# SAP security inside the system or beside it - which architecture holds up

> SAP security as an add-on inside SAP or an appliance beside it? Visibility, stopping an attack, stored credentials, the SIEM and five vendor questions.

- Source: https://snok.ai/en/news/blog/sap-security-inside-or-beside-the-system/
- Author: Jacek Bugajski
- Published: 2026-11-17

---
## Key takeaways

- An SAP security add-on sees the Security Audit Log, ABAP runtime events, transports and RFC calls as they happen; an external tool collects the same data afterwards, through interfaces and a privileged account.
- A tool inside SAP can stop an operation mid-flight: demand a second authentication factor, lock the account or end the session; an external tool, SecurityBridge argues, can only observe and alert.
- Once an SAP system is compromised, any console reading from it shows the attacker's version of events; the defence is evidence that reached the SIEM before the attacker could change it.
- An external appliance brings its own operating system and database, a single store of credentials for every SAP system, and new network paths into the landscape.
- Without a live event stream to the SIEM, the native architecture loses its strongest argument, so SOC integration is a precondition rather than an extra.

On 10 November SAP released 42 items on SAP Security Patch Day, nine of them rated critical. A day later the SecurityBridge research team reported that in a single month it had sent SAP more than 40 previously unknown vulnerabilities, roughly twenty times its usual rate, after handing the reading of ABAP code to an AI model and keeping people on validation. Those fixes have yet to ship. At that pace, where you place the tool meant to notice an attack on SAP is no longer an argument for architects alone.

On 12 November Ivan Mans, co-founder of SecurityBridge, set out why SAP security belongs inside the system. He was answering vendors of tools that sit beside SAP, who argue that code written in ABAP shares the fate of the system it protects. Both sides make reasonable points, so I have put them to the test: each argument checked against the facts, with a clear view of where the native approach wins and what that depends on.

## Two models for protecting SAP

There are two models. In the first, the security tool is installed in SAP as an SAP-certified add-on, through routine Basis work, and is governed by the same controls as everything else: the authorisation concept, the Security Audit Log, change documents and transports. In the second, the tool lives on a separate virtual machine or appliance, connects to each monitored system and pulls data out of it.

The external camp makes three serious points. A tool written in ABAP is exposed to the same vulnerabilities as the system, so an attacker with an exploit can blind it. A system cannot objectively audit itself, and an administrator with full authorisations can switch the tool off or wipe the logs. And any code running in production competes with business processes for resources, so a faulty update can destabilise the system.

Mans answers with the principle he puts at the top of his piece: security tools belong where the visibility is. Each of the objections can be tested against the evidence.

## What the EDR market shows

Microsoft Defender runs inside Windows. CrowdStrike Falcon, SentinelOne and Defender for Endpoint run on the host they protect, many of them in the kernel. All of them are exposed to the vulnerabilities of that host. The market chose this model anyway, because processes, memory, users and files are only visible from the host.

The strongest counter-example comes from the same market. On 19 July 2024 a faulty CrowdStrike content update affected 8.5 million Windows devices, by Microsoft's count. Nobody moved EDR off the endpoint afterwards. CrowdStrike introduced staged rollouts and let customers choose where and when updates land. Deployment risk was answered with deployment discipline, and the visibility stayed. The same question belongs in any conversation with an SAP add-on vendor: how are updates tested, and who decides when they reach production?

## Visibility: what a tool inside SAP sees that one beside it does not

Mans starts with the layers. Firewalls, VPNs, EDR and cloud security cover the infrastructure. None of them sees the application layer, where SAP transactions, roles, custom code and business processes live. A tool inside the SAP stack sees the Security Audit Log, ABAP runtime events, transports, RFC calls and user context as they occur. A tool outside sees them later, partially, and only through the interfaces SAP exposes.

This matters for the SIEM. The add-on evaluates an event at source, with user and process context, and can filter and enrich it before anything leaves the system. The SOC receives alerts it can act on instead of raw log volume. We covered this in more detail in [SecurityBridge as a SIEM in ABAP](https://snok.ai/en/news/blog/securitybridge-siem-in-abap/).

