DUFT: Device-Unique Feedback Traversal for Entropy-Estimated Response Extraction and Mutual Authentication using SRAM PUFs
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Security Radio. Generated commentary on the latest security and cryptography papers.
Nadia: We're looking at the paper: "DUFT: Device-Unique Feedback Traversal for Entropy-Estimated Response Extraction and Mutual Authentication using SRAM PUFs".
Elias: Here's the gist of the paper. Static Random Access Memory (SRAM) Physical Unclonable Functions (PUFs) are a promising technology in the realm of random number generation for key generation, but for PUF-based authentication, they have been sidelined due to their limited challenge space as a reservoir of Challenge-Response Pairs (CRPs) This paper develops a seed-based, round-iterated feedback mechanism called Device-Unique Feedback Traversal (DUFT) to extract an estimated min-entropy response and provide a larger logical input space for SRAM PUFs The DUFT Mechanism The core novelty of the paper is the device key, the traversal seed, and the round number are concatenated with the round’s stability mask and hashed using an unkeyed BLAKE2s-128 function to derive the next MSRAM address The process involves iteratively hashing a combination of these elements to select a unique traversal path over the physical source of SRAMs A collision-check step is introduced before accepting a candidate address to ensure no round within a single challenge’s traversal revisits a per-chip sub-address, which is managed by incrementing a per-round counter The accumulated response is then hashed using the BLAKE2s unkeyed function to serve as the fuzzy extractor privacy amplification step Entropy Estimation and Error Correction The number of rounds in the traversal scales directly with measured source-entropy, allowing security to be increased for weaker sources at the cost of running time The authors estimate 128 bits of estimated min-entropy on reference hardware under stated assumptions and experimental estimates Error correction with SRAM PUFs involves a two-stage approach where unstable bits are eliminated using a masking scheme and random errors in the stable bits are corrected using Bose–Chaudhuri–Hocquenghem (BCH) codes The sizing of the BCH code parameters, such as the number of errors corrected per window, is derived from measured mean and standard deviation of the window error count rather than assuming independence Protocol Design and Security Analysis The proposed protocol is a lightweight mutual authentication and key-encapsulation scheme that requires no pre-shared key or public-key cryptography The protocol utilizes the entropy-estimated response to support secure key establishment on resource-constrained nodes Security analysis shows that the probability of an adversary correctly predicting the DUFT response is bounded by 2−er + AdvPRF BLAKE2s(A) The protocol achieves mutual authentication by having both nodes compute a Hash-based Message Authentication Code (HMAC) over a session-bound message keyed with the shared session key Experimental Validation The evaluation was performed on physical IoT/embedded hardware, including a discrete-SRAM node pair and an embedded-SRAM microcontroller to realize the mechanism end-to-end Performance comparisons show that the total execution time of DUFT falls directly with minentropy, as the number of rounds falls with it The performance on the Cypress CY62256 is comparable to other PUF-based protocols, and single-digit-millisecond performance on the STM32F407 is promising DUFT provides a large input space, but the physical entropy remains that of the original SRAMs, and no wrapper changes the cost of characterizing that source In this paper, we addressed the need for extracting responses of a flexible size from SRAM PUFs containing an estimated or established amount of min-entropy, and for supplying a non-repeating stream of challenge-response pairs from a small physical source We utilize this improved response in an effective mutual authentication scheme in the IoT space as a lightweight alternative to traditional key-exchange schemes A novel mechanism called DUFT was utilized to extract this response while providing a large input space to prevent database exhaustion The feedback chain is driven by the device’s stability mask and device key hashed from its PUF array, which are itself manufacturing fingerprints, so each chip walks a different sequence of cells while keeping the feedback traversal secret A physical-source dependent calculation determines the number of rounds and the correction capability needed for a target security level, and thorough error-correction testing was performed such as stability mask profiling and BCH window error statistics Some limitations exist, DUFT provides a large input space, but the physical entropy is still that of the original SRAMs, and no wrapper changes the cost of characterizing that source The assumption that the physical adversary has no address-level access to the SRAMs is a strong one This is an assumption that is possible to meet and will be required for most weak-PUF based schemes Our entropy figures come from thirty-two Cypress CY62256 devices and twenty-eight STM32F407 devices and we do not claim they hold across process corners or manufacturing lots With these limitations, this work still provides a robust, secure, and lightweight protocol for IoT devices with the minimal resources of SRAMs REFERENCE arXiv:2611.15958v1
cs.CR: 11 Nov 2026 (This is the arXiv identifier, not a page number) Page 3 --- Syed Talha Imam, Amna Zafar, and Arslan Munir (Authors' Contact Information) Page 4 --- Syed Talha Imam, Amna Zafar, and Arslan Munir (Authors' Contact Information) Page 5 --- Table 1.
