DNP3 Secure Authentication Version 5 (SAv5) — the challenge-response extension IEEE 1815-2012 added to the DNP3 protocol — makes SHA-256 the mandatory default hash algorithm for its HMAC message authentication codes and permits AES-256 Key Wrap for protecting key material, replacing the pre-shared-key-only model of the earlier SAv2 with a remote key-update mechanism. A 2017 formal analysis using the Tamarin prover found that SAv5's core design meets its intended security properties and debunked a previously claimed replay attack, but that same body of research and independent technical reporting agree on where the real problem sits: not the cryptography, but the operational burden of getting update keys to every outstation across a utility's service territory, which is consistently cited as the reason SAv5 adoption remains thin relative to how long the mechanism has existed.
Key takeaways
- It adds per-message proof of origin and integrity for critical requests, and nothing else.
- SAv5 still lives in IEEE 1815-2012, but that standard is no longer active and its successors use a different mechanism.
- A device challenges each critical request and acts only if the reply carries a valid MAC.
- Session keys change often and automatically; update keys change rarely and are the hard part.
What does SAv5 actually add that base DNP3 doesn't have?
It adds per-message proof of origin and integrity for critical requests, and nothing else. Base DNP3 has no field for a credential, a session key or a signature, so any device that can reach an outstation can issue a function code it accepts. The DNP Users Group's SAv5 release note (1 December 2011) states that SAv5 "supersede[s] all earlier versions, including the Version 2 procedures (SAv2) that were included in IEEE 1815-2010", that SAv2 "should not be deployed in new installations", and that "SAv2 and SAv5 procedures are not directly interoperable".
SAv5 is not encryption. The DNP Users Group's own talking points list "Not encryption" and point to TLS for IP networks, and a Trout Software write-up puts it plainly: "Nothing is encrypted. Process values, point names and the structure of the operation stay readable on the wire." Confidentiality needs DNP3 over TLS (IEC 62351-3), a separate project with its own certificates; the protocol background is in our DNP3 protocol security guide.
Where do IEEE 1815, IEC 62351-5 and the next DNP3 security layer stand?
SAv5 still lives in IEEE 1815-2012, but that standard is no longer active and its successors use a different mechanism. The current state:
| Document | Status (read 2026-11-14) | Security mechanism |
|---|---|---|
| IEEE 1815-2012 | "Inactive-Reserved", inactivated 2023-03-30 | SAv5 (challenge-response, aggressive mode) |
| IEEE P1815 | Active project; IEEE NesCom recommended extending it to 31 December 2027 | To include DNP3-SL |
| IEC TS 62351-5:2013 | Replaced | Basis of SAv5; added remote Update Key change |
| IEC 62351-5:2023 | International Standard | "Challenge/Reply procedure and concepts were removed", aggressive mode "replaced with the Secure Data message exchange mechanism", "Authenticated encryption of application data was added" |
The DNP Users Group's DNP3-SL flyer describes DNP3-SL, "formerly Secure Authentication Version 6", as "an implementation of the new IEC 62351-5 International Standard" that "Supports message encryption" 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". Until that edition is published, SAv5 is the interoperable option products ship; our IEC 62351 guide covers the other parts of the series.
How does the challenge-response flow actually work end to end?
A device challenges each critical request and acts only if the reply carries a valid MAC. SAv5 is carried inside normal DNP3 traffic using Group 120 objects (g120v1–v15, as named in the open-source opendnp3 object definitions). The flow, per the SAv5 specification text the DNP Users Group released:
ClassificationTable 7-7 marks function codes as critical. Write, Select, Operate, Direct Operate, Cold and Warm Restart, Initialize/Start/Stop Application, Enable/Disable Unsolicited Responses, Record Current Time, the file functions and Activate Configuration are MANDATORY critical; Read and Unsolicited Response are optional, so routine polling is usually not authenticated. ChallengeThe receiver holds the request and sends an Authentication Challenge (g120v1) with random data and a Challenge Sequence Number. ReplyThe sender returns an Authentication Reply (g120v2) with a MAC over the challenge and the original request, computed with the current session key for that direction. VerificationThe receiver computes the same MAC and executes the request only on a match; failures are counted, and the challenger stops sending error messages after a configurable limit. Aggressive modeThe sender may "anticipate" the challenge and send the MAC (g120v3 + g120v9) in the same message. The specification says all DNP3 implementations "are required to support it" and must "permit it to be disabled by configuration".
