AMORT-V1: Checkpointed Post-Quantum Attestation for Industrial Controllers
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Security Radio. Generated commentary on the latest security and cryptography papers.
Nadia: I'm Nadia.
Elias: And I'm Elias, here with Priya.
Nadia: We're looking at the paper: "AMORT-V1: Checkpointed Post-Quantum Attestation for Industrial Controllers".
Elias: Here's the gist of the paper. Programmable logic controllers often remain deployed for decades, making post-quantum attestation a long-term requirement AMORT-V1, a checkpointed attestation protocol that moves public-key signing and durable commitment from individual reports to periodic checkpoints.
Nadia: The Core Problem Addressed Shor’s algorithm allows a quantum computer to forge RSA and elliptic-curve signatures Programmable logic controllers (PLCs) are exposed to that forgery risk for longer than most systems, because operators keep them in service for decades.
Elias: Remote attestation defends against such alterations by letting a verifier check the software state of a PLC.
Nadia: If a quantum adversary can forge those signatures, the verifier cannot detect the tampering.
Paper summary: Nadia: So we're looking at this paper called "AMORT-V1: Checkpointed Post-Quantum Attestation for Industrial Controllers". The big idea here is that programmable logic controllers, or PLCs, are stuck in the field for decades. This means we need a way to prove their software hasn't been tampered with even when quantum computers can forge the digital signatures they use.
Elias: Exactly. And the problem with current methods is that if a quantum adversary can forge those signatures, our verifier can't tell if anything has changed on the PLC because the signature itself looks valid to them. This paper proposes moving public-key signing and durable commitment from every single report to periodic checkpoints instead.
Priya: So they're trying to make the process more efficient by not doing heavy public-key math for every little report, which sounds like it could save a lot of time in the field, but I wonder if that efficiency comes at a cost to what we can actually trust about the device's origin.
Nadia: That's right. The paper claims that this checkpointed protocol, AMORT-V1, uses an ML-KEM epoch to establish per-device evidence keys and then authenticates each report with a symmetric message authentication code under those keys. Then they commit reports through hash-chained Merkle checkpoints signed with ML-DSA or SLH-DSA.
Elias: The core security property they're hitting is that shared symmetric MACs don't provide a transferable proof of device origin, which means the controller still holds the key and can't prove authorship on its own. So AMORT-V1 adds a device-held disclosure chain that only the device can compute, which lets an outside party verify authorship after commitment.
Priya: That sounds like it tries to fix the problem of proving who made what, even when you ditch the public key signing for every single message. But what about how long it takes for that third party to actually check this disclosure chain?
Nadia: They measure the delay and verification cost of that disclosure chain, and they find that in-order traversal keeps verification in the microsecond range, but adding that key disclosure adds one checkpoint interval of delay.
Elias: And they look at three research questions. RQ1 is about how the checkpoint interval trades amortized postquantum cost against omission detection delay. RQ2 asks if AMORT-V1 can recover publicly transferable device origin authentication after replacing per-report signatures with symmetric evidence, and at what cost? And RQ3 looks at whether these per-device epochs can preserve failure isolation and support recovery from things like revocation or rekeying without bringing back the need for durable state.
Priya: So they're trying to make sure that if one part of the system fails, it doesn't take down the whole attestation mechanism, which is important when you're dealing with these industrial controllers. But they also have to worry about how much data this checkpointing adds to the overall cost per report.
Nadia: They do look at the controller cost per report and find that checkpointing reduces the ML-DSA sixty-five cost per report by ninety times without a durable log, and by eight hundred seventy-eight times with one.
Paper summary: Elias: That's a big reduction in the signing overhead, but they also have to manage how many records are between checkpoints to expose omissions. They found that the next checkpoint exposes an omission after a median of about half the records, or about thirty-six and sixty four records when you use sixty four records for k.
Priya: So increasing that checkpoint interval k reduces the per-report cost by spreading out those checkpoint operations over more evidence, but they say it proportionally increases the maximum delay before an omission becomes visible in a checkpoint. It's a trade off between speed and detection time, I guess.
Nadia: They also looked at scaling across different controller deployments. The amortized controller cost drops from four hundred ninety one point zero microseconds per report at ten identities down to six point one eight microseconds per report at ten thousand identities because the fixed checkpoint work gets spread over more records.