Nadia: Comparison of PUF-based authentication schemes and SRAM-PUF wrappers.
Elias: Schemes are identified by citation.
Nadia: ✓ = satisfied; ✗ = not satisfied; – = not applicable, not reported, or not directly comparable.
Paper summary: Nadia: So, wrapping up this discussion about DUFT: Device-Unique Feedback Traversal for Entropy-Estimated Response Extraction and Mutual Authentication using SRAM PUFs. This paper addresses the need to get responses of a flexible size from SRAM PUFs that already have an estimated amount of min entropy >
Elias: It also solves the problem of supplying a non-repeating stream of challenge response pairs from a small physical source, which is something that has been sidelining these technologies before >
Priya: The implication for me is that this gives us a way to use SRAM PUFs in IoT devices for mutual authentication, which is a lightweight alternative to traditional key exchange schemes >
Nadia: It provides the large input space needed to prevent database exhaustion in authentication systems and uses this improved response as a protocol for secure key establishment on resource-constrained nodes >
Elias: The mechanism relies on the device's stability mask and device key being hashed from its SRAM PUF array, which are manufacturing fingerprints, so each chip walks a different sequence of cells while keeping the feedback traversal secret >
Priya: The paper highlights that you need to perform thorough error correction testing, like stability mask profiling and BCH window error statistics, to get those numbers right for a target security level >
Nadia: Yes. And the authors give some figures from thirty-two Cypress CY62256 devices and twenty-eight STM32F407 devices for their entropy estimates, but they are careful not to claim those figures hold across different process corners or manufacturing lots >
Elias: So, the final thought is that while DUFT gives you a robust and secure protocol for IoT devices with minimal resources, you still have that limitation about the physical entropy being exactly what it was from the original SRAMs >
Conclusion: Nadia: So, we're looking at the end of this paper on DUFT, which is about using these SRAM PUFs for mutual authentication in IoT devices that have limited resources.
Elias: Yeah, it’s titled "DUFT: Device-Unique Feedback Traversal for Entropy-Estimated Response Extraction and Mutual Authentication using SRAM PUFs." It sounds a bit like a mouthful.
Priya: It does sound technical, but the core idea is they figured out how to get a response from an SRAM PUF that has an estimated amount of min entropy, even if it's not perfect.
Nadia: Exactly. They developed this device unique feedback traversal mechanism, DUFT, to extract a response that fits the entropy level they measured on the hardware.
Elias: And the main goal is to give these SRAM PUFs a larger input space than they have naturally, which stops you from running out of challenge response pairs quickly.
Priya: It’s about making sure that when you use this extracted response for key establishment, it’s secure, even with imperfect entropy.
Nadia: Right. The authors showed how the number of rounds in this traversal scales directly with the measured source entropy; so if the source is weaker, they just run more rounds to get better security.
Elias: They estimate about one hundred twenty eight bits of estimated min entropy on their reference hardware under their test assumptions, which is a decent starting point for these small memory chips.
Priya: Before we look at the results, I want to make sure we understand that they have to do some heavy lifting with error correction.
Nadia: They use a two-stage approach for the SRAM PUFs; first, they use a masking scheme to get rid of unstable bits, and then they correct those remaining random errors using something called Bose–Chaudhuri–Hocquenghem or BCH codes.
Elias: And they’re smart about that because instead of assuming every error is independent, they size the BCH code parameters based on the actual measured mean and standard deviation of the window error count.
Priya: That sounds like a pretty rigorous way to handle noise, which is what you really need when you're dealing with physical hardware imperfections.
Nadia: So for someone listening, it means this protocol lets IoT devices use SRAM PUFs securely for authentication without needing any pre-shared keys or complex public key math.
Elias: It’s a lightweight way to set up secure sessions on devices that have very little power and processing capability.
Priya: The security analysis shows the probability of an attacker guessing the DUFT response is limited by two to the exponent of the error plus the complexity of their hash function.
Nadia: That bound helps us understand how robust this system is against an adversary trying to predict what’s coming next in a traversal sequence.
Elias: But there are caveats, and you have to be aware of them because they're important.
Priya: They point out that the physical entropy itself is still just what the original SRAMs have, and no wrapper they put around it changes the underlying cost of characterizing that source.
Nadia: That’s a big limitation; DUFT can provide a huge input space, but you can't magically create more physical randomness if it wasn't there to begin with.
Elias: And the authors are careful about their claims too; they got their entropy figures from thirty-two Cypress CY62256 devices and twenty-eight STM32F407 devices, so those numbers aren't guaranteed across every manufacturing lot or process corner.