On the wire, an illustrative exchange (synthetic values) for one breaker command over TCP/IP looks like this:
| Step | Direction | Function code | Objects |
|---|---|---|---|
| 1 | Master → outstation | 0x05 Direct Operate | g12v1 control relay output block |
| 2 | Outstation → master | 0x83 Authentication Response | g120v1 Challenge, CSQ 41, random challenge data |
| 3 | Master → outstation | 0x20 Authentication Request | g120v2 Reply, CSQ 41, 16-octet HMAC-SHA-256 |
| 4 | Outstation → master | 0x81 Response | g12v1 status, after the MAC checks out |
In aggressive mode, steps 1–3 collapse into one Direct Operate carrying g120v3, g12v1 and g120v9.
MAC lengths depend on the link: HMAC-SHA-256 "shall be the default MAC algorithm" and is truncated to 16 octets over TCP/IP or 8 on serial links; HMAC-SHA-1 must be implemented for slow devices, is "not the preferred MAC algorithm", and every device must let it be disabled.
How do session keys and update keys change?
Session keys change often and automatically; update keys change rarely and are the hard part. The specification defines:
| Key | Default lifetime | Changed by |
|---|---|---|
| Session keys, one per direction | 15 minutes or 1000 authentication messages, whichever comes first | Master, wrapped with the update key (g120v6) |
| Update key, per user | Until changed; expiry configured by the Authority | Out of band, or remotely through an Update Key Change (g120v13) |
| Authority key (symmetric) or Authority key pair (asymmetric) | Long-term | Out of band |
Remote update key change was the main SAv5 addition: the release note says it "can use either symmetric or asymmetric methods". The specification lets a device skip remote change altogether, but any device that implements it must support Key Change Method 3 (AES-128 Key Wrap with HMAC-SHA-1, which must be disableable) and Method 4 (AES-256 Key Wrap with HMAC-SHA-256), and "This method shall be the default." Asymmetric methods (RSA with DSA signatures) are optional. The specification also warns against stretching the session key interval beyond 30 minutes without increasing the MAC length.
How do SAv2, SAv5 and DNP3-SL compare?
Each generation fixed the previous one's key or algorithm limits. The release note's appendix lists what SAv5 changed: it "changes the minimum length of an HMAC to 8 octets instead of 4", "makes SHA-256 a mandatory hash algorithm, and the default", makes it a requirement that SHA-1 "can be disabled by configuration", moves the mandatory TLS cipher suite to one supported by TLS 1.2, and "optionally permits" AES-256 Key Wrap and AES-GMAC. It also says the changes were needed for IEEE 1815 to enter the Smart Grid Interoperability Panel catalog.
| SAv2 (IEEE 1815-2010) | SAv5 (IEEE 1815-2012) | DNP3-SL (draft, IEEE P1815) | |
|---|---|---|---|
| Update keys | Pre-shared only | Pre-shared or remote change, symmetric or asymmetric | No pre-set keys; X.509 and ML-KEM key establishment, per the DNP3-SL flyer |
| Default MAC | SHA-1 permitted as default | HMAC-SHA-256 | HMAC and/or AEAD |
| What is authenticated | Critical requests | Critical requests | All messages |
| Encryption | No | No | Supported |
| Interoperability | — | Not with SAv2-only devices, unless the SAv5 device can be set to SAv2 | "Can co-exist with DNP3-SAv5" |
For a product team, the practical reading is: ship SAv5 now with SHA-256 defaults, keep the SAv2 fallback switchable and off by default, and design the key store so a later security layer can reuse it.
Did independent researchers find real weaknesses in SAv5, or is it sound as specified?
