Two operators at a control-room desk with a row of monitors, a laptop and keyboards

OT/ICS security

DNP3 protocol security: what Secure Authentication covers, and what it doesn't

Last updated:

DNP3 (Distributed Network Protocol) runs most of the telecontrol links between control centres, substations and field equipment across North America's electric, water and gas utilities, and its base specification — standardized by IEEE as IEEE Std 1815, most recently revised as IEEE 1815-2012 — carries no authentication or encryption by default, the same structural gap Modbus has. The 2012 revision added an optional Secure Authentication layer, but a decade of CISA ICS advisories against DNP3 implementations shows that most documented failures are implementation bugs that run underneath whatever authentication layer is, or isn't, turned on.

Standards status comes from the IEEE SA pages for 1815-2012 and P1815 and the DNP Users Group's January 2026 documents; advisory facts from CISA's advisories and machine-readable CSAF files; all read on 12 November 2026. Where a statement rests on commentary rather than a primary document, the text says so.

Key takeaways

  • Standards status comes from the IEEE SA pages for 1815-2012 and P1815 and the DNP Users Group's January 2026 documents; advisory facts from CISA's advisories and machine-readable CSAF files; all read on 12 November 2026.
  • On status, IEEE lists 1815-2012 as an "Inactive-Reserved Standard" since 30 March 2023.
  • A DNP3 master polls or commands an outstation, usually over TCP port 20000, which IANA registers as dnp (port 19999 is dnp-sec, "Distributed Network Protocol - Secure").
  • The IOServer and GE Proficy findings trace to Adam Crain (Automatak) and Chris Sistrunk's 2013 fuzz-testing research.

What did IEEE 1815-2012 actually add, and is it still current?

IEEE 1815-2012 ("IEEE Standard for Electric Power Systems Communications — Distributed Network Protocol (DNP3)", approved 8 June 2012, published 10 October 2012) specifies "the DNP3 protocol structure, functions, and interoperable application options (subset levels)", from a simplest level "intended for low-cost distribution feeder devices" to the most complex "for full-featured systems". None of that core structure includes a field for a credential, session key or message signature; any device that can reach a DNP3 master or outstation can issue a function code it accepts. What the 2012 revision changed was adding Secure Authentication Version 5 (SAv5) as an optional feature, based on IEC/TS 62351-5 — an improvement over the 2010 edition's Secure Authentication, which supported pre-shared keys only. Both versions are opt-in: a master and outstation must both implement and explicitly negotiate Secure Authentication for it to apply.

On status, IEEE lists 1815-2012 as an "Inactive-Reserved Standard" since 30 March 2023. Its revision, project P1815, has been an "Active PAR" since 11 June 2015, and as of 12 November 2026 no newer edition is published. In practice, 1815-2012 is still the edition that devices, conformance tests and procurement specifications cite.

What does a DNP3 exchange look like on the wire?

A DNP3 master polls or commands an outstation, usually over TCP port 20000, which IANA registers as dnp (port 19999 is dnp-sec, "Distributed Network Protocol - Secure"). Each application-layer message carries a function code. Zeek's DNP3 analyzer names them in consts.zeek; these are the ones a defender watches:

Code Zeek name What it does Why it matters
0x01 READ Polls data objects Normal master traffic
0x02 WRITE Writes data objects, such as the outstation clock Changes outstation state
0x03, 0x04 SELECT, OPERATE Two-step control of a breaker, pump or setpoint Changes the process
0x05, 0x06 DIRECT_OPERATE, DIRECT_OPERATE_NR One-step control, the second with no response Changes the process with no select step
0x0D, 0x0E COLD_RESTART, WARM_RESTART Restarts the outstation Loss of view and control during restart
0x12 STOP_APPL Stops an application on the outstation Stops a function the site relies on
0x15 DISABLE_UNSOLICITED Stops report-by-exception Blinds the master to events
0x20, 0x21 AUTHENTICATE_REQ, AUTHENTICATE_REQ_NR Secure Authentication requests Shows SA messages on the link
0x83 AUTHENTICATE_RESP Secure Authentication response Shows the outstation answering SA challenges

Without Secure Authentication, nothing in an OPERATE or COLD_RESTART request proves who sent it. With SA, the outstation challenges critical requests, and the SA function codes appear in the traffic.

Where does DNP3 security actually fail in deployed stacks?