## Responding to an attack inside SAP

An alert tells you something has happened, but it will not stop the operation itself. In Mans's account, an external tool can observe and raise an alert, while a tool inside SAP can act during the operation, on rules the customer defines:

- require a second authentication factor before a sensitive action, such as exporting a large payment file, not just at logon,
- lock an account that behaves suspiciously,
- terminate a session in which an attack is under way.

The first matters most. Attackers usually log in with genuine, stolen credentials. If they cannot re-authenticate before the action, the action never completes and there is no incident to handle.

This is where the architecture question meets November's SAP Security Patch Day. By its own count, SecurityBridge arrives at 47 notes for November and calls it the highest figure in more than eight years, while conceding that one month does not make a trend. Joris van de Vis, who leads its research, points to the other edge of the same technique: a model that compares old and new code also shortens the path from a published fix to a working exploit. That is a vendor's assessment, though a hard one to dispute: the gap between a patch and the first exploitation attempt will keep narrowing.

In-system response has its limits, though. A second factor stops an attacker holding stolen credentials, but not a flaw that can be exploited without logging on. In that case an add-on can at most spot and interrupt what the attacker does next: creating a user, granting authorisations, running a transaction. The flaw itself still needs the patch, and by SecurityBridge's count more than 20 of November's notes require a kernel fix, which means a maintenance window and a restart.

## Where an external tool gets its data

The isolation argument is persuasive until you ask where an external tool gets its data. It logs on to every monitored system as a privileged technical user, calls RFC and API functions, and reads tables, logs and configuration. Every finding it reports comes from the system it claims to be isolated from. In Mans's words, "agentless" describes the packaging, not the privileges.

If an attacker takes over an SAP system, they can tamper with the functions an external tool relies on, and the isolated console will faithfully display their version of events. What decides the outcome is how quickly trustworthy evidence leaves the system. An add-on streams events to the SIEM the moment they occur, and an attacker cannot recall what has already arrived there. According to Mans, an attempt by a fully authorised administrator to disable the add-on generates its own events - authorisation failures, Security Audit Log entries, change documents - which reach the SIEM before the attempt is complete. That is a strong answer, but it covers events that have already left. If an attacker takes over an application server at kernel level, events still waiting in a queue may never leave.

## What an external appliance adds to the attack surface

Mans tests the charge that an add-on widens the attack surface against the alternative. In his view, an external appliance brings three things into an SAP landscape:

- a foreign operating system and database: a Linux image, container runtime, web console and vendor services, each with its own CVE stream and patch schedule, usually outside your vulnerability management,
- a single store of credentials for every SAP system: whoever compromises the appliance inherits access to the whole landscape,
- new network paths: firewall rules from the appliance to every application server, message server and gateway, and often outbound connections to the vendor.

Inside the protected systems, an add-on introduces none of these. The central system that collects data still has to connect to the monitored systems, but it is an SAP system: it runs under the SAP authorisation concept, leaves a trail in change documents and transports, and Basis patches the whole set through the standard process, in the same maintenance window as the rest of SAP. A team that knows the ABAP stack can harden it, audit it and support it at two in the morning. Isolation remains available: where segregation of duties requires it, the native controller can run on a separate, hardened SAP system in its own network segment, with its own administrators.

## Where the native approach has limits

An add-on is production software, so update discipline and performance testing have to be proven in your own landscape. The central system that collects data from the monitored systems is itself an SAP system and needs hardening like any other.

The most important limit concerns the SIEM. The entire defence against the scenario in which an attacker takes over SAP and disables the tool rests on events leaving the system as they happen. If the SIEM integration is not configured, nobody watches it, or events sit in a queue, the native case weakens at exactly the point the other side attacks. That is why I treat SOC integration as a condition of any deployment.

## Five questions for any SAP security vendor

Mans closes with five questions and says they should be put to every vendor, SecurityBridge included. I would endorse all of them:

1. Where exactly does your detection data come from, and what happens to it if the SAP system is compromised?
2. Which credentials does your platform store, for how many systems, and with which authorisations?
3. Which operating system, runtime and database does your platform run on, and who patches them?
4. Which new network connections does it need, inbound and outbound?
5. Can your existing team operate, audit and troubleshoot it without the vendor's support desk?