The core design holds, with specification gaps. Cremers, Dehnel-Wild and Milner's ESORICS 2017 report (extended in the Journal of Computer Security 27(2), 2019) modelled the full protocol in the Tamarin prover and found "the core DNP3: SAv5 design meets its intended security properties". It also showed that a replay attack previously reported by Amoah, Çamtepe and Foo (2016) "does not work, as an outstation will not accept a non-aggressive mode message replayed into the Aggressive Mode".
Its recommendations for future versions:
- Update Key Change messages (g120v13) should name the intended outstation.
- The specification "must clarify the use of Challenge Sequence Numbers"; if they need not strictly increase, "replay attacks of Aggressive Mode Requests become possible".
- Devices should support asymmetric cryptography, preferably elliptic-curve, for key changes.
- "Deprecate HMAC-SHA-1"; "HMAC-SHA-256 should be required at minimum."
- Clarify parsing, HMAC calculation and how long challenge data is kept.
The ambiguity points matter for implementers: two compliant stacks can behave differently on sequence numbers.
Why does SAv5 reportedly stay rare in deployed utility networks?
The evidence points at key logistics and legacy hardware rather than any rejection of the cryptography:
- Key distribution at scale. Update keys still have to reach every outstation, often over slow serial links across a service territory.
- Bootstrapping left open. A 2015 post on DNP3 security progress by a participant in the interoperability work says EPRI testing found implementations "interoperate almost seamlessly", but that key management needed a separate DNP3 Key Management Protocol (DKMP), expected to become "a non-normative annex to the next edition of IEEE 1815".
- Legacy fleets. Many RTUs and IEDs cannot run SAv5 firmware, so utilities adopt it on new installations and leave older outstations unauthenticated.
- Mixed networks. SAv2, SAv5 and unauthenticated devices can share one network, so an attacker targets whichever outstation was never upgraded.
No DNP Users Group survey publishes an adoption percentage, so none is claimed here.
Should you use SAv5, DNP3 over TLS, or both?
Both, on IP links that carry anything confidential; SAv5 alone on serial. The SAv5 specification says implementations "over TCP/IP requiring confidentiality shall use both this application layer mechanism and the Transport Layer Security (TLS)", and over UDP/IP "shall use this application layer authentication mechanism by itself". The DNP Users Group's talking points add that SAv5 "Works end-to-end in mixed networks of IP and serial, including serial radios", which TLS cannot do.
| SAv5 | DNP3 over TLS (IEC 62351-3) | |
|---|---|---|
| Protects | Origin and integrity of critical requests | Confidentiality and integrity of the whole TCP session |
| Works over serial and radio | Yes | No |
| Survives protocol gateways | Yes, end to end at the application layer | No, ends at each TLS endpoint |
| Keys | Symmetric session/update keys, optional asymmetric | X.509 certificates |
| Port | DNP3 port 20000 | The IANA registry lists "Distributed Network Protocol - Secure" on 19999 |
Worked example: what does an SAv5 configuration review find?
Synthetic example, not a real utility. A distribution operator reviews one master and four outstations:
| Device | Finding | Action |
|---|---|---|
| Outstation A (2019 RTU) | SAv5, HMAC-SHA-256, aggressive mode enabled | Keep; confirm CSQ handling with the vendor |
| Outstation B (2012 RTU) | SAv5 with HMAC-SHA-1 still enabled | Disable SHA-1; check the device can handle SHA-256 |
| Outstation C (2008 IED) | No Secure Authentication | Compensating controls: segmentation and allow-listed masters |
| Outstation D (2015 gateway) | Session key interval raised to 2 hours, 8-octet MAC on a serial link | Return to 15 minutes or lengthen the MAC |
| Master | Update keys loaded by hand at commissioning, never changed | Define an update key change schedule and a Method 4 process |
Firmware age matters as much as configuration here: the gateway's own software components can be matched against CVE and advisory data from an SBOM, viewed with the SBOM viewer.
What should asset owners and RTU or master vendors verify?