Advisory Product CVE(s) CVSS What it actually allowed
ICSA-13-161-01 (2013, rev. 2018) IOServer DNP3 driver (v1.0.19.0) CVE-2013-2783 7.1 (v2) Specially crafted TCP packet on port 20000 put the outstation into an infinite loop, requiring manual restart
ICSA-13-297-02 (2013) GE Proficy DNP3 I/O Driver (v7.20j and prior), used in iFIX and CIMPLICITY; ICSA-14-287-01 later listed 7.20k as still affected by the TCP flaw, fixed in 7.20L, 8.2.62.7 and 9.0.62.7 CVE-2013-2811, CVE-2013-2823 7.1 / 4.7 (v2) Crafted TCP packet (IP-based) or serial frame caused an infinite loop and process crash
ICSA-20-105-02 (2020, rev. 2022) Triangle MicroWorks DNP3 Outstation Libraries (v3.16.00–3.25.01) CVE-2020-6996 7.5 (v3, CISA); 9.8 (v3.1, NVD) Specially crafted message caused a stack-based buffer overflow; no authentication required

The IOServer and GE Proficy findings trace to Adam Crain (Automatak) and Chris Sistrunk's 2013 fuzz-testing research. Crain's Project Robus page says "During 2013, Chris Sistrunk and I reported a large number of defects in implementations of the DNP3 protocol", and its table lists 32 CISA (then ICS-CERT) advisories, from ICSA-13-161-01 to ICSA-16-299-01. That is a well-documented case of an entire protocol ecosystem sharing the same class of parsing weakness, not one vendor's isolated bug. The 2020 Triangle MicroWorks finding shows the pattern recurring seven years later in a widely licensed outstation library. In every case, the vulnerability sits in message-parsing code that executes regardless of whether Secure Authentication is negotiated on that link — a correctly configured SAv5 session narrows who can reach that code, it does not patch the code itself.

Newer advisories move the problem sideways rather than away. In ICSA-22-167-05 (June 2022), Siemens' EN100 Ethernet module "DNP3 IP variant" had CVE-2022-30937 (CVSS 3.1 8.6 from Siemens, 7.5 from NVD), a memory corruption in parsing HTTP requests to its /txtrace endpoint; the bug sits in the module's web interface, next to the DNP3 stack. For a device maker, the attack surface is the whole communication module: web server, TLS library and DNP3 stack together, which is what a firmware SBOM is for.

What happened to Secure Authentication after SAv5?

The work moved from IEEE to the IEC and back into DNP3, and the name changed along the way:

Version Where it is defined What changed
Secure Authentication in IEEE 1815-2010 IEEE 1815-2010 First DNP3 SA; pre-shared keys only
SAv5 IEEE 1815-2012 The SA version in the current published IEEE edition
IEC 62351-5:2023 IEC TC 57 International Standard replacing IEC TS 62351-5 (2013)
DNP3-SL, formerly SAv6 DNP Users Group, for IEEE P1815 "an implementation of the new IEC 62351-5 International Standard"

The IEC 62351-5:2023 foreword lists the technical changes. Security now works per controlling/controlled station association. User management and the symmetric update-key change were removed, and so were the challenge/reply procedure and concepts. Aggressive Mode became a Secure Data message exchange, and "Authenticated encryption of application data was added." That last item matters: SAv5 authenticates messages but does not encrypt them.

The DNP Users Group's DNP3-SL sheet (January 2026) says DNP3-SL "Authenticates all messages and eliminates previous inefficient challenge-response" and "Can co-exist with DNP3-SAv5". Its January 2026 membership update says the "Draft specification is complete and will be released with IEEE P1815~2026." Until that edition is published, SAv5 remains the only Secure Authentication version in a published IEEE DNP3 standard. Not everyone wants the new layer either: Crain's own company argues in a 2026 post for protecting the IP path with TLS and segmenting instead.

Why does secure DNP3 stay switched off, according to CISA?

CISA's 10 February 2026 report "Barriers to Secure OT Communication: Why Johnny Can't Authenticate" names DNP3 to DNP3-SAv5 as its first example of secure protocol versions that "have existed since the early 2000s". The DNP Users Group is one of its listed research partners. Its interviews covered water and wastewater, transportation, chemical, energy, and food and agriculture operators. The findings that matter for DNP3:

  • Cost and availability come first. "The two main barriers were cost and the availability concerns stemming from encrypting data." Availability fears break down into observability, latency and bandwidth, and confidence.
  • PKI is the sticking point. "Almost every operator interviewed mentioned challenges with deploying and maintaining PKI." CISA notes that pre-shared keys, an alternative, were never mentioned by interviewees.
  • Newer DNP3 versions help. CISA writes that newer protocol versions (it names DNP SAv2, DNP SAv5 and DNP3-SL) favor "increasing automation that lessens the complexity for the end user."