I would add a sixth: how many seconds does it take for an SAP event to reach your SIEM, and who is alerted when the stream stops?

## How SNOK deploys SecurityBridge

We have been deploying SecurityBridge since December 2023. In one deployment, confirmed by the customer's reference letter, the platform protects five SAP systems: SAP ERP ECC, SAP ERP HCM, SAP BW on HANA, SAP Solution Manager and SAP PO, and we ran the installation together with workshops for the customer's team. During deployment we agree with the customer how the SIEM integration works and who responds to an SAP alert, and how fast. Alongside that, [SAP penetration testing](https://snok.ai/en/offer/sap-security/sap-penetration-testing/) shows whether the detection rules catch real attack paths.

If you are comparing SAP security platforms, start with the five questions above. Our [SecurityBridge offering](https://snok.ai/en/offer/sap-security/securitybridge/) describes how a deployment runs, and our piece on the [RISE with SAP Security Score](https://snok.ai/en/news/blog/rise-with-sap-security-score-what-it-measures/) covers what a compliance score alone cannot see.

## Frequently asked questions

### Is it safe to run a security tool inside SAP?

Yes, provided updates are tested and rolled out in stages. EDR runs on the host it protects on the same principle. The add-on is subject to the SAP authorisation concept and transports, and streams events to the SIEM as they occur.

### Can an attacker who has compromised SAP hide from an add-on?

They cannot recall or alter events that have already reached the SIEM. The risk lies with events that have not yet left the system, so the stream must run without delay and someone must notice when it stops.

### Does an "agentless" external platform leave no footprint in SAP?

It does leave one. It connects to every monitored system with a privileged technical user over RFC or APIs and usually needs configuration in each system. It also keeps credentials for all of those systems in one place.

### Does a native add-on need extra servers?

It needs no new operating system, database or container platform. Data is collected by a central system that is itself an SAP system; where segregation of duties requires it, we place it in a separate, hardened segment.

### Does SecurityBridge work with RISE with SAP?

SecurityBridge holds an SAP-certified integration with SAP S/4HANA Cloud Private Edition, the environment on which RISE with SAP is built. The add-on installs through the standard SAP process, with no new network paths to negotiate through SAP's managed network.

## Sources

- Ivan Mans, SecurityBridge, "Why Native Is the Winning Architecture for SAP Security", 12.11.2026 - [SecurityBridge blog](https://securitybridge.com/blog/native-sap-security-architecture/) (accessed 16.11.2026)
- Joris van de Vis, SecurityBridge, "The flood wave is here: How AI is changing SAP vulnerability research, and your patch backlog", 11.11.2026 - [SecurityBridge blog](https://securitybridge.com/blog/the-flood-wave-is-here-ai-and-saps-patch-backlog/) (accessed 16.11.2026)
- SAP Support Portal, SAP Security Notes, November 2026 - [SAP page](https://support.sap.com/en/my-support/knowledge-base/security-notes-news/november-2026.html) (accessed 16.11.2026)
- Microsoft, "Helping our customers through the CrowdStrike outage", 20.07.2024 - [Microsoft blog](https://blogs.microsoft.com/blog/2024/07/20/helping-our-customers-through-the-crowdstrike-outage/) (accessed 16.11.2026)
- CrowdStrike, "Channel File 291 Incident Root Cause Analysis", 06.08.2024 - [CrowdStrike report (PDF)](https://www.crowdstrike.com/wp-content/uploads/2024/08/Channel-File-291-Incident-Root-Cause-Analysis-08.06.2024.pdf) (accessed 16.11.2026)
- SAPinsider, SecurityBridge certification for SAP S/4HANA Cloud Private Edition, 06.08.2026 - [article](https://sapinsider.org/articles/securitybridge-sap-s4hana-cloud-private-edition-certification/) (accessed 16.11.2026)
