HHK: A Hardware-Oriented Cross-Location PPG Key Generation Architecture for Body Area Networks
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Electrical Engineering. The day in electrical engineering: signals, systems and the hardware that carries them.
Rosa: We're looking at the paper: "HHK: A Hardware-Oriented Cross-Location PPG Key Generation Architecture for Body Area Networks".
Dev: Here's the gist of the paper. Body area networks require lightweight session key establishment, yet public key exchange imposes computation and energy costs that exceed the budgets of deeply constrained wearable nodes The proposed datapath extracts inter-beat intervals (IBIs) from green-light PPG at multiple body sites, aligns beat timestamps across locations, applies Gray-coded equal-frequency quantization, and employs a rate-1/3 polar code fuzzy commitment (N = 128, K = 42) to reconcile residual timing mismatches.
Rosa: Post-implementation synthesis on a Xilinx XC7Z020 maps the complete datapath to 18 760 lookup tables and 20 971 flip-flops with no multipliers or embedded memories, giving 48 µJ per key generation event and 0.4 µW average power over a 120-second acquisition window.
Dev: Wireless body area networks connect wearable and implantable sensing nodes for continuous health monitoring, sports performance tracking, and closed-loop therapy.
Rosa: Securing these links is not optional: an unprotected BAN exposes sensitive health data to passive eavesdroppers and is vulnerable to active command injection in implantable devices.
Title and authors: Rosa: So we're diving into the paper "HHK: A Hardware-Oriented Cross-Location PPG Key Generation Architecture for Body Area Networks". We've talked about the basic concept of using body signals to create a shared secret key across different locations, and now let's look at what the authors actually built.
Dev: The title itself tells us a lot, emphasizing that this is a hardware-oriented architecture. It’s not just some software idea; they implemented it in synthesizable register-transfer level RTL on an Xilinx XC7Z020 FPGA.
Rosa: Right, so the hardware aspect means they focused intensely on making sure it runs efficiently and uses very little power, which is crucial for those deeply constrained wearable nodes we're talking about.
Taro: I'm interested in the fact that they chose to focus on inter-beat intervals instead of just looking at the raw waveform shape or amplitude features. That seems like a smart move because waveform morphology can change a lot depending on where you are, which makes it unreliable for key generation twelve <ref:2605.24991#pg2>.
Dev: That’s exactly what they point out. They argue that IBIs are physiologically invariant across body locations and much more robust to cross-location distortion than relying on the raw signal features themselves.
Rosa: So, the summary of what they actually did is that they created a complete datapath—a fixed-point preprocessor, a beat detector, an IBI timer, a quantizer using Gray code, and finally this polar encoder with a decoder—all done in hardware.
Dev: They claim this system works by broadcasting some helper data between the nodes and then having the receiving node compute something that allows them to recover the shared key material.
Taro: And they handle those timing mismatches using a rate one-third polar code fuzzy commitment, with one hundred twenty-eight bits of input and forty-two bits of output for that reconciliation step <ref:2605.24991#pg1>. That sounds like a specific mathematical tool for fixing errors.
The paper's summary: Rosa: Moving on to the summary section, what they’re really emphasizing is how they bridged the gap between theory and practical implementation in a way that hasn't been done before. They aren't just proposing an algorithm; they are showing a concrete system.
Dev: They highlight that biosignal processors for heart rate estimation or ECG authentication exist, but nothing has yet implemented a complete cross-location PPG key generation pipeline entirely in synthesizable RTL and validated it on actual silicon.
Rosa: So the summary stresses that HHK closes both those gaps at once—it’s a full system built from scratch that works in hardware. They specifically mention the challenges they tackled, like motion artifact impact on the wrist PPG and how their gating mechanism rejects bad data points <ref:2605.24991#pg3>.
Taro: The paper shows examples of challenging real-world conditions they tested it under, including things like cable cars, mountain trains, and varying solar radiation throughout the day. That shows they didn't just test it in a controlled lab setting; they pushed the limits of ambulatory monitoring.