Elias: That scaling behavior is interesting, showing that the fixed checkpoint work is distributed effectively across a larger set of reports, which is good for massive deployments. But they also noted that this cost reduction only happens because the checkpoint work is fixed per interval, not because they eliminated it entirely.
Priya: So for someone listening who just cares about what this means in practice, the paper suggests that post-quantum attestation doesn't always need a public-key signature and a durable write for every report. But they also flagged that checkpointing alone isn't enough to prove device origin to a third party without adding that disclosure chain.
Nadia: Right. The authors conclude that checkpointing can amortize the authentication and persistence work, but it cannot reduce the cost of repeatedly measuring a large software image itself. So you still have to measure that big image periodically anyway, just less frequently for signing purposes.
Elias: It means they aren't eliminating the need for measurement; they are just changing how often you do the heavy cryptographic lifting related to proof of identity. This points toward future work combining this AMORT-V1 approach with incremental measurement and hardware-assisted measurement.
Priya: So, what this means for us is that while we can get closer to practical post-quantum attestation for long-lived industrial equipment, we still need more steps—like using hardware assistance or smarter measurement caching—to make the whole system truly robust against those quantum threats.
Nadia: That’s the direction they point toward. The checkpointing is a way to manage the heavy cryptographic load over time, but it doesn't solve every single problem in one go. This paper shows a solid way to handle the long-term deployment challenge for PLCs in the context of post-quantum security.
Elias: So we have this protocol AMORT-V1 that trades per-report signing for periodic, authenticated checkpoints, while carefully managing the trade offs between speed and detection latency. It’s a concrete step toward making these long-term industrial systems more secure against quantum attacks.
Conclusion: Nadia: So we're wrapping up on "AMORT-V1: Checkpointed Post-Quantum Attestation for Industrial Controllers." Basically, this paper says you don't have to sign every single report with a public key anymore when dealing with these long-lived industrial controllers.
Elias: Right. The authors are looking at how to handle the quantum threat—Shor’s algorithm breaking RSA signatures—without making every single attestation round trip take forever.
Priya: But they're moving the heavy math around, not eliminating it entirely, so we need to see what that actually means for trust and privacy in the field.
Nadia: They introduce this checkpointing system, which moves the public-key stuff from every report to periodic checkpoints. It’s a way to save massive amounts of computation on the controller side.
Elias: And they’re using ML-KEM epochs to create device keys and then authenticating reports with symmetric codes under those keys, which is much faster than doing full post-quantum signatures for every message.
Priya: The paper talks about six security properties they aim for, like freshness and device binding, which sounds promising if it works in the real world. But I want to know how that symmetric evidence actually proves who made the report to someone outside the system.
Nadia: That’s where they add a device-held disclosure chain that only the controller can compute, which lets an outside party verify authorship after commitment, restoring that device origin authentication.
Elias: And they show that this verification process is fast, keeping it in the microsecond range for in-order traversal. But there is a tradeoff: adding that key disclosure adds one checkpoint interval of delay to the verification time.
Priya: So you're trading a little bit of immediate speed for a verifiable proof of origin later on. That’s something we need to really look at before we trust this as the final answer.
Nadia: Exactly. The conclusion is that checkpointing helps amortize the authentication and persistence work over time, but it doesn't magically make measuring a large software image cheaper per report itself.
Elias: It means you still have to measure that whole big software image repeatedly, but you only do the heavy signing math less often than before. That’s a practical way to manage the long-term deployment challenge for PLCs against quantum threats.
Priya: So, the authors are proposing this as a solid step forward in managing post-quantum security for industrial stuff by spreading out the cost of proof generation over time, but they've flagged that future work needs to combine this with incremental measurement.
J. Warren McClure School of Emerging Communication Technologies, Scripps College of Communication · School of Electrical Engineering and Computer Science, Russ College of Engineering and Technology, Ohio University
cs.CR
Submitted: 2026-11-05
Updated: 2026-11-05
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 88/100
The gist: Programmable logic controllers often remain deployed for decades, making post-quantum attestation a long-term requirement AMORT-V1…
Key concepts
- Post-Quantum Attestation
- This is a security process that verifies the software state of a device, specifically designed to resist attacks from future quantum computers capable of breaking current public-key cryptography like RSA. It ensures that an industrial controller has not been tampered with or altered.
