There are two paths to stability ― carefully correct the variation, or minimize it from the start.
▲ img source: Generated with ChatGPT
Our last post walked through how a VIA actually forms during chip fabrication, and why it matters that connection and disconnection land close to a balanced split rather than leaning heavily one way. Before manufacturing, which physical state a given VIA will settle into is genuinely hard to predict. Once the chip is fabricated, though, that outcome becomes a fixed, permanent trait of that specific chip.
That raises an obvious follow-up question.
"If a value is hard to predict, doesn't that mean it's also unstable — coming out slightly different every time you measure it?"
For most PUFs, yes. A PUF response is a physical measurement, not a calculated number, so it naturally drifts a little across changes in temperature, voltage, and aging. The rate at which repeated readings differ is called the Bit Error Rate (BER), and the standard way to correct that drift and reconstruct the exact same key every time is a Fuzzy Extractor — a combination of an Error-Correcting Code (ECC) and Helper Data.
So here's the question this post actually asks: why can VIA PUF™ produce a stable key without that standard correction machinery at all? The answer isn't that VIA PUF™ corrects drift especially well. It's that VIA PUF™ derives its value from a point in the physical process where there's far less drift to begin with.
1. VIA PUF™ Starts From a Different Kind of Physical Signal
SRAM PUFs and Ring Oscillator (RO) PUFs derive their 0s and 1s by comparing a tiny analog difference between two devices — which of two near-identical voltages or frequencies edges out the other. The narrower that margin, the easier it is for a small shift in temperature or voltage to flip the outcome.
VIA PUF™'s physical structure starts from a fundamentally different place. Each VIA PUF™ cell is a passive metal structure — no active transistors are involved at all. Rather than comparing which of two circuits runs faster or which voltage is marginally higher, each cell's value is set by whether a conductive path actually formed through a microscopic via hole in the insulating layer (Connection, or "short") or never formed at all (Disconnection, or "open") — two physically distinct states, not two points on a shared continuum.
A structure that settles into one of two states as far apart as "fully connected" and "fully open" has much less room for a shifting environment to flip the outcome after manufacturing. This isn't a clever way of canceling out drift — it's choosing, from the start, a physical phenomenon that has very little drift to cancel.
2. So Bit Values Hold Steady Across Repeated Readings
That structural difference shows up in real data, too. ICTK's public vPUF Security IP datasheet reports no bit-value changes across more than 9 billion vPUF cells in mass production, evaluated against the Steadiness requirement in ISO/IEC 20897-1 and 20897-2. Randomness has likewise been validated against standard statistical test suites such as NIST SP 800-90B.
That said, this result shouldn't be read as a mathematical guarantee of zero errors under every conceivable condition — no physical system can promise that. What the mass-production data supports is narrower and more useful: at production scale and under real operating conditions, the kind of bit drift that would require a separate correction step essentially doesn't show up.
3. So ECC and Helper Data Aren't Really Needed Here
Both ECC and Helper Data only earn their keep if there's drift to correct in the first place. When repeated measurements already come back consistent, the case for a dedicated correction layer — and the auxiliary data that layer depends on — largely falls away.
That's not just "skip a step." An ECC block adds its own silicon area, computation time, and power draw, plus another layer of design and verification complexity. Helper Data has its own overhead too — it has to be computed at enrollment, stored somewhere, and retrieved and processed on every single reconstruction, which means securely managing that storage becomes part of the system design. Because VIA PUF™ needs comparatively little of either layer, the path from silicon to key can stay considerably simpler.
4. In Summary: Two Different Roads to the Same Stability
One balance worth striking here: none of this makes Fuzzy Extractors, ECC, or Helper Data inferior technology. Quite the opposite — it's a well-established, thoroughly validated approach for turning a noisy physical signal into a trustworthy cryptographic key, grounded in solid theory and years of practical implementation. For PUFs like SRAM or RO PUF that read out an analog threshold, that stabilization layer remains genuinely necessary.
VIA PUF™'s difference isn't in how it solves that problem — it's in where the problem starts. Because it builds on Connection and Disconnection, two physically distinct states rather than a fine-grained comparison, it can reach the same goal — a key that reconstructs reliably every time — through a considerably lighter architecture. Stability can come from careful correction after the fact, or from choosing a physical foundation that doesn't drift much in the first place. Both are valid paths, and VIA PUF™ happens to take the second one.
Frequently Asked Questions
What does it actually mean to say VIA PUF™ has essentially no BER?
It doesn't mean the error probability is mathematically zero. Based on ICTK's published datasheet, it means no bit-value changes have been observed across mass-production data covering more than 9 billion vPUF cells — a result validated against the Steadiness requirement in ISO/IEC 20897-1 and -2.
Doesn't skipping ECC and Helper Data create a security weakness?
