIEM implementation project consists of 8 phases that turn a purchased platform into a working detection capability. A Security Information and Event Management (SIEM) system collects log data from many systems, correlates it, and raises alerts. The project makes that capability real for one specific organisation.

PhaseResultOwner
1. Scoping and use casesApproved scope statement and a list of starting use casesCISO or ISB, IT lead
2. Legal and works council trackBetriebsvereinbarung (works agreement), DSGVO basis, Löschkonzept (deletion concept)Datenschutzbeauftragte, HR, Betriebsrat
3. Log source wavesPlanned sources connected in ordered wavesSIEM engineer, system owners
4. Parsing and normalisationEvery source readable in one common schemaSIEM engineer
5. Correlation and ATT&CKTested rules mapped to attacker techniquesSOC analyst, detection engineer
6. TuningAlert volume the team can processSOC lead
7. IntegrationTicketing, SOAR and CMDB connectedSOC lead, IT operations
8. Handover and KPIsSteady state with measured performanceSOC lead, CISO

The legal track runs in parallel with phases 1 to 3 from the first week. Phase 2 decides when the first log data from employees may be processed. Planning the project without it produces a platform that cannot go live.

Which German rules and BSI documents shape a SIEM project?

Four sources shape a SIEM project in Germany: the BSIG (the German act on the Federal Office for Information Security), the BSI IT-Grundschutz modules, the BSI guidance on attack detection, and the BSI minimum standard on logging. None of them names a SIEM product. They define duties and give reference points for the project.

Section 30 Abs. 1 BSIG requires suitable, proportionate and effective technical and organisational measures, and the company must document that it meets this duty. Section 30 Abs. 2 Satz 2 Nr. 2 BSIG lists the handling of security incidents as a minimum content of those measures. Nr. 6 lists concepts and procedures for assessing the effectiveness of the measures. A SIEM supports both, because it detects incidents and produces evidence of how well detection works. Section 38 Abs. 1 BSIG makes the implementation and monitoring of these measures a duty of the Geschäftsleitung (the management board). The Geschäftsleitung may assign the work to others, such as a CISO or a service provider, but it keeps the final responsibility. Under § 38 Abs. 2 BSIG, managing directors who culpably breach this duty are liable to their own entity for the resulting damage, under the company-law rules for its legal form, such as § 43 GmbHG or § 93 AktG. The NIS2 directive and its obligations explain the wider duty set.

The BSI assessment shows the gap: in the BSI report on the state of IT security (Die Lage der IT-Sicherheit in Deutschland 2025), about 80% of operators of critical infrastructure run an information security management system at maturity level 3 or higher, while only 48% reach level 3 or higher for attack detection systems. Detection lags behind the management layer.

SourceWho it bindsWhat it means for the project
BSIG § 30Entities in scope of the NIS2 implementation lawDocument measures for incident handling and effectiveness testing
BSIG § 31 Abs. 2Betreiber kritischer Anlagen (operators of critical facilities, KRITIS)Deploy a System zur Angriffserkennung (SzA, attack detection system)
IT-Grundschutz DER.1 and OPS.1.1.5Reference for everyone, mandatory only where a company has adopted GrundschutzDetection requirements and central logging
BSI Orientierungshilfe SzAAddressed to KRITIS operators and auditorsChecklist for implementation and audit
BSI Mindeststandard Protokollierung und DetektionFederal administrationDetailed reference for logging and detection

What does § 31 BSIG require from KRITIS operators?

Section 31 Abs. 2 BSIG requires operators of critical facilities to deploy attack detection systems. The systems must capture and evaluate parameters and characteristics from live operation continuously and automatically, and they must follow the state of the art. Section 2 Nr. 41 BSIG defines an attack detection system as processes supported by technical tools and organisational integration that detect attacks on information technology systems. A SIEM together with its operating process fits this definition, but the statute does not use the word SIEM.

The evidence cycle is a separate duty. Section 39 BSIG requires operators to prove their measures on a date set by the BSI and then every 3 years. The project consequence is direct: document the detection scope for the critical systems first, because auditors will ask for exactly that.

Which BSI documents bind whom: Grundschutz modules, Orientierungshilfe, Mindeststandard?

Only one of the three BSI document types binds anyone by law: the Mindeststandard binds the federal administration. The documents differ in who they address.

