Manufacturing & Machine Building · NIS2 & IEC 62443

Sometimes all it takes is a USB stick on a desk

If you manufacture machinery, electronics, electrical equipment, vehicles or medical devices, you fall under NIS2 as an important entity from 50 employees or more than 10 million EUR in both annual revenue and balance sheet total. In practice, the technical benchmark for that is usually IEC 62443, backed up by the BSI's ICS Security Compendium. We test exactly where IT meets production: PLCs and SCADA systems, remote-maintenance access, and the factory floor itself, without putting your production at risk in the process.

The question every plant manager asks first

We test without stopping the line. Active testing happens only inside windows you agreed to beforehand, and on OT-adjacent systems we read along instead of writing.

  • Testing methodology aligned with IEC 62443
  • Experience with OT/ICS & production environments
  • Free initial consultation
Plant section Example scenario · Break room · 06:52 · early shift
  1. USB stick, phishing
  2. 01Office IT
  3. IT/OT fault line
  4. 02Production controlRemote maintenance
  5. 03Control layer
  6. 04Machines & field level

Four layers, two ways in. The second one bypasses everything the upper layer protects.

How common real incidents are

24%

Roughly one in four companies in mechanical and plant engineering reported a significant cybersecurity incident within the past two years.

That makes an incident an operating risk like a machine failure rather than an exception. Most plants still treat the two differently: the machine has a maintenance plan, the network behind it rarely has a test plan.

Source: VDMA study Industrial Security und Produktpiraterie 2024, Fraunhofer AISEC, 100 surveyed VDMA member companies

The core view

The plant section: four layers and two ways in

A plant can be cut open like a building. Office IT on top, with workstations, goods receiving, and the break room. Below it production control, with control systems and remote-maintenance access. Below that the control layer, with PLCs and fieldbuses. At the bottom, the machines themselves. Between the first and the second layer sits the point everything turns on.

USB stick, phishing 01 Office IT IT/OT fault line 02 Production control Remote maintenance 03 Control layer 04 Machines & field level
Schematic drawing. Which layers you actually have and how cleanly they are separated is settled in the scoping workshop, before the first test step.

Two ways in

USB stick or phishing mail

This path arrives at the top, in office IT, and works its way down from there. On the way it runs into everything you have built: endpoint control, permissions, segmentation. Each of those costs an attacker time and leaves traces behind.

Remote maintenance without a second factor

The uncomfortable one. The machine vendor's access does not arrive at the top, it arrives from the side and directly below the fault line. It skips every measure on the upper layer, and it is reachable from outside without anyone inside the building clicking on anything.

01

Office IT

Workstations, goods receiving, break room, Active Directory

A prepared USB stick in the break room, a phishing mail to shift planning, a door held open at goods receiving. Whoever lands here is standing in a network that, in many plants, is still connected to everything else.

Media dropping and phishing tailored to shift work and third-party contractors, plus entry attempts and the classic office IT inside the plant including Active Directory. On this layer we test actively.

See social engineering
IT/OT fault line The fault line
02

Production control

Control systems, SCADA, remote-maintenance access

Machine builders and system integrators need remote access for service and maintenance. If that access sits on the network without multi-factor authentication or with overly broad permissions, it leads past everything secured above it.

Remote-maintenance access, control systems, and their permissions, scheduled around your production windows. Active steps run only inside time windows agreed beforehand.

See OT/ICS pentest
03

Control layer

PLCs, fieldbuses, engineering stations

PLCs, control systems, and fieldbuses are often built for uptime rather than security, and can run for years without a patch. Factory-default credentials on an industrial device are no rarity here.

We read along, we do not write. Reachability, firmware levels, and credentials are collected passively. No write access to a running controller takes place.

See infrastructure pentest
04

Machines & field level

Machines, sensors, industrial IoT

Sensors and networked machines widen the attack surface, frequently still on factory-default settings nobody ever changed. Whoever gets here is at the point where a mistake affects material and people.

We check field devices and industrial IoT for reachability, default credentials, and exposed services, as a rule passively and never without plant management at the table.

See custom pentest

The fault line

How far an incident travels is decided between office IT and production

Exactly where classic office IT touches the production layer, the network segmentation that would otherwise stop the spread is often missing. That single property decides whether a compromised office workstation becomes an annoyance or a production standstill.

Without segmentation

The path goes through

Without a separation between office IT and the production network, PLCs, control systems, and the machine vendor's remote-maintenance access are all reachable from the compromised workstation. The way down is open, and nobody notices as long as nobody looks.

With segmentation

The path ends here

Once the boundary is drawn and enforced, the incident stays in office IT. The damage is then an IT problem with a recovery effort, not a manufacturing problem with idle lines and delivery dates at risk.

How it reads in the report

No network segmentation between office IT and the production network

How we test it

We take a foothold in office IT and try to move from there toward the production network. Whatever turns out to be reachable gets documented rather than exploited. The result is a list of the crossings that stand open, with an assessment of which one should be closed first.