Which version runs on each link:SAv5, not SAv2, and which devices have none at all. HMAC-SHA-1 disabledwherever every peer supports HMAC-SHA-256, the specification default. Key Change Method 3 and asymmetric SHA-1 methods disabled, with Method 4 or an asymmetric SHA-256 method in use. Session key interval and MAC lengthat or near the defaults; long intervals need longer MACs. Critical function listcovering at least the MANDATORY codes in Table 7-7. Update key process documented:who holds the Authority key, how often keys change, how a compromised key is replaced. No "SAv5 = encrypted" entriesin audit records; confidentiality needs TLS (IANA lists DNP-Secure as port 19999). SAv5 inside a segmented network, never instead of one; see NERC CIP-005 and NIST SP 800-82.
What are the common deployment pitfalls?
- Leaving HMAC-SHA-1 on "for compatibility" after the last SHA-1-only device has gone.
- Hand-loaded update keys that never change, so one leaked commissioning file compromises every outstation it covered.
- Long session key intervals with short MACs on serial links, against the specification's 30-minute warning.
- Assuming two compliant stacks agree on Challenge Sequence Numbers; test the exact master-outstation pair.
- Treating aggressive mode as broken. It is mandatory to support and the formal analysis did not break it; disable it only by policy.
- Waiting for DNP3-SL before deploying anything, while IEEE has published no edition that contains it.
What can a passive capture show, and what can it not?
SoftWSP's OT Snapshot builds a per-host inventory from a passive capture, and for DNP3 it shows which hosts spoke DNP3 and which sat on the outstation side. It does not decode Group 120 objects, so it does not report the Secure Authentication version, MAC algorithm or key settings; those come from device configuration and vendor documentation. For RTU or gateway firmware, a device & firmware security scan extracts a CycloneDX SBOM and matches it against NVD, CISA KEV, FIRST EPSS and CISA ICS advisories, with an AI-drafted, cited report that an engineer reviews.
This guide draws on the DNP Users Group's SAv5 release note, talking points and 2026 DNP3-SL documents, the SAv5 specification text as released, IEEE SA records for 1815-2012 and P1815, the IEC 62351-5:2023 abstract and TS 62351-5:2013 preview, the Cremers, Dehnel-Wild and Milner analysis, and the technical write-ups cited above, read on 2026-11-14.
Put this guide to work
Frequently asked questions
What is DNP3 Secure Authentication, and how is it different from just using DNP3?
DNP3 Secure Authentication (DNP3-SA) is an optional, additive extension to the base DNP3 protocol, standardized inside IEEE Std 1815 rather than as a separate specification — the current version, Secure Authentication Version 5 (SAv5), is based on IEC/TS 62351-5 and was added in the IEEE 1815-2012 revision, superseding the pre-shared-key-only Secure Authentication Version 2 (SAv2) that shipped in IEEE 1815-2010. Base DNP3 has no field for a credential, session key, or message signature at all, so any device that can reach a master or outstation can issue a function code it accepts. DNP3-SA does not change that base protocol or wrap it in a tunnel; instead it defines application-layer objects (DNP3 Object Group 120) and a state machine that let a master and outstation challenge each other and verify, via a keyed message authentication code, that a specific critical request actually came from the party holding the right key and was not altered in transit. Industry technical write-ups describing the mechanism are explicit that this buys authentication and integrity, not confidentiality: process values, point names, and the structure of the operation remain readable on the wire whether or not SAv5 is running, because DNP3-SA was never designed to encrypt traffic — that is a separate problem IEC 62351-3's DNP3-over-TLS profile addresses, with its own, independent key management burden.
How does the SAv5 challenge-response mechanism actually work?
SAv5 runs as a two-pass challenge-response exchange modeled on ISO/IEC 9798-4, layered inside normal DNP3 traffic via Group 120 authentication objects (g120v1 through g120v15) rather than a separate wire protocol. In the baseline flow: a device designates certain incoming requests as critical (operate commands, cold/warm restarts, configuration changes, and security-relevant functions like key changes); when a critical request arrives, the receiving device issues a Challenge object containing a random challenge value; the originator of the critical request computes a Message Authentication Code (MAC) over that challenge using a shared session key and returns it in a Reply object; the receiver independently computes the same MAC and only acts on the original request if the two values match. SAv5 also defines an Aggressive Mode, where the request and its authentication data are sent together in a single message rather than requiring the full two-pass round trip, trading a small amount of security margin for lower latency and bandwidth use — a deliberate option given how constrained some serial DNP3 links still are. Non-critical traffic, including routine polling and most unsolicited responses, is not authenticated by default under this model, which is a scope decision, not an oversight: authenticating every poll on a bandwidth-constrained serial link was judged too expensive for the protection it would add.