  • Mindeststandard Protokollierung und Detektion (Version 2.1): binds the Einrichtungen der Bundesverwaltung under § 44 BSIG. A private company is not bound, but it may use the standard as a detailed reference.
  • Orientierungshilfe zum Einsatz von Systemen zur Angriffserkennung (Version 1.1, 18 November 2024): addresses KRITIS operators and auditors as a reference point for implementation and audit. Any company may use it as a free checklist.
  • IT-Grundschutz modules DER.1 and OPS.1.1.5 (Kompendium Edition 2023): serve as a reference for every organisation. DER.1 covers the detection of security-relevant events, and OPS.1.1.5 covers logging. The BSI is moving to a new methodology called Grundschutz++, so the edition status needs to be checked at the time of the project.

How do you define scope and use cases for a SIEM?

You define scope and use cases by starting from risks and duties, not from the list of available log sources. A use case is a specific attack or misuse scenario that the SIEM must detect, for example, a compromised account, a brute force attack, or a change to a privileged group. The scope statement lists the use cases, the systems they cover, the owners and the success criteria.

Three inputs set the scope: the incident handling duty in § 30 Abs. 2 Satz 2 Nr. 2 BSIG, the critical assets of the company, and the threat scenarios that apply to its sector. The asset inventory and the criticality rating should come from the CMDB (configuration management database, the register of systems and their owners), as described in how a CMDB strengthens SIEM and vulnerability management.

The plan should reject the approach of collecting everything first. It raises cost, produces noise, and collects personal data without a purpose. Priorities matter because budgets are limited: Bitkom reports that the average share of IT security in the IT budget is 18% in 2026. Five to eight starting use cases are enough for the first release. Further use cases follow once the first ones run stably.

Which use cases come first, and which log sources do they need?

The first use cases are those that stop the most common attack paths: account takeover, remote access abuse, privilege changes, ransomware preparation, log tampering and cloud administration abuse. Each one needs a small set of sources.

Use caseRequired sourcesATT&CK tacticExample event
Compromised accountIdentity logsInitial AccessLogin from two countries within minutes
Brute force or password spraying against remote accessVPN, firewallCredential AccessMany failed logins from one address
Changes to privileged groupsDirectory servicePrivilege EscalationNew member added to an administrator group
Ransomware precursorsEDR, serversImpactMass file renaming, backup deletion
Log tampering or log source silenceAll collectorsDefense EvasionAudit log cleared, source stops sending
Unusual cloud administration actionsCloud audit logsPersistenceNew access key created outside change window

What do the works council and data protection law require before logs are collected?

The works council and data protection law require agreements and documented decisions before the first employee-related log data is collected. A SIEM collects log data that can be assigned to named employees: logins, file access, web requests, administrator actions. Three parties have a role in this: the employer, the Betriebsrat and the Datenschutzbeauftragte (data protection officer).

The employer has a security duty under § 30 and § 38 BSIG and needs the logs to meet it. The Betriebsrat protects employees from monitoring of behaviour and performance. The Datenschutzbeauftragte checks the legal basis, data minimisation, and retention. All three interests are legitimate, and the project plan should treat this track as the critical path, because go-live depends on its result.

How does § 87 Abs. 1 Nr. 6 BetrVG apply to a SIEM?

Section 87 Abs. 1 Nr. 6 BetrVG gives the Betriebsrat a right of co-determination over the introduction and use of technical devices that are intended to monitor the behaviour or performance of employees. The Bundesarbeitsgericht (the Federal Labour Court) held in its decision of 6 December 1983 (1 ABR 43/81) that a technical device is intended for monitoring when it records behaviour or performance data that can be attributed to employees, regardless of the purpose the employer pursues. A SIEM records such data, so its introduction requires the Betriebsrat to take part. When both sides do not reach an agreement, the Einigungsstelle (conciliation board) decides under § 87 Abs. 2 BetrVG. The right applies only in companies that have a Betriebsrat.

A Betriebsvereinbarung for a SIEM normally covers these points:

  • Purpose: limitation to information security, with no performance control.
  • Data: which log sources and which data fields are collected.
  • Roles: which roles have access, and who may see which data.
  • Safeguards: a four-eyes principle for resolving pseudonyms, where the company uses them.
  • Evaluation: rules for how the logs may be evaluated, and which evaluations are excluded.
  • Retention and audit: the retention periods and the audit rights of the Betriebsrat.