A scenario from real assessments

The story in five steps

Media dropping is an established social-engineering technique we run in a controlled form at Access Granted on a regular basis. Not a specific incident at a specific customer, just a typical sequence that shows how little it can take. Every step belongs to one of the four layers of the plant section.

  1. 01

    Layer 01 · office IT

    The stickBreak room

    A prepared USB stick turns up, apparently by chance, in the break room or at goods receiving

    Placed unobtrusively, often carrying a company logo or a label designed to spark curiosity.

  2. 02

    Layer 01 · office IT

    The clickOffice workstation

    An employee plugs it into a company computer, out of curiosity

    Out of curiosity, a wish to be helpful (“maybe it belongs to someone”), or plain habit, with no bad intent at all.

  3. 03

    Layer 01 · office IT

    The first footholdOffice network

    Without network segmentation, that's enough for a first foothold in your internal network

    From there, the real assessment checks how far that access could be pushed toward the production network.

  4. 04

    Fault line · layer 02 and below

    From office IT to the shop floorProduction network

    Where IT and OT aren't cleanly separated, the path leads straight on to the controllers

    Without segmentation between office IT and the production network, PLCs, control systems, and the machine vendor's remote-maintenance access are all reachable from the compromised workstation. We document that path rather than walking it to the end: on OT-adjacent systems we work passively, and actively only in pre-agreed time windows.

  5. 05

    All four layers

    The report on the tableMeeting room · a few weeks later

    What ends up on the table isn't an incident, it's a report

    Every finding rated by severity, exploitability, and potential impact on production, with a management summary for executive, IT, and plant leadership and a mapping to NIS2 and IEC 62443. This is what an excerpt can look like:

We measure how high the pickup rate actually is at your company, and whether missing segmentation turns that into more than a footnote.

What ends up on the table

The report assigns every finding to the layer it was found on and names the matching measure. Executive management reads the summary, IT leadership reads the technical part, plant management reads the part that affects the lines.

Manufacturing pentest report · Sample excerpt Sample
Severity Finding Remediation area
CriticalUSB-drop test: 3 out of 10 planted sticks were plugged in (example figure)Awareness & endpoint control
CriticalNo network segmentation between office IT and the production networkIT/OT segmentation
HighRemote-maintenance VPN for a machine vendor reachable without multi-factor authenticationRemote-maintenance access
HighPLC firmware left unpatched for years across multiple linesPatch & firmware management
MediumTailgating: entry into the production hall possible without a badge check (example figure)Physical access control
LowIndustrial device reachable with factory-default credentialsCredential hardening
6 findings · remediation plan included Mapped to NIS2 & IEC 62443

Illustrative example – no real customer data.

Are you in scope

When NIS2 applies to your plant, and what that means on each layer

Six manufacturing branches are listed in Annex 2 of the BSIG, including machinery, electronics and vehicles. Anyone active there who reaches the size thresholds counts as an important entity. The order in which that is checked is set out in § 28 BSIG, and it asks about the activity first and the size second.

Size thresholds

Employees

50 and above

The company is counted, not the individual site.

or

Annual revenue and balance sheet

10 million EUR, each more than

Revenue and balance sheet total must both be exceeded; alternatively, headcount alone is enough.

Result

Important entity

Machine building, electronics, automotive parts, and medical devices are all listed entity types within manufacturing. The German NIS2 implementation act has been binding law since 6 December 2025. Not sure whether you cross the threshold? The free scope check walks you through the assessment in order.

What applies then, layer by layer

  1. Layer 01 · office IT

    Risk management and access control

    The requirement catalog under Article 21(2) asks for access control, personnel security, and training, among other things. A pentest on this layer demonstrates exactly that technically instead of asserting it.

  2. Fault line

    Network security and containment

    Segmentation between office IT and production is the point where risk management becomes measurable. Evidence for it comes from an attempt to defeat it, not from a policy document.

  3. Layers 02 to 04

    State of the art in an OT context

    IEC 62443 is the practically recognised benchmark for what counts as appropriate in automation and control systems. Our reports map findings to it accordingly.

  4. All layers

    Reporting duty for significant incidents

    § 32 BSIG requires staged reporting: an early warning within 24 hours and an incident report within 72 hours. Knowing in advance which layer is reachable how makes it far easier to say what happened when it counts.

Access Granted delivers the penetration test and the technical evidence. Registration with the BSI, the legal classification, and advice on either are not part of that and stay with you or your legal counsel.

Ground floor

People and premises: the path that bypasses technology entirely

The plant section does not start at a screen, it starts at the fence. Four stations on the site decide how easily somebody gets anywhere near layer 01 in the first place.

  1. Station 01

    Goods receiving

    Parcel deliveries, suppliers, truck drivers: unfamiliar people and unfamiliar objects come through here every day, often without anyone asking who actually belongs.

  2. Station 02

    Third-party contractors

    Service technicians, cleaning staff, and temp workers move through the plant with their own laptops or their own badges. Their access and devices rarely go through the same controls as your own.