Dev: They used three channels: head, wrist, and ankle. The sternum channel was excluded because their beat visibility there is not good enough for reliable IBI extraction during movement, which is a key limitation they admit to <ref:2605.24991#pg3>.
Rosa: So the summary boils down to this: they took the concept of using timing differences from multiple body sites and built a robust, hardware-based system that works on real-world data under tough conditions.
The paper's improvements: Rosa: Now let's talk about what they call their contributions or improvements. They frame it around making the system practical for real devices.
Dev: One major improvement is moving away from waveform morphology and amplitude features, which the paper says are highly site dependent and attack vulnerable twelve <ref:2605.24991#pg2,are highly site dependent and attack vulnerable 12>. HHK instead exploits IBI timing, which is physiologically invariant across body locations and more robust to distortion.
Taro: That’s important because if the signal changes based on where you are—like temperature or motion—then relying on that shape to generate a key is risky. Timing, however, should be more consistent.
Rosa: The second major improvement is the entire datapath being expressed in synthesizable RTL and validated on silicon using an AMD/Xilinx XC7Z020 programmable logic device <ref:2605.24991#pg2>. That’s huge because it proves it can run reliably in real hardware, not just on a simulator.
Dev: And they emphasize the efficiency gains too. They achieved this entire implementation with no multipliers or embedded memories, which is what allows them to achieve a very low energy per key generation event of forty-eight microjoules and an average power draw of only zero point four microwatt over a hundred and twenty seconds <ref:2605.24991#pg1>.
Taro: The hardware metrics are compelling because they show that this isn't just a theoretical concept; it has been mapped to concrete logic gates and flip-flops, which means we can actually design for it in constrained systems.
Rosa: So the improvements really center on robustness—timing invariance—and efficiency—achieving key generation with minimal hardware resources.
Conclusion: Dev: So wrapping up the HHK paper, what this means practically is that we have a demonstrable way for deeply constrained wearable nodes to establish secure sessions using only fixed-point arithmetic and hardware structures.
Rosa: It shows that you don't need massive computational budgets for security operations; you can get robust key agreement by focusing on timing differences derived from green-light PPG signals.
Taro: For someone just listening, the big picture is that this architecture sets a new benchmark for what’s possible in multi-site biometric key agreement systems tested on free-living data across head, wrist, and ankle.
Dev: The validation results are pretty strong too: they saw raw inter-beat interval bit agreement between eighty-six point zero and ninety point one across sixteen participants in a real ambulatory dataset one <ref:2605.24991#pg1,a real ambulatory dataset 1>. And the key agreement rate was in that range as well, between fifty one point three and sixty nine point eight percent.
Rosa: It’s the first time we’ve seen this specific approach validated on free-living data from multiple locations with such high agreement rates, which is a big deal for setting an open benchmark.
Taro: The limitation they mention is that they excluded the sternum channel because their beat visibility there isn't good enough for reliable IBI extraction during motion, so if you are trying to deploy this on a system where sternum PPG is your primary sensor, that method might not work as well.
Dev: That’s a fair caveat. But overall, HHK proves that by using this fixed-point pipeline and hardware polar codec, you can reconcile residual timing mismatches entirely in hardware without needing complex DSP multipliers or embedded memories for key generation.
Rosa: So to finish up on "HHK: A Hardware-Oriented Cross-Location PPG Key Generation Architecture for Body Area Networks", we’ve seen a method that is robust against cross-location distortion, works on real ambulatory data, and runs with extremely low power consumption.
Dev: It’s a solid piece of work showing how to get security features onto the edge devices where they actually need them most.
Taro: It opens the door for using these kinds of timing-based synchronization methods across different physiological signals in future wearable security applications.