Which DSGVO basis, retention period, and Löschkonzept does a SIEM need?

A SIEM needs a documented legal basis, a purpose for each data class, and a Löschkonzept that sets retention per data class. The usual basis is Art. 6 Abs. 1 lit. f DSGVO (legitimate interests), which applies unless the interests of the data subject override them. For the part of the processing that serves a legal duty, such as the KRITIS duty in § 31 Abs. 2 BSIG, Art. 6 Abs. 1 lit. c DSGVO (legal obligation) applies.

Art. 5 Abs. 1 lit. c DSGVO limits collection to what is necessary. Art. 5 Abs. 1 lit. e DSGVO allows storage only as long as the purpose requires. Art. 88 Abs. 1 DSGVO allows rules on employee data through collective agreements, which is why the Betriebsvereinbarung matters. The company should base SIEM processing on Art. 6 Abs. 1 lit. f DSGVO and, where it can, on a Betriebsvereinbarung. Since the judgment of the Court of Justice of the European Union of 30 March 2023 (C-34/21), there are serious doubts whether § 26 Abs. 1 Satz 1 BDSG can serve as a basis on its own. This point belongs on the review list of the Datenschutzbeauftragte. The company should also check whether a data protection impact assessment is required under Art. 35 Abs. 1 DSGVO. The check is needed because the law requires an assessment only where processing is likely to result in a high risk.

No statutory retention period applies to all SIEM logs. The company sets a documented period per data class, based on the purpose. The BSI Mindeststandard version 2.1 requires a defined retention period for the federal administration, and the Orientierungshilfe states that legal deletion periods must be observed.

Data classPurposeRetention rule to be set
Authentication and identity logsDetect account takeoverPeriod set by the company and documented
Network and firewall logsDetect attacks from outsidePeriod set by the company and documented
Endpoint and server logsInvestigate intrusionsPeriod set by the company and documented
Alerts and incident recordsEvidence of incident handlingPeriod set by the company and documented

In which order do you onboard log sources?

You onboard log sources in waves ordered by criticality and attack path, starting at the network boundary and working inwards. This follows the BSI Orientierungshilfe, which says that data sources should be prioritised by criticality and opened from the network boundary inwards.

WaveSourcesWhyTypical pitfall
1Identity (directory service, identity provider), firewall, VPNMost attacks use stolen credentials or remote accessService accounts missing from the data
2EDR, servers (Windows event logs, Linux authentication logs)Shows what happens after the first accessAudit policies not enabled on servers
3Cloud and SaaS audit logs, e-mailCovers systems outside the network boundaryLogging switched off by default
4OT and special systems, where relevantProtects production and critical processesCollection that disturbs availability

Wave 1 comes first for a measurable reason: 55% of attacked companies name inadequate identity and access management as a contributing factor (Bitkom, Wirtschaftsschutz 2026, n=580). Each wave ends with an acceptance test. Every planned source must deliver data, and the data must parse correctly. Only then does the next wave start.

Licensing affects the order. Licensing by data volume or EPS (events per second, the unit many SIEMs bill by) pushes teams to leave sources out, which breaks the link between use cases and sources. The hidden costs of SIEM and ingest-based pricing describe this effect. Licensing without EPS or data volume limits, as SecureVisio offers, removes that pressure from the source decision.

When does OT belong in the early waves?

OT (operational technology, the systems that control machines and plants) belongs in the early waves for KRITIS operators and plant operators whose critical processes depend on it. The Orientierungshilfe says that, at the system level, collection should start from the central critical systems, such as process control. The collection must be passive and must not affect availability. All other companies add OT last, after the identity, endpoint and cloud waves.

How do you parse, normalise and enrich log data?

You parse, normalise, and enrich log data in three steps: read each source format, convert the fields into one common schema, and add context. A platform collects data in three ways: passively (the source sends logs), actively (the platform queries the source) or through an agent (software on the source). Which data a correlation engine collects shows how one platform lists these methods.