Its recommendations are concrete. Owners should use signing on all OT communications and, if worried about downtime, "consider a phased approach beginning with signing only and not dropping packets". Manufacturers should "Default to the secure option for all configurations" and build in cryptographic agility. Because SAv5 signs without encrypting, DNP3 already matches the signing-first approach, and inspection keeps working.

How do you watch DNP3 traffic passively?

Zeek's built-in analyzer writes one dnp3.log line per request/response pair with fc_request, fc_reply and the outstation's internal indications (iin). It does not log object groups. CISA's ICSNPP-DNP3 package adds dnp3_control.log (the control block of SELECT, OPERATE and DIRECT_OPERATE) and dnp3_objects.log (objects in READ and RESPONSE). Authentication requests are logged "Header Only". That is enough for these rules:

  • A new master. Any host other than the known front-end processors sending requests to an outstation on 20000.
  • Control from the wrong place. SELECT, OPERATE or DIRECT_OPERATE from a host that does not normally control that outstation, or outside the expected schedule.
  • Disruptive functions. COLD_RESTART, WARM_RESTART, STOP_APPL, DISABLE_UNSOLICITED or WRITE outside a change window. These map to ATT&CK for ICS techniques such as T0816 Device Restart/Shutdown; see the MITRE ATT&CK for ICS guide.
  • SA that is expected but absent. On a link that should run SAv5, control requests with no AUTHENTICATE_REQ / AUTHENTICATE_RESP exchanges mean SA is off or bypassed.
  • DNP3 crossing a zone boundary, such as port 20000 reaching an outstation from a business network.

What should an asset owner or device maker check first?

  1. Patch the stack firstThe CISA advisories above are vendor-specific implementation bugs; Secure Authentication does not substitute for a current, patched DNP3 driver or outstation library.
  2. Check whether Secure Authentication is actually negotiated on each link, not whether the hardware's datasheet says it supports DNP3-SA. Unenabled support is no different from no support, and the SA function codes in a capture are the evidence.
  3. Segment and filter regardless of the first two answersIsolate DNP3 traffic (commonly TCP port 20000) behind firewalls from business networks, and require VPNs for remote access — CISA's standing recommendation across every advisory cited here.
  4. Map your protocol inventory against [IEC 62351](/guides/iec-62351-power-systems-communications-security/), the broader TC 57 series IEEE 1815-2012 itself draws on for Secure Authentication, before assuming one part covers a mixed DNP3/IEC 61850/IEC 60870-5-104 environment.
  5. For device makers: ship SA enabled where the customer can support keys, and know your stackRecord which DNP3 library and version each product carries (licensed outstation libraries were behind ICSA-20-105-02), fuzz the parser the way Project Robus did, and decide how you will support IEC 62351-5:2023 and DNP3-SL alongside SAv5.
  6. Keep the compensating architecture in placeThe Purdue Model and NIST SP 800-82 remain the layer that has to hold regardless of what any individual DNP3 link is configured to do. NIST SP 800-82r3 itself names DNP3 among the protocols that OT firewalls and intrusion detection systems parse.

Where does SoftWSP fit in a DNP3 review?

DNP3's 2012 revision gave the protocol a real, standards-based answer to its original lack of authentication — but like Modbus Security and OPC UA's built-in security model, that answer only helps on links where it's actually implemented, patched and turned on. SoftWSP's OT Snapshot reads Zeek and ICSNPP logs from a recorded capture, including dnp3.log, and lists which hosts speak DNP3, which of them sit on the outstation side, and which remote-access sessions (SSH, Telnet, RDP, VNC) cross from outside the private address range. It does not decode function codes or Secure Authentication state, and a capture does not show an outstation's model or firmware. Once you have those from a component list, SoftWSP matches them against NVD, CISA KEV, FIRST EPSS and CISA's ICS advisories, and a watchlist can alert on a vendor, product or CVE. AI drafts the cited report in plain language, and an engineer reviews every finding before it is sent.

This guide draws on the IEEE SA listings for IEEE 1815-2012 and P1815, the DNP Users Group's January 2026 documents, the IEC 62351-5:2023 preview, CISA's advisories and its February 2026 Barriers report, Zeek and ICSNPP documentation, the IANA port registry and Adam Crain's Project Robus page. It is not a substitute for the DNP Users Group's own specifications, CISA's own advisories, or an engineer's assessment of a specific deployment's configuration and patch level.

Put this guide to work

Frequently asked questions

Does DNP3 have built-in authentication or encryption?