- Checkpointing
- Instead of signing every single status report individually, AMORT-V1 groups reports into periodic checkpoints. These checkpoints are signed using strong post-quantum algorithms, which drastically reduces the computational load on the controller for each update.
- Device Binding
- This mechanism links a specific report to a device's identity and the time period (epoch) it was generated. It ensures that evidence is tied to a known hardware entity, making it harder for an attacker to substitute reports from another device.
Terminology
Summary
Programmable logic controllers often remain deployed for decades, making post-quantum attestation a long-term requirement AMORT-V1, a checkpointed attestation protocol that moves public-key signing and durable commitment from individual reports to periodic checkpoints. p. 0 edu
Abstract—Programmable logic controllers often remain deployed for decades, making post-quantum attestation a long-term
requirement. Prior measurements of OpenPLC attestation show
that post-quantum signing can consume 88–99 % of round-trip
latency and that durable state can further limit throughput.Where the paper says it
The Core Problem Addressed
Shor’s algorithm allows a quantum computer to forge RSA and elliptic-curve signatures Programmable logic controllers (PLCs) are exposed to that forgery risk for longer than most systems, because operators keep them in service for decades. Remote attestation defends against such alterations by letting a verifier check the software state of a PLC. p. 0 An attacker who alters that software can tamper
the physical process and hide the alteration, as Stuxnet did
when it reprogrammed centrifuge logic [3]. Remote attestation
defends against such alterations by letting a verifier check the
software state of a PLC. The check depends on signatures
over the reported evidence. If a quantum adversary can forge
those signatures, the verifier cannot detect the tampering. A prior study measured that per-report work for postquantum attestation on OpenPLC [18]. Hash-based signing
consumed 88-99 % of each attestation round trip.Where the paper says it
Where the paper says it
Where the paper says it
Protocol Architecture and Security Properties
AMORT-V1 moves asymmetric authentication and durable commitment from individual reports to periodic checkpoints. The protocol establishes per-device evidence keys from an ML-KEM epoch, authenticates each report with a MAC under that key, and commits reports through hash-chained Merkle checkpoints signed with ML-DSA or SLH-DSA. p. 1 • Checkpointed post-quantum attestation. AMORT-V1
derives a per-device evidence key from an ML-KEM
epoch, authenticates each report with a MAC under that
key, and commits reports through hash-chained Merkle
checkpoints signed with ML-DSA or SLH-DSA. Containment limits the effect of a device compromise to that
device and its current epoch. Finally, detectable equivocation
allows conflicting checkpoint histories to serve as evidence
of controller misbehavior once the conflicting views are compared.Where the paper says it
Where the paper says it
Performance Analysis and Tradeoffs
The checkpoint model defines the amortized controller cost per report as cost(k) = Cmac + Cleaf + Csign + Cfsync k. Checkpointing reduces the ML-DSA-65 cost per report by 90× without a durable log and by 878× with one. p. 5 Table II gives the endpoints. Checkpointing reduces the MLDSA-65 cost per report by 90× without a durable log and
by 878× with one, and it reduces the SLH-DSA-128f cost by
2979×. 9 %, which puts k
⋆
at 12739. In
the omission trials, the next checkpoint exposed the omission
after a median of about k/2 records and never after more than
k: 36 and 64 records at k = 64, and 2014 and 4079 records
at k = 4096. These results answer RQ1. Increasing k reduces per-report
cost by amortizing checkpoint operations over more evidence,
but proportionally increases the maximum delay before a
checkpoint-visible omission can be exposed.Where the paper says it
Where the paper says it
Where the paper says it
Restoring Device Origin
AMORT-V1 adds a device-held disclosure chain that only the device can compute, which lets a third party verify authorship after commitment. This mechanism restores transferable device-origin authentication without a public-key operation on each report. p. 6 These results answer RQ2. AMORT-V1 restores transferable
device-origin authentication without a public-key operation
on each report. AMORT-V1 restores transferable
device-origin authentication without a public-key operation
on each report. In-order traversal keeps verification in the
microsecond range, while key disclosure adds one checkpoint
interval of delay.Where the paper says it
Where the paper says it
Recovery and Scaling Capabilities