King Abdullah University of Science and Technology
eess.SP
Submitted: 2026-05-24
Updated: 2026-11-02
Code: https://github.com/joseiltonfilho/hhk
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 86/100
The gist: Body area networks require lightweight session key establishment, yet public key exchange imposes computation and energy costs that exceed the budgets of deeply constrained wearable nodes The…
Key concepts
- Inter-Beat Intervals (IBIs)
- IBIs are the time differences between consecutive heartbeats measured from green-light photoplethysmography (PPG) signals at different body locations. Instead of analyzing complex waveform shapes, HHK focuses on these timing intervals, which are more stable across different monitoring sites and less susceptible to distortion caused by movement or sensor placement.
- Fixed-Point Arithmetic
- This is a type of digital math that uses integers with a fixed number of bits to represent fractional numbers. The paper uses Q15 fixed-point arithmetic (16-bit signed, 15 fractional bits) throughout the datapath. This approach allows complex calculations, like filtering and timing alignment, to be performed entirely using simple addition and subtraction operations on standard flip-flops, completely eliminating the need for hardware multipliers.
- Polar Code Fuzzy Commitment
- This is a specific error correction technique used in the reconciliation step to fix small timing mismatches between two nodes. It involves encoding data into a rate-1/3 polar code and using this structure to correct residual errors after initial alignment. This hardware mechanism ensures that even when timing is slightly off due to movement, the final key agreement succeeds.
- Hardware Synthesis on FPGA
- The entire key generation process is designed as synthesizable RTL (Register-Transfer Level) code and implemented on an Xilinx XC7Z020 FPGA. This means the design can be directly mapped onto physical hardware using only lookup tables and flip-flops, resulting in a highly efficient circuit with no embedded memories or multipliers, proving its feasibility for deeply constrained wearable devices.
Terminology
Summary
Body area networks require lightweight session key establishment, yet public key exchange imposes computation and energy costs that exceed the budgets of deeply constrained wearable nodes The proposed datapath extracts inter-beat intervals (IBIs) from green-light PPG at multiple body sites, aligns beat timestamps across locations, applies Gray-coded equal-frequency quantization, and employs a rate-1/3 polar code fuzzy commitment (N = 128, K = 42) to reconcile residual timing mismatches. p. 0 This brief presents HHK, a
hardware-oriented cross-location photoplethysmography (PPG)
key generation architecture for BANs. The proposed datapath
extracts inter-beat intervals (IBIs) from green-light PPG at
multiple body sites, aligns beat timestamps across locations,
applies Gray-coded equal-frequency quantization, and employs
a rate-1/3 polar code fuzzy commitment (N = 128, K = 42)
to reconcile residual timing mismatches. The proposed datapath
extracts inter-beat intervals (IBIs) from green-light PPG at
multiple body sites, aligns beat timestamps across locations,
applies Gray-coded equal-frequency quantization, and employs
a rate-1/3 polar code fuzzy commitment (N = 128, K = 42)
to reconcile residual timing mismatches. Post-implementation
synthesis on a Xilinx XC7Z020 maps the complete datapath to
18 760 lookup tables and 20 971 flip-flops with no multipliers
or embedded memories, giving 48 µJ per key generation event
and 0. Post-implementation
synthesis on a Xilinx XC7Z020 maps the complete datapath to
18 760 lookup tables and 20 971 flip-flops with no multipliers
or embedded memories, giving 48 µJ per key generation event
and 0. 4 µW average power over a 120-second acquisition window.Where the paper says it
Where the paper says it
Where the paper says it
Introduction to the Problem
Wireless body area networks connect wearable and implantable sensing nodes for continuous health monitoring, sports performance tracking, and closed-loop therapy. Securing these links is not optional: an unprotected BAN exposes sensitive health data to passive eavesdroppers and is vulnerable to active command injection in implantable devices. p. 0 INTRODUCTION
WIRELESS body area networks (BANs) connect
wearable and implantable sensing nodes for