The Orientierungshilfe states that data must be filtered, normalised, aggregated and correlated. Four tasks make this concrete:

  • Parse: extract fields such as user, source address and action from each raw message.
  • Normalise: map the fields into a common schema, so that one rule works across firewalls, servers and cloud services.
  • Fix time: use synchronised time sources and store the time zone, because correlation across sources fails with wrong timestamps.
  • Enrich: add asset criticality from the CMDB and user context from the directory service.

Two checks keep the data usable. The parse success rate per source shows how many messages were read correctly. An alert for a silent source reports when a source stops sending.

How do you build correlation rules and map them to MITRE ATT&CK?

You build correlation rules from use cases, give each rule one ATT&CK technique, and test it against a real or simulated event. A correlation rule is a condition that links events into an alert. Four rule types cover most needs: single event rules, threshold rules, sequence rules, and behaviour rules.

Rule typeLogicExample
Single eventOne event triggers the alertAudit log cleared
ThresholdA count within a time windowMany failed logins within 5 minutes
SequenceEvents in a defined orderFailed logins followed by a success
BehaviourDeviation from a baselineAdministrator logs in at an unusual time

Each rule carries an ATT&CK technique ID, an owner, and a test case. The MITRE ATT&CK framework is a public catalogue of attacker tactics and techniques, and the project uses it as a coverage map. The map shows which techniques have a rule and which do not. Coverage is a planning tool and not a quality score, because one tested rule can be worth more than five untested ones.

How do you baseline and tune a SIEM against alert fatigue?

You baseline and tune a SIEM by observing normal behaviour first and then adjusting rules until the team can process every alert. Alert fatigue is the state in which analysts receive so many alerts that real incidents are missed or ignored. The baseline period before alerting goes live is a planning assumption of several weeks, and its length depends on the size of the company and the number of sources.

Tuning follows a fixed order:

  1. Observe the alerts and record what they contain.
  2. Classify each alert as true positive, benign, or unclear.
  3. Suppress known-good activity, for example, a vulnerability scanner.
  4. Tune thresholds against the recorded data.
  5. Retire rules that produce noise and no value.
  6. Re-test the changed rules with the original test cases.

Every rule has an owner and a review date. A rule without an owner is a rule that nobody tunes. The false positive rate in the KPI section measures the result.

How do you integrate the SIEM with SOAR, ticketing, and the CMDB?

You integrate the SIEM with SOAR, ticketing, and the CMDB in three stages: ticketing and enrichment first, automated containment last. SOAR (security orchestration, automation, and response) runs defined response steps from playbooks. SOAR playbooks and what they automate cover the design of those playbooks.

IntegrationWhat flowsBenefit
TicketingAlert becomes a ticket, status returns to the SIEMTraceable handling and evidence
SOAREnrichment and containment steps run from playbooksFaster and consistent response
CMDBAsset criticality and owners appear in the alertCorrect prioritisation
IdentityMapping of user accounts to personsAttribution, subject to the Betriebsvereinbarung

Automated containment, such as disabling an account or isolating a host, needs approval rules. The rules state which actions run without approval, which need a human decision and who decides. Teams that skip this step risk disruption to production systems.

Which operating model fits: in-house SOC, managed SIEM or hybrid?

The operating model that fits is the one the company can staff, because a SIEM only works when someone analyses its alerts. The deciding question is who analyses alerts outside business hours. SOC versus SOC-as-a-Service compares the models in detail.

ModelWho analyses alertsFits when
In-house SOCOwn staff, including coverage outside business hoursThe company can hire and keep a team for round-the-clock coverage
Managed SIEMAn external providerThe company has no capacity for a SOC team
HybridOwn staff by day, provider at night and weekendsThe company has a small team and needs more coverage

Every model needs a named internal owner for the SIEM, for the rules and for decisions about response. The choice of the product is a separate decision and is not part of this guide.

What does a realistic SIEM implementation timeline look like?

A realistic SIEM implementation takes about 4–9 months for a mid-sized company and longer for KRITIS operators with OT. These durations are a planning assumption from typical project plans, not measured data. The real duration depends on the size of the company and the number of log sources. The technical deployment of the platform itself is the shorter part of this period: the legal track, the source waves and tuning set the overall duration.