The base DNP3 protocol does not. DNP3 was standardized by IEEE as IEEE Std 1815, first approved in 2010 and revised as IEEE 1815-2012 ("IEEE Standard for Electric Power Systems Communications — Distributed Network Protocol (DNP3)", approved 8 June 2012, published 10 October 2012). The core protocol — the structure, function codes and interoperable application levels the standard defines, from simple distribution feeder devices up to full-featured master stations — carries no field for a credential or a signature, the same gap Modbus has. What changed in the 2012 edition is that it added an optional add-on layer, Secure Authentication Version 5 (SAv5), where the prior 2010 edition's Secure Authentication only supported pre-shared keys. Secure Authentication is a separate, optional feature a specific master and outstation must both implement and enable — a DNP3 link running without it, which is the common case on older field equipment, has none of this protection regardless of what the standard offers.

What do real, documented DNP3 security failures actually look like?

Mostly implementation bugs in specific vendors' DNP3 driver software, not a cryptographic break of Secure Authentication itself. CISA's advisory ICSA-13-161-01 (first published 2013, last revised 6 September 2018) covers IOServer's DNP3 driver (v1.0.19.0) and assigns CVE-2013-2783 (CVSS v2 7.1, CWE-20 Improper Input Validation) to a finding that the software "does not validate or incorrectly validate input on Port 20000/TCP," letting a specially crafted TCP packet put the outstation into an infinite loop requiring a manual restart — discovered by Adam Crain (Automatak) and Chris Sistrunk, whose 2013 fuzz-testing research against dozens of DNP3 master and outstation implementations produced a wave of similar CISA advisories that year. The same class of finding recurs in CISA's ICSA-13-297-02 (GE Intelligent Platforms' Proficy DNP3 I/O Driver, version 7.20j and prior), which assigns CVE-2013-2811 (IP-based, CVSS v2 7.1) and CVE-2013-2823 (serial-based, CVSS v2 4.7), both CWE-20, to "an attacker [who] could cause the software to go into an infinite loop by sending a specifically crafted TCP packet, causing the process to crash." CISA's 2014 follow-up ICSA-14-287-01 lists driver 7.20k as still affected by CVE-2013-2811 and names fixed drivers 7.20L, 8.2.62.7 and 9.0.62.7. More recently, CISA's ICSA-20-105-02 covers Triangle MicroWorks' DNP3 Outstation Libraries (versions 3.16.00 through 3.25.01) and assigns CVE-2020-6996 (CISA: CVSS v3 7.5, CWE-121 Stack-based Buffer Overflow; NVD's Primary score is 9.8 with CWE-787) to a finding that "a specially crafted message may cause a stack-based buffer overflow" where "authentication is not required to exploit this vulnerability." None of these are failures of Secure Authentication's cryptography — they are parsing bugs in code that, in most deployments, runs with no authentication layer enabled at all to even gate who can send the crafted message.

How widely deployed is DNP3 Secure Authentication in the field?

Industry commentary associated with the DNP Users Group — the vendor-independent body IEEE 1815-2012 itself credits for DNP3's base specifications and that maintains the protocol's ongoing evolution — has described SAv5 adoption as limited to a handful of large projects rather than broad-based utility demand, and the DNP Users Group's own January 2026 membership update says the successor, first called Secure Authentication Version 6 (SAv6) and renamed the DNP3 Security Layer (DNP3-SL), has a complete draft specification to be released with the next edition of IEEE 1815 (project P1815). Whatever the precise adoption number, the same structural problem shows up here as in [Modbus](/guides/modbus-protocol-security-ot-ics/) and [OPC UA](/guides/opc-ua-security-ot-ics/): a secure option existing in the specification is not the same claim as a secure option running on a given link, and NIST SP 800-82 Revision 3 treats the lack of authentication in legacy field protocols like DNP3 as a structural OT risk to manage with network-layer compensating controls, not something to assume a protocol revision has already fixed in deployed equipment.

What should an asset owner or device maker running DNP3 actually check?