continuous health monitoring, sports performance tracking,
and closed-loop therapy [2], [3]. Securing these links is not
optional: an unprotected BAN exposes sensitive health data
to passive eavesdroppers and is vulnerable to active command
injection in implantable devices. Session key establishment
must be renewed frequently to provide forward secrecy, yet
public key cryptography imposes computation and energy
costs that exceed the budgets of deeply constrained biomedical
nodes [4], and symmetric pre-distribution provides no forward
secrecy after compromise. Physiological signal-based key
generation offers an alternative: two nodes on the same body
simultaneously observe the same underlying physiology and
independently derive matching keys without transmitting
any secret [4], [5]. Physiological signal-based key
generation offers an alternative: two nodes on the same body
simultaneously observe the same underlying physiology and
independently derive matching keys without transmitting
any secret [4], [5]. Photoplethysmography (PPG) sensors,
already embedded in consumer smartwatches, fitness bands,
and ear clips, are the most accessible cardiac modality
for BANs. 1 illustrates the central challenge for cross-location key
agreement: simultaneous PPG recordings capture the same
cardiac rhythm but differ in waveform shape and motion
artifact content. The first critical gap in the literature is ambulatory,
multi-location validation.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
HHK Architecture Overview
Two BAN nodes at different body locations execute the same datapath independently and exchange only public helper data; no secret is ever transmitted. Raw 16-bit PPG samples enter a causal fixed-point prefilter that attenuates baseline wander and high-frequency noise without DSP multipliers. p. 1 HHK ARCHITECTURE
Two BAN nodes at different body locations execute the
same datapath independently and exchange only public helper
data; no secret is ever transmitted. Raw 16-bit PPG samples enter a causal fixed-point prefilter
that attenuates baseline wander and high-frequency noise
without DSP multipliers. Raw 16-bit PPG samples enter a causal fixed-point prefilter
that attenuates baseline wander and high-frequency noise
without DSP multipliers. All arithmetic is Q15 fixed-point
(16-bit signed, 15 fractional bits), enabling a multiplier-free
implementation. RTL IMPLEMENTATION AND SILICON VALIDATION
A. Architecture
The HHK datapath comprises six pipelined modules: a
fixed-point preprocessor, a peak-based beat detector, an IBI
timer, a Gray-coded quantizer, a polar encoder, and an
SC decoder with partial-sum propagation, all wrapped with
an AXI4-Lite control interface.Where the paper says it
Where the paper says it
Where the paper says it
Key Generation Protocol Details
The process involves several steps to reconcile timing mismatches between locations. The IBI is mapped to a 2-bit Gray code via equal-frequency bin edges, and concatenated to bA, bB. p. 1 Only
matched pairs contribute to IBI sequences, eliminating
beat-desynchronization errors that dominate naive sequential
comparison. The IBI range is partitioned into four
equal-frequency bins using dataset-wide percentiles, and each
Algorithm 1 IBI Extraction and Cross-Location Alignment
Require: PPG streams xA(t), xB(t) at fs = 128 Hz; tolerance δ =
300 ms
Ensure: Aligned IBI pairs; Gray-coded bit strings bA, bB ∈
{0, 1}
128
1: for each site u ∈ {A, B} do
2: Filter xu(t) with causal Q15 IIR bandpass filter
3: Detect local maxima above prominence threshold; enforce
400 ms refractory period
4: Compute IBIu(n) = tu(n + 1) − tu(n)
5: end for
6: M ← ∅
7: for each timestamp tA(i) do
8: j
∗ = arg minj |tA(i) − tB(j)|
9: if |tA(i) − tB(j
∗
)| ≤ δ then
10: M ← M ∪ {(IBIA(i), IBIB(j
∗
))}
11: end if
12: end for
13: Map each IBI in M to 2-bit Gray code via equal-frequency bin
edges; concatenate to bA, bB
Algorithm 2 Polar Fuzzy Commitment Key Agreement
Require: Bit strings bA, bB ∈ {0, 1}
128; polar code
(N=128, K=42)
Ensure: Shared key mˆ at node B, or FAIL
1: Node A: draw m
$← {0, 1}
42
2: Node A: c ← PolarEncode(m) ▷ rate- 1
3
, butterfly F
⊗7
3: Node A: broadcast helper h = bA ⊕ c
4: Node B: r ← bB ⊕ h
5: Node B: mˆ ← SCDecode(r) ▷ SC, partial-sum propagation
6: if decode ok then
7: return mˆ
8: else
9: return FAIL ▷ retry with next 120 s window
10: end if
IBI is mapped to a 2-bit Gray code. Algorithm 1 summarises the