PhaseIndicative duration (planning assumption)Depends on
Scoping and use cases2–4 weeksNumber of stakeholders
Legal and Betriebsrat track4–12 weeks, in parallelExisting agreements, Betriebsrat workload
Wave 13–6 weeksNumber of identity and network systems
Waves 2 and 34–8 weeks eachNumber and variety of sources
Parsing, rules and tuningOverlaps with the waves; tuning 8–12 weeks after the first alertsAlert volume, team capacity
Handover to steady stateAfter the last wave passes its acceptance testTest results

The legal track is the critical path. A delay there delays the go-live, however fast the technical work runs. A project plan should therefore start the Betriebsrat discussion in the first week.

Which KPIs show that the SIEM works?

Six KPIs show that a SIEM works: MTTD, MTTR, false positive rate, log source coverage, source health and rule coverage. Each one needs a definition that the team agrees on before go-live. The project then uses them as acceptance criteria per phase.

KPIDefinitionHow to measureDirection
MTTD (mean time to detect)Time from event to alertCompare event time with alert timeLower is better
MTTR (mean time to respond)Time from alert to containment or resolution; define once and keep itCompare alert time with ticket closureLower is better
False positive rateAlerts closed as benign divided by all alertsTicket classificationLower is better
Log source coverageConnected planned sources divided by planned sourcesSource inventoryHigher is better
Source healthSources silent longer than a set periodSilent source alertLower is better
Rule coverageATT&CK techniques with a tested ruleCoverage mapHigher, as a planning tool

The article sets no target values, because targets depend on the risk profile of the company. The cost of poor detection is high: the IBM Cost of a Data Breach Report 2026 (vendor research) gives a global average of USD 4.99 million, up 12% on the previous year and driven by higher detection, escalation, and lost business costs.

Why do SIEM projects fail?

SIEM projects fail for seven recurring reasons, and each one has a prevention step. The reasons follow the project phases.

  • Collecting everything first, because it raises cost and noise without a purpose; prevent it by starting from use cases.
  • Starting the legal track late, because the Betriebsrat and the Datenschutzbeauftragte then block the go-live; prevent it by starting both in week one.
  • Having no named owner or run capacity, because alerts then stay unread; prevent it by naming an owner and a staffed operating model before purchase.
  • Excluding sources because of licensing, because the use cases then lose their data; prevent it by choosing the scope first and the licence model second.
  • Never parsing sources, because unparsed data cannot be correlated; prevent it with a parse success check per source.
  • Having no tuning budget, because untuned rules cause alert fatigue; prevent it by planning tuning time of several weeks after the first alerts.
  • Never testing detection, because untested rules give false confidence; prevent it with a purple team exercise (attackers and defenders working together) or an attack simulation.

Frequently asked questions

A SIEM (Security Information and Event Management) system collects log data from servers, network devices, and applications, correlates the events, and raises alerts on suspected attacks. It also stores the data for investigations and as evidence for auditors and regulators.

A SIEM collects and correlates data from many different sources across the whole company. XDR (extended detection and response) focuses on endpoint, network and cloud telemetry, usually from the tools of one vendor. The two approaches can complement each other.

No law names a SIEM, but the duty to detect and handle incidents applies to entities in NIS2 scope. Section 30 BSIG requires them to document incident handling measures. Section 31 Abs. 2 BSIG requires KRITIS operators to deploy attack detection systems, and a SIEM with an operating process can fulfil this duty.

Yes, in companies with a Betriebsrat. Section 87 Abs. 1 Nr. 6 BetrVG gives the Betriebsrat a right of co-determination over technical devices that are suitable to monitor employee behaviour or performance. Without agreement, the Einigungsstelle decides.

Yes, through a managed or hybrid operating model. A provider then analyses the alerts. The company still needs a named internal owner who decides on rules, scope, and response actions.

To implement a SIEM in a German company, follow the step-by-step order of this guide: start from risks and duties, run the legal track with the Betriebsrat and the Datenschutzbeauftragte in parallel, onboard log sources in waves, and then parse, correlate, and tune. Measure the result with agreed KPIs and name an owner for steady state. The Bitkom data shows what is at stake: 59% of attacked companies name insufficient detection as a contributing factor. A SIEM implementation that follows these steps closes that gap in a way that auditors and the works council can verify.

Table of contents


Please contact us if you have any questions.

Learn more about SecureVisio and the benefits it offers.
Poland
Poland
+48 17 779 6246
Germany
Germany
+49 4186‑895991‑0

Fill out the form to contact us