What cryptographic algorithms does SAv5 actually use, and what changed from SAv2?
SAv5's most consequential change from SAv2 is less about the challenge-response logic itself and more about key distribution and algorithm defaults. On algorithms: SAv5 makes SHA-256 the mandatory, default hash algorithm for HMAC-based message authentication, while the older SHA-1 option can still be configured but is no longer the default; it also permits the AES-256 Key Wrap algorithm for protecting key material in transit and the AES-GMAC algorithm as an alternative MAC construction to HMAC-SHA-256. The key hierarchy has three tiers: short-lived session keys actually used to compute per-message MACs; medium-term update keys used to securely deliver new session keys; and long-term authority or pre-shared keys used to establish update keys in the first place. On distribution: SAv2 supported only pre-shared (out-of-band) update keys, which meant physically or manually provisioning key material to every outstation before authentication could run at all. SAv5's defining addition, per the DNP Users Group's own release materials, was a remote Update Key change mechanism — letting a central Authority deliver new update keys to outstations over the same DNP3 link, in response to explicit requests from the DNP3 user community to reduce the operational burden of out-of-band key distribution. That remote mechanism reduced the logistical cost, but did not eliminate it, which is the gap most of the real-world adoption commentary below is about.
Has independent research found actual weaknesses in the SAv5 protocol design itself?
The most rigorous published analysis available is "Secure Authentication in the Grid: A Formal Analysis of DNP3: SAv5" by Cas Cremers, Martin Dehnel-Wild, and Kevin Milner, presented at ESORICS 2017 and later extended in the Journal of Computer Security (Vol. 27, No. 2, 2019). The authors used the Tamarin prover, a symbolic protocol-verification tool, to model SAv5's full state machine, including its sub-protocols, its Aggressive Mode, and the possibility of cross-protocol interactions — the first analysis to cover the complete protocol rather than an isolated piece of it. Their headline finding is reassuring on protocol design: the core SAv5 design meets its intended security properties as formally modeled, and a previously claimed replay attack combining aggressive and non-aggressive modes (reported in earlier academic work on DNP3-SA) does not actually succeed against the standard as written, because an outstation will not accept the non-aggressive message under the conditions that attack requires. Where the paper is more cautionary is on algorithm agility: it is consistent with the broader cryptographic community's position that HMAC-SHA-1, still a configurable (non-default) option under SAv5, should be deprecated in favor of requiring HMAC-SHA-256 at minimum, given SHA-1's known collision weaknesses. In short — published formal analysis has not found a protocol-logic break in SAv5 as specified, but it does reinforce that any deployment still permitting the legacy SHA-1 option is carrying risk the standard's own default configuration no longer requires.
Why does SAv5 reportedly see low adoption in deployed utility OT networks, despite the protocol being over a decade old?
Multiple independent technical sources converge on the same answer: key management overhead, not distrust of the cryptography. SAv5 requires update-key and session-key material to reach every outstation across a utility's service territory, often over slow, intermittent, or serial-only links, and — per commentary associated with people involved in the DNP3 Technical Committee's own interoperability testing — large parts of the "behind-the-scenes" bootstrapping work needed to get Secure Authentication running at scale were left largely unspecified by the base standard, prompting a separate effort (the DNP3 Key Management Protocol, DKMP, developed with EPRI and Triangle MicroWorks involvement) aimed at standardizing that bootstrapping rather than leaving it to each vendor. That DKMP work was expected to land as a non-normative annex rather than a mandatory part of IEEE 1815, which by its own non-normative status is unlikely to drive adoption without separate regulatory pressure. Compounding this, a large installed base of DNP3 outstations — RTUs and IEDs from the 1990s and 2000s — cannot run SAv5 firmware at all, so even utilities that have adopted SAv5 commonly report doing so only on newer installations, leaving the pre-existing fleet unauthenticated and creating a mixed environment where a capable attacker can simply target the devices that were never upgraded. None of this is a claim this guide can attach a precise industry-wide adoption percentage to — no primary DNP Users Group survey publishes a hard adoption number — but the direction and mechanism of the gap (key logistics and legacy hardware, not a rejection of the cryptography) is corroborated across vendor technical writing, protocol-committee-adjacent commentary, and the academic literature alike.