Priya: So, the world changes by giving us a practical protocol for using these specific hardware components in a real authentication scheme today.
Nadia: It’s a robust and lightweight approach for IoT devices, but we still have to respect the physical limits of the memory source itself.
SYED TALHA IMAM, AMNA ZAFAR, ARSLAN MUNIR
Florida Atlantic University
cs.CR
Submitted: 2026-11-11
Updated: 2026-11-11
Comments: 29 pages, 8 figures, 5 tables
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 78/100
The gist: Static Random Access Memory (SRAM) Physical Unclonable Functions (PUFs) are a promising technology in the realm of random number generation for key generation, but for PUF-based authentication…
Key concepts
- SRAM PUFs
- Static Random Access Memory Physical Unclonable Functions are hardware structures that generate unique, unclonable responses based on the physical characteristics of the memory cells. These responses serve as a source of randomness for key generation and authentication, but they often have limited challenge space.
- DUFT Mechanism
- DUFT is a seed-based, round-iterated feedback mechanism that uses device keys and stability masks to iteratively hash and select a unique traversal path over the SRAM physical source. This process generates a larger set of responses from the original PUF, increasing the effective input space for security.
- Entropy Estimation
- The number of rounds in DUFT scales with the measured source entropy, allowing security to be increased even when using weaker entropy sources. The authors estimate 128 bits of min-entropy on reference hardware, which helps determine how many iterations are needed for a target security level.
- BCH Codes
- A two-stage error correction process is used: masking eliminates unstable bits, and Bose–Chaudhuri–Hocquenghem (BCH) codes correct random errors in the stable bits. The parameters of these codes are determined by measuring the actual error distribution of the window to ensure effective data recovery.
Terminology
Summary
Static Random Access Memory (SRAM) Physical Unclonable Functions (PUFs) are a promising technology in the realm of random number generation for key generation, but for PUF-based authentication, they have been sidelined due to their limited challenge space as a reservoir of Challenge-Response Pairs (CRPs) This paper develops a seed-based, round-iterated feedback mechanism called Device-Unique Feedback Traversal (DUFT) to extract an estimated min-entropy response and provide a larger logical input space for SRAM PUFs
The DUFT Mechanism
The core novelty of the paper is the device key, the traversal seed, and the round number are concatenated with the round’s stability mask and hashed using an unkeyed BLAKE2s-128 function to derive the next MSRAM address The process involves iteratively hashing a combination of these elements to select a unique traversal path over the physical source of SRAMs A collision-check step is introduced before accepting a candidate address to ensure no round within a single challenge’s traversal revisits a per-chip sub-address, which is managed by incrementing a per-round counter The accumulated response is then hashed using the BLAKE2s unkeyed function to serve as the fuzzy extractor privacy amplification step
Entropy Estimation and Error Correction
The number of rounds in the traversal scales directly with measured source-entropy, allowing security to be increased for weaker sources at the cost of running time The authors estimate 128 bits of estimated min-entropy on reference hardware under stated assumptions and experimental estimates Error correction with SRAM PUFs involves a two-stage approach where unstable bits are eliminated using a masking scheme and random errors in the stable bits are corrected using Bose–Chaudhuri–Hocquenghem (BCH) codes The sizing of the BCH code parameters, such as the number of errors corrected per window, is derived from measured mean and standard deviation of the window error count rather than assuming independence
Protocol Design and Security Analysis
The proposed protocol is a lightweight mutual authentication and key-encapsulation scheme that requires no pre-shared key or public-key cryptography The protocol utilizes the entropy-estimated response to support secure key establishment on resource-constrained nodes Security analysis shows that the probability of an adversary correctly predicting the DUFT response is bounded by 2−er + AdvPRF BLAKE2s(A) The protocol achieves mutual authentication by having both nodes compute a Hash-based Message Authentication Code (HMAC) over a session-bound message keyed with the shared session key
Experimental Validation
The evaluation was performed on physical IoT/embedded hardware, including a discrete-SRAM node pair and an embedded-SRAM microcontroller to realize the mechanism end-to-end Performance comparisons show that the total execution time of DUFT falls directly with minentropy, as the number of rounds falls with it The performance on the Cypress CY62256 is comparable to other PUF-based protocols, and single-digit-millisecond performance on the STM32F407 is promising DUFT provides a large input space, but the physical entropy remains that of the original SRAMs, and no wrapper changes the cost of characterizing that source
Conclusion