complete extraction and alignment procedure. Node A draws a random key m ∈ {0, 1}
42, encodes it
to codeword c under a rate-1/3 SC polar code (N = 128,
K = 42), and broadcasts helper data h = bA ⊕ c. Node A draws a random key m ∈ {0, 1}
42, encodes it
to codeword c under a rate-1/3 SC polar code (N = 128,
K = 42), and broadcasts helper data h = bA ⊕ c. Node
B computes r = bB ⊕ h and SC-decodes to recover mˆ . Reconciliation succeeds when mˆ = m. Three successful
blocks yield 126 bits of shared key material, which SHA-256
compresses into an AES-128 session key.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Experimental Validation Results
Validation across 16 participants from a real ambulatory dataset yields 86.0–90.1% raw IBI bit agreement and 51. p. 0 Validation across 16 participants from a real ambulatory dataset
[1] (216 hours, head, wrist, and ankle) yields 86. 1% raw IBI
bit agreement and 51. 1% raw IBI
bit agreement and 51. 8% key agreement. 8% key agreement per reconciliation window. Head–wrist is the best performing pair at 90% raw agreement All prior systems operate
entirely in software and evaluate under controlled or laboratory
conditions. HHK is the only system providing a synthesizable
RTL implementation validated on silicon, and the only one
evaluated on free-living ambulatory data.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Hardware Implementation Metrics
Post-implementation synthesis on an AMD/Xilinx XC7Z020 maps HHK to 18 760 lookup tables and 20 971 flip-flops with no multipliers or embedded memories. Dynamic power on the FPGA prototype is 15 mW at 10 MHz, resulting in an energy per key event of 48 µJ. p. 4 X, XXXX 5
TABLE II
HHK POST-IMPLEMENTATION RESOURCE UTILIZATION AND POWER
(PYNQ-Z2, AMD/XILINX XC7Z020, VIVADO 2025. 2, 10 MHZ)
Module LUTs FFs
BPF Preprocessor 335 113
Foot Detector 178 84
IBI Timer & Quantizer 441 209
Polar Encoder & Control 358 556
SC Decoder 17 076 19 359
AXI4-Lite Interface 90 265
HHK IP Total 18 378 20 508
AXI Interconnect 364 426
Dynamic Power 15 mW at 10 MHz
Energy per Key Event 48 µJ
Avg. 2 ms = 48 µJ. Amortized over the 120-second PPG acquisition window, this
yields an average power draw of 0. The IBI-based pipeline tolerates
cross-location waveform distortion inherent to ambulatory
settings, and polar code reconciliation compensates residual
timing mismatches entirely in hardware. HHK is the first
synthesizable RTL PPG key generation architecture validated
on silicon across multiple body locations on free-living data. The IBI-based pipeline tolerates
cross-location waveform distortion inherent to ambulatory
settings, and polar code reconciliation compensates residual
timing mismatches entirely in hardware. HHK is the first
synthesizable RTL PPG key generation architecture validated
on silicon across multiple body locations on free-living data.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Conclusion
This brief has presented HHK, a multiplier-free RTL cross-location PPG key generation architecture for BANs. The fixed-point IBI pipeline and hardware polar codec deliver 86. p. 4 CONCLUSION
This brief has presented HHK, a multiplier-free RTL
cross-location PPG key generation architecture for BANs. The
fixed-point IBI pipeline and hardware polar codec deliver
86. Validation across 16 participants from a real ambulatory dataset
[1] (216 hours, head, wrist, and ankle) yields 86. 1% raw IBI
bit agreement and 51. 1% raw bit agreement and 51. 8% key agreement
across head, wrist, and ankle on 16 ambulatory participants,
with no multipliers, DSP slices, or embedded memories,
and 48 µJ per key event. 8% key agreement
across head, wrist, and ankle on 16 ambulatory participants,
with no multipliers, DSP slices, or embedded memories,
and 48 µJ per key event. The IBI-based pipeline tolerates
cross-location waveform distortion inherent to ambulatory
settings, and polar code reconciliation compensates residual
timing mismatches entirely in hardware.Where the paper says it
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
-
Hardware-Oriented Key Generation for Constrained Edge Devices This architecture allows deeply constrained wearable nodes to generate session keys using only fixed-point arithmetic, lookup tables, and flip-flops with no multipliers or embedded memories, achieving 48 µJ per key generation event and 0.4 µW average power over a 120-second window.