What should an asset owner or RTU/master vendor actually verify about their DNP3-SA deployment?
Five checks, roughly in order of how often each one is actually the gap in practice. First, confirm SAv5 — not SAv2 — is the negotiated version on a given link; SAv2's pre-shared-key-only model carries the operational weaknesses SAv5 was built to fix, and both versions can coexist on the same network per EPRI interoperability testing, so a mixed deployment can quietly be running the weaker of the two. Second, check whether HMAC-SHA-1 is still enabled as a configurable option on any device, and disable it in favor of the SHA-256 default, since that is the one place a credible formal analysis (Cremers, Dehnel-Wild, and Milner, ESORICS 2017 / JCS 2019) flags a concrete algorithm-level recommendation rather than a logic flaw. Third, inventory which outstations are actually running SAv5-capable firmware versus which are legacy hardware that cannot be upgraded at all, since a mixed fleet's unprotected legacy devices are the easiest target regardless of how well the newer devices are configured. Fourth, treat SAv5 as authentication and integrity only, never as confidentiality: a configuration audit or compliance checklist that records 'SAv5 enabled' as equivalent to 'traffic is encrypted' is recording a false control, and separate transport-layer protection (DNP3 over TLS, per IEC 62351-3) is a distinct project with its own certificate and key management overhead. Fifth, ask whether the update-key distribution process for new or re-keyed outstations is manual and out-of-band, or uses SAv5's remote Update Key change mechanism — the manual path is exactly the operational burden that industry commentary consistently cites as the reason SAv5 rollouts stall at a handful of sites rather than scaling across a full service territory.
Sources
- DNP Users Group: Further Information Regarding the Release of DNP3 Secure Authentication Version 5 (dnp.org)
- IEEE 1815-2012: IEEE Standard for Electric Power Systems Communications — Distributed Network Protocol (DNP3) (standards.ieee.org)
- Cremers, Dehnel-Wild & Milner, "Secure Authentication in the Grid: A Formal Analysis of DNP3: SAv5" (ESORICS 2017 technical report, cs.ox.ac.uk)
- DNP3 Security: SAv5, Encryption and What It Misses (Trout Software)
- DNP3 Secure Authentication Version 5 (SAv5): Strengthening SCADA Communication Security (scadaprotocols.com)
- Progress in DNP3 Security — DNP3 Key Management Protocol (DKMP) and EPRI interoperability testing (rlc.vlinder.ca)
- IEEE SA: IEEE P1815 active PAR (revision of IEEE 1815-2012)
- IEEE SA NesCom recommendations, 22 October 2025 (P1815 extension to 31 December 2027)
- IEC 62351-5:2023, Security for IEC 60870-5 and derivatives (IEC webstore abstract and change list)
- IEC TS 62351-5:2013 preview (iTeh): foreword and change list
- DNP Users Group: DNP3 Secure Authentication talking points
- DNP Users Group: DNP3 Security Layer (DNP3-SL) and AMP flyer
- DNP Users Group: membership update, January 2026 (DNP3-SL draft and IEEE P1815)
- DNP3 Secure Authentication Version 5 specification text, 2011-11-08 (as released by the DNP Users Group; scadahacker.com mirror)
- Cremers, Dehnel-Wild & Milner, Journal of Computer Security 27(2), 2019 (author preprint, CISPA)
- opendnp3 (open-source DNP3 implementation, Group 120 object definitions)
- IANA Service Name and Transport Protocol Port Number Registry (dnp 20000, dnp-sec 19999)
About this guide
SoftWSP OT/ICS Desk
Covers the OT/ICS security cluster: 87 guides.

In this cluster
Questions about this page?