In this paper, we addressed the need for extracting responses of a flexible size from SRAM PUFs containing an estimated or established amount of min-entropy, and for supplying a non-repeating stream of challenge-response pairs from a small physical source We utilize this improved response in an effective mutual authentication scheme in the IoT space as a lightweight alternative to traditional key-exchange schemes A novel mechanism called DUFT was utilized to extract this response while providing a large input space to prevent database exhaustion The feedback chain is driven by the device’s stability mask and device key hashed from its PUF array, which are itself manufacturing fingerprints, so each chip walks a different sequence of cells while keeping the feedback traversal secret A physical-source dependent calculation determines the number of rounds and the correction capability needed for a target security level, and thorough error-correction testing was performed such as stability mask profiling and BCH window error statistics Some limitations exist, DUFT provides a large input space, but the physical entropy is still that of the original SRAMs, and no wrapper changes the cost of characterizing that source The assumption that the physical adversary has no address-level access to the SRAMs is a strong one This is an assumption that is possible to meet and will be required for most weak-PUF based schemes Our entropy figures come from thirty-two Cypress CY62256 devices and twenty-eight STM32F407 devices and we do not claim they hold across process corners or manufacturing lots With these limitations, this work still provides a robust, secure, and lightweight protocol for IoT devices with the minimal resources of SRAMs
REFERENCE
arXiv:2611.15958v1 [cs. p. 0CR] 11 Nov 2026 (This is the arXiv identifier, not a page number)
--- Page 3 ---
Syed Talha Imam, Amna Zafar, and Arslan Munir (Authors' Contact Information)
--- Page 4 ---
Syed Talha Imam, Amna Zafar, and Arslan Munir (Authors' Contact Information)
--- Page 5 ---
Table 1. Comparison of PUF-based authentication schemes and SRAM-PUF wrappers. p. 4 Schemes are identified by citation. p. 4 = satisfied; = not satisfied; – = not applicable, not reported, or not directly comparable. p. 4
--- Page 6 ---
Syed Talha Imam, Amna Zafar, and Arslan Munir (Authors' Contact Information)
--- Page 7 ---
3 The Proposed Mechanism and Protocol
--- Page 8 ---
4 Syed Talha Imam, Amna Zafar, and Arslan Munir (Authors' Contact Information)
--- Page 9 ---
5.1 Hardware
--- Page 10 ---
5.2 Software
--- Page 11 ---
5.3 Performance Evaluation and Results
--- Page 12 ---
6 Conclusion
--- Page 13 ---
References (List of citations) (This is the reference list, not a page number)
--- Page 14 ---
Figure 3. Key-exchange and mutual authentication protocol based on our DUFT-PUF. p. 1
--- Page 15 ---
Table 4. Execution time comparison for SRAM PUF DUFT (CPU time). p. 0
--- Page 20 ---
Algorithm 3 (not present in the text)
--- Page 26 ---
Fig. 7. Error-distribution of the BCH correction window over 10,000 evaluations for the Cypress CY62256. p. 8
--- Page 27 ---
Fig. 8. Error-distribution of the BCH correction window over 1,000 evaluations for the STM32F407. p. 8
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
-
Device-Unique Feedback Traversal (DUFT) can be used to extract responses of an estimated min-entropy from SRAM PUFs, achieving up to 128 bits of estimated min-entropy on reference hardware under stated assumptions. This allows for a more secure response extraction mechanism than traditional methods.
-
The DUFT mechanism provides a large logical input space by deriving each round’s address from
feeding back the device’s stability-mask fingerprint output, the device’s PUF array hash, the input traversal-seed, and round number for a unique and secret traversal.
This prevents adversary exhaustion of the challenge space. -
A lightweight mutual authentication protocol can be implemented based on DUFT that requires
no pre-shared key or public-key cryptography,
making it suitable for resource-constrained embedded devices. This enables secure key establishment without heavy cryptographic overhead like elliptic curve cryptography. -
The system can support a PUF-based key encapsulation protocol that uses the entropy-estimated response to provide secure key establishment on resource-constrained nodes. This allows for robust communication security in IoT scenarios with minimal computational cost.
-
The DUFT mechanism incorporates an entropy estimation method that
uses the device’s measured per-bit min-entropy and measured error statistics to determine the number of rounds for our feedback-traversal wrapper for a target min-entropy response.
This ensures that the security level is tailored to the actual physical source characteristics. -
The system can be designed with an address traversal schedule that uses
the stability mask, seed and round number, but also a secret hash of the complete corrected array of the three SRAMs Kdev,
which makes each chip walk a different sequence of cells while keeping the feedback traversal secret. This provides a unique and device-specific traversal. -
The protocol can achieve mutual authentication by using
a Hash-based Message Authentication Code (HMAC) over a session-bound message J keyed with Ks
to confirm that both nodes possess the shared session key Ks, thereby ensuring mutual possession.
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