  3. Station 03

    Break room

    Where colleagues are relaxed and happy to help, an orphaned USB stick or an offhand “I'm new here” is at its most convincing.

  4. Station 04

    Tailgating

    A door held open, a smokers' exit, a badge check that only happens in the morning: the way into the production hall often bypasses technology entirely.

Physical testing runs on a written engagement with a named contact on file who can vouch for our people if it comes to that. Nobody on your team ends up in a situation they cannot resolve.

Our approach

How we test without stopping the line

Keeping production available matters more than test depth. Before we touch anything, we agree with IT and OT stakeholders which systems get actively tested, which are only analyzed passively, and which are excluded entirely. On PLCs, control systems, and other OT-adjacent systems we work passively as a rule: we read along, we don't write.

Where active testing is needed, it runs in pre-agreed production windows, with a defined emergency contact on both sides and a firm rule to stop immediately at any unexpected effect. Testing machines and lines never happens without plant management at the table, so no test step accidentally halts production.

A working day in shift operation
Production running Agreed test window

You set the window, not us. With round-the-clock shift operation, active steps move to planned downtime, maintenance windows, or a test cell away from the line.

Three release levels, settled in advance

Active
Office IT, physical entry, Active Directory, named remote-maintenance access. Here we work with the same means an attacker would.
Passive
PLCs, fieldbuses, engineering stations. We collect reachability and version levels without changing a state.
Excluded
Everything you rule out. Excluded systems appear in the report as a deliberate gap, so nobody later mistakes them for tested.

Test plan

The test sequence, ordered along the same layers

The same structure that carries the plant section also decides what gets tested when. Going down, the approach gets more careful, because the possible consequences get larger.

Layer Module Mode Limit
Ground floor · premises Physical pentest Active Written engagement, named contact on file, no confrontation.
Layer 01 · office IT Social engineering & infrastructure Active No content from private mailboxes, no pressure on individuals.
Fault line · IT/OT Network segmentation Active Reachability is demonstrated, the crossing is not walked to the end.
Layer 02 · production control OT/ICS pentest Active in the window Only inside the agreed production window, emergency contact reachable.
Layers 03 & 04 · control and machines OT/ICS pentest Passive No write access to running controllers, no restarts, no firmware changes.
  • Active: we act as an attacker would, within the released scope.
  • Passive: we observe and document without intervening.

The assignment is the starting point, not the outcome. After the scoping workshop, the report states for every system which mode it was tested in.

Pentesting for manufacturing & production: your questions

Probably, once you have at least 50 employees or both annual revenue and balance sheet total exceed 10 million EUR. Manufacturing, including machine building, electronics, automotive parts, and medical devices, falls under the NIS2 sectors and is then classified as an important entity. Not sure whether you cross that threshold? We'll work it out with you in a free initial consultation.

In OT, availability comes before everything else: a crashed controller can shut down a production line for hours. So we test more carefully, often passively rather than actively, and with a narrower scope. A clearly defined emergency contact is in place before any active testing ever touches a PLC.

Yes, and for OT-adjacent systems that's actually the norm. We agree in advance on which systems get actively tested, which are only analyzed passively, and which are excluded entirely. A defined emergency contact and an immediate stop at the first sign of trouble are standard parts of the plan.

IEC 62443 is the relevant standards series for securing industrial automation and control systems. There's usually no direct legal mandate to follow it, but it's the practically recognized benchmark for what “state of the art” means in an OT context, which makes it relevant for your NIS2 evidence.

It depends heavily on scope: number of sites and production lines, whether OT/ICS systems are included, and whether physical and social-engineering testing are part of it. After a free initial consultation, we'll know your scope and you'll get a transparent fixed-price quote.

Media dropping is a documented, actively used attack technique, not a theoretical exercise. Ransomware at manufacturing companies typically gets in through the IT side and then spreads toward OT-adjacent systems. A dropped USB stick is one of the simplest ways to simulate exactly that first step.

Typically IT management, plant or OT leadership, and, depending on scope, executive management. The coordination between IT and OT is often the trickiest part, since responsibilities in practice frequently aren't cleanly split between the two. We bring the relevant stakeholders to the table from day one.

Yes. Remote-maintenance access for machine vendors and system integrators is one of the most common weaknesses we find in practice, often without multi-factor authentication or with overly broad permissions. We test that access specifically as part of the OT/ICS and infrastructure pentest.

We listen on the network, evaluate the responses to ordinary queries, and compare firmware levels and credentials against known weaknesses. We send no control commands, restart nothing, and install nothing. The difference from active testing is not how thoroughly we evaluate what we see, but whether we change the state of a running system.

One contact from IT and one from production, a rough overview of sites and lines, the information on which remote-maintenance access exists, and a decision on which systems stay excluded. Network diagrams help but are not a prerequisite: if they are missing or out of date, that is already a finding in itself.

One vulnerability is enough to stop a production line for hours.

Talk to us before someone else finds the gap first. A short call settles scope and NIS2 relevance, then you get a transparent fixed-price quote.