Three things, in order of what actually stops an incident. First, patch level: the CISA advisories above are vendor-specific implementation bugs, and a correctly configured Secure Authentication session does not patch a stack-level buffer overflow or input-validation flaw in the message parser underneath it. Second, whether Secure Authentication is actually negotiated on a given link, not just whether the master and outstation "support DNP3-SA" on a datasheet — if it isn't, that link has no authentication, full stop. Third, network segmentation and function/port-level filtering regardless of the answer to the first two: CISA's advisories consistently recommend minimizing network exposure, isolating DNP3 traffic (commonly TCP port 20000) behind firewalls from business networks, and using VPNs for any remote access, because a patched, authenticated DNP3 link is still worth protecting at the network layer. [IEC 62351](/guides/iec-62351-power-systems-communications-security/) is the broader standards series DNP3-family telecontrol security sits inside (IEC/TS 62351-5 is the basis IEEE 1815-2012 itself cites for its Secure Authentication work), and the [Purdue Model](/guides/purdue-model-ot-ics-network-architecture/) and [NIST SP 800-82](/guides/nist-sp-800-82-ot-security/) describe the segmentation layer that has to carry the weight a legacy protocol's own security option often doesn't.

Sources

  1. IEEE 1815-2012: IEEE Standard for Electric Power Systems Communications — Distributed Network Protocol (DNP3) (standards.ieee.org)
  2. CISA ICS Advisory ICSA-13-161-01: IOServer DNP3 Improper Input Validation (last revised 6 September 2018; CVE-2013-2783)
  3. CISA ICS Advisory ICSA-13-297-02: GE Proficy DNP3 I/O Driver (last revised 17 December 2013; CVE-2013-2811, CVE-2013-2823)
  4. CISA ICS Advisory ICSA-14-287-01: GE Proficy HMI/SCADA DNP3 Driver Input Validation (CVE-2013-2811; driver 7.20k affected, fixes 7.20L, 8.2.62.7, 9.0.62.7)
  5. CISA ICS Advisory ICSA-20-105-02: Triangle MicroWorks DNP3 Outstation Libraries (last revised 27 October 2022; CVE-2020-6996)
  6. NVD: CVE-2020-6996 (Primary CVSS v3.1 9.8, CWE-787)
  7. DNP Users Group: Secure Authentication and DNP3 Secure Authentication Version 6 overview materials (dnp.org)
  8. NIST Special Publication 800-82 Revision 3, Guide to Operational Technology (OT) Security (September 2023)
  9. IEEE P1815: IEEE Draft Standard for Electric Power Systems Communications — DNP3 (Active PAR, approved 11 June 2015)
  10. DNP Users Group: January 2026 Membership Update (DNP3-SL status, IEEE P1815)
  11. DNP Users Group: The DNP Security Session Layer (DNP3-SL), formerly Secure Authentication Version 6
  12. IEC 62351-5:2023 preview (foreword and change list), via iTeh Standards
  13. CISA: Barriers to Secure OT Communication: Why Johnny Can't Authenticate (10 February 2026)
  14. Zeek: DNP3 function-code constants (consts.zeek)
  15. cisagov ICSNPP-DNP3 Zeek package (dnp3_control.log, dnp3_objects.log)
  16. Step Function I/O: Project Robus (Adam Crain, 1 April 2014) — table of the resulting CISA advisories
  17. IANA Service Name and Transport Protocol Port Number Registry (dnp 20000, dnp-sec 19999)
  18. Siemens SSA-693555: EN100 Ethernet module /txtrace memory corruption (CVE-2022-30937)
Was this guide helpful?

In this cluster

DNP3 Secure Authentication (SAv5): IEEE 1815's challenge-response layerDNP3 Secure Authentication v5 mandates SHA-256 and AES-256 Key Wrap as its default algorithms, but a 2017 Tamarin-prover formal analysis and practitioner reporting both point to the same bottleneck: key management, not cryptography, is why SAv5 stays rare in the field.EtherCAT security: no native authentication in the wire protocol, and what fills the gapEtherCAT's base protocol (ETG.1000, IEC 61158 Type 12) carries no authentication or encryption of its own — security depends on hardware frame processing, a closed topology, and now an IEC 62443-3-3 Security Level 2 certification layered on top. What academic MITM research and the one confirmed CISA advisory touching EtherCAT actually found.EtherNet/IP and CIP Security: why Rockwell's most-deployed protocol still ships unauthenticated by defaultEtherNet/IP and the Common Industrial Protocol (CIP) it carries run a large share of North American discrete manufacturing, and CIP has no authentication unless a site explicitly deploys ODVA's CIP Security extension. What CIP Security's five profiles actually add, what a 2021-to-2026 run of Rockwell Automation CISA advisories shows about real failures, and the controls that compensate where it isn't deployed.GE Vernova UR relay vulnerabilities: the 2025 CVE chain, explainedNozomi Networks disclosed five 2025 CVEs in GE Vernova's UR protection-relay family and EnerVista UR Setup software, two of which chain into a man-in-the-middle-to-malicious-firmware attack.

DNP3 protocol security: what Secure Authentication covers, and what it doesn't