-
Cross-Location Biometric Key Agreement for Body Area Networks This system enables two nodes at different body sites to independently derive matching keys by extracting inter-beat intervals (IBIs) from green-light PPG, aligning beat timestamps within a 300 ms tolerance, and using a rate-1/3 polar code fuzzy commitment to reconcile residual timing mismatches.
-
Robustness Against Cross-Location Distortion This method exploits IBI timing instead of waveform morphology or amplitude features, which are
highly site dependent and attack vulnerable,
making the key generation robust against the distortions inherent in ambulatory settings like head, wrist, and ankle monitoring. -
Hardware Acceleration for Security Operations FPGA Implementation The complete datapath is implemented in synthesizable register-transfer level RTL on an FPGA (Xilinx XC7Z020), providing a
multiplier-free fixed-point IBI pipeline with Gray-coded quantization and a hardware polar codec
that validates simulation results withzero disagreement between RTL simulation and physical execution.
-
Real-World Ambulatory Data Validation Benchmarking This architecture establishes the highest reported key agreement in both software (51.3–69.8%) and hardware across head, wrist, and ankle locations on 216 hours of free-living ambulatory data from sixteen participants, serving as an
open benchmark for multi-location PPG key generation.
Abstract
Body area networks (BANs) require session keys that wearable nodes establish without user interaction and with proof that both nodes are on the same body. This paper presents a multi-site photoplethysmography (PPG) key-agreement processor for BANs, incorporating signal conditioning, band-pass filtering, pulse onset detection, beat alignment, Gray-coded quantization and a polar-code fuzzy commitment. The algorithm reaches 95.6% key agreement on resting wrist and fingertip recordings, and on 16 ambulatory participants the processor reaches 69.2, 54.8 and 50.0% between head and wrist, wrist and ankle, and head and ankle, respectively, against 2.3% for the best of five prior methods. Implemented and tested on an AMD XC7Z020 FPGA, the processor consumes 261 μ W. Synthesized with the NanGate 45 nm library, it occupies 0.394 mm squared, operates at up to 258 MHz and requires 0.208 μ J per decoding. We further attack the system and show that the public helper data reveals more bits than the inter-beat intervals carry, so an eavesdropper who knows only the mean heart rate recovers 39.4% of the session keys. To our knowledge, the processor is the first synthesizable RTL architecture for inter-beat-interval-based PPG key agreement in BANs evaluated on free-living ambulatory data.
Sources
Related papers
- Open-H-RF: A Large-Scale Ultrasound Raw Channel Dataset Initiative
- Scalable Attention for 5G NR Channel Estimation
- Low-Altitude Wireless Networks: A Comprehensive Survey
- Sensing-Assisted OFDM with Index Modulation: A Waveform for Optical Wireless Integrated Communication and Localisation
- Iteration-Free Candidate Pruning and an Index-Domain VLSI Architecture for Sort-Free MIMO Box Decoding up to 1024-QAM
- A Foundation Model for Large-Scale Wireless Network Planning, Operation and Optimization