AMORT-V1 provides two recovery paths: under freshepoch recovery, a restarted device discards its current disclosure state and opens a new epoch, while under rollback-protected recovery, the device anchors its sequence state to a monotonic hardware counter and resumes the current epoch after restart. The protocol also supports scaling across controller deployments, where amortized controller cost falls from 491. p. 6 The experiments in
this section maintains a fixed device population, so the next
experiment asks whether the amortized cost survives as the
device count grows. Scaling Across Controller Deployments
While deploying a series of controllers each device adds
a sequence range and subtree root to every checkpoint. Transferable device-origin verification costs 2. 95 µs per report,
while controller-side scaling to 10 000 identities requires 6. 0 µs per report at 10 identities to 6. 18 µs at 10 000 because
fixed checkpoint work is spread across more records.Where the paper says it
Where the paper says it
Where the paper says it
Conclusion and Future Directions
AMORT-V1 shows that post-quantum attestation does not require a public-key signature and durable write for every report. Checkpointing alone is not sufficient because symmetric evidence cannot prove device origin to a third party. p. 8 This structure turns repeated post-quantum cost
into a configurable tradeoff between attestation overhead and
detection delay. Checkpointing alone is not sufficient because symmetric
evidence cannot prove device origin to a third party. We also show checkpointing can amortize authentication
and persistence, but it cannot reduce the cost of repeatedly
measuring a large software image. Future work should therefore combine AMORT-V1 with incremental measurement,
hardware-assisted measurement, and trusted measurementstate caching. These mechanisms provide a path toward long-lived
post-quantum attestation for controller fleets without placing
expensive cryptographic or durable-state operations on every
report. We also show checkpointing can amortize authentication
and persistence, but it cannot reduce the cost of repeatedly
measuring a large software image.Where the paper says it
Where the paper says it
Where the paper says it
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Checkpointed post-quantum attestation allows programmable logic controllers to remain deployed for decades by moving public-key signing and durable commitment from individual reports to periodic checkpoints, which reduces controller cost by 90 times relative to per-report ML-DSA without durable logging and by 878 times when durable logging is required.
-
The AMORTV1 protocol enables transferable device origin authentication under symmetric evidence by adding a device-held disclosure chain whose keys are released only after a checkpoint commitment, allowing a third party to verify authorship after commitment.
-
The system supports rollback resistant restart through TPM backed monotonic state without placing durable hardware operations on the per report path, providing recovery paths like fresh epoch recovery and TPM backed reservation schemes to stop sequence reuse under rollback.
Abstract
Programmable logic controllers often remain deployed for decades, making post-quantum attestation a long-term requirement. Prior measurements of OpenPLC attestation show that post-quantum signing can consume 88--99% of round-trip latency and that durable state can further limit throughput. We introduce AMORT-V1, a checkpointed attestation protocol that moves public-key signing and durable commitment from individual reports to periodic checkpoints. An ML-KEM epoch establishes per-device evidence keys, devices authenticate reports with MACs, and the controller commits ordered evidence through hash-chained Merkle checkpoints signed with ML-DSA or SLH-DSA. Because shared MACs do not provide transferable proof of device origin, AMORT-V1 adds a device-held disclosure chain whose keys are released after checkpoint commitment. We validate the protocol with a 27-case attack matrix, 149 differential-test vectors, and a seven-node Raspberry Pi 5 testbed with five physical OpenPLC provers. Checkpointing reduces controller cost by 90 times relative to per-report ML-DSA without durable logging and by 878 times when durable logging is required. Transferable device-origin verification costs 2.95micro sec per report, while controller-side scaling to 10,000 identities requires 6.18 micro sec and 155 octets per report. AMORT-V1 also supports rollback-resistant restart through TPM-backed monotonic state without placing durable hardware operations on the per-report path.
Related papers
- From Strategy Search to Policy Learning: MCTS-Guided Adversarial Co-Evolution for LLM Safety Alignment
- Privacy Leakage from Recurrent State in LLM Inference
- Incrementally Verifiable Oracle Aided Computation
- Privilege and confidentiality in generative AI workflows
- BadMoE: Backdooring Mixture-of-Experts LLMs via Routing-Aware Triggers and Task-Underutilized Experts
- CoVault: Secure, Scalable Analytics of Personal Data