If anything, it's closer to the opposite. Without a separate error-correction layer and Helper Data storage/reconstruction pipeline, there's less added circuitry, less stored auxiliary data, and fewer processes to manage overall. A simpler architecture generally means fewer places that need to be verified.
Could other PUFs be built without ECC too?
It depends on the physical mechanism behind the PUF. PUFs that derive their values from analog threshold comparisons tend to see more environmental drift, and for those architectures, ECC and Helper Data-based stabilization remains a valid, well-proven solution.
Are ECC and Helper Data bad technology, then?
Not at all. They're a widely validated, standard method — used across academia and industry — for turning a noisy physical signal into a trustworthy key. VIA PUF™ simply needs less of that layer because of its underlying physical characteristics; that says nothing about the value of the technology itself.
References
See How VIA PUF™ Works
Curious about VIA PUF™'s Steadiness data, or how a noisy physical signal normally gets turned into a stable key in the first place?
☑️ Why Does a PUF Give a Different Reading Every Time?
☑️ Explore the VIA PUF™ Security IP
Read more
There are two paths to stability ― carefully correct the variation, or minimize it from the start.
▲ img source: Generated with ChatGPT
Our last post walked through how a VIA actually forms during chip fabrication, and why it matters that connection and disconnection land close to a balanced split rather than leaning heavily one way. Before manufacturing, which physical state a given VIA will settle into is genuinely hard to predict. Once the chip is fabricated, though, that outcome becomes a fixed, permanent trait of that specific chip.
That raises an obvious follow-up question.
"If a value is hard to predict, doesn't that mean it's also unstable — coming out slightly different every time you measure it?"
For most PUFs, yes. A PUF response is a physical measurement, not a calculated number, so it naturally drifts a little across changes in temperature, voltage, and aging. The rate at which repeated readings differ is called the Bit Error Rate (BER), and the standard way to correct that drift and reconstruct the exact same key every time is a Fuzzy Extractor — a combination of an Error-Correcting Code (ECC) and Helper Data.
So here's the question this post actually asks: why can VIA PUF™ produce a stable key without that standard correction machinery at all? The answer isn't that VIA PUF™ corrects drift especially well. It's that VIA PUF™ derives its value from a point in the physical process where there's far less drift to begin with.
1. VIA PUF™ Starts From a Different Kind of Physical Signal
SRAM PUFs and Ring Oscillator (RO) PUFs derive their 0s and 1s by comparing a tiny analog difference between two devices — which of two near-identical voltages or frequencies edges out the other. The narrower that margin, the easier it is for a small shift in temperature or voltage to flip the outcome.
VIA PUF™'s physical structure starts from a fundamentally different place. Each VIA PUF™ cell is a passive metal structure — no active transistors are involved at all. Rather than comparing which of two circuits runs faster or which voltage is marginally higher, each cell's value is set by whether a conductive path actually formed through a microscopic via hole in the insulating layer (Connection, or "short") or never formed at all (Disconnection, or "open") — two physically distinct states, not two points on a shared continuum.
A structure that settles into one of two states as far apart as "fully connected" and "fully open" has much less room for a shifting environment to flip the outcome after manufacturing. This isn't a clever way of canceling out drift — it's choosing, from the start, a physical phenomenon that has very little drift to cancel.
2. So Bit Values Hold Steady Across Repeated Readings
That structural difference shows up in real data, too. ICTK's public vPUF Security IP datasheet reports no bit-value changes across more than 9 billion vPUF cells in mass production, evaluated against the Steadiness requirement in ISO/IEC 20897-1 and 20897-2. Randomness has likewise been validated against standard statistical test suites such as NIST SP 800-90B.
That said, this result shouldn't be read as a mathematical guarantee of zero errors under every conceivable condition — no physical system can promise that. What the mass-production data supports is narrower and more useful: at production scale and under real operating conditions, the kind of bit drift that would require a separate correction step essentially doesn't show up.
3. So ECC and Helper Data Aren't Really Needed Here
Both ECC and Helper Data only earn their keep if there's drift to correct in the first place. When repeated measurements already come back consistent, the case for a dedicated correction layer — and the auxiliary data that layer depends on — largely falls away.
That's not just "skip a step." An ECC block adds its own silicon area, computation time, and power draw, plus another layer of design and verification complexity. Helper Data has its own overhead too — it has to be computed at enrollment, stored somewhere, and retrieved and processed on every single reconstruction, which means securely managing that storage becomes part of the system design. Because VIA PUF™ needs comparatively little of either layer, the path from silicon to key can stay considerably simpler.
4. In Summary: Two Different Roads to the Same Stability
One balance worth striking here: none of this makes Fuzzy Extractors, ECC, or Helper Data inferior technology. Quite the opposite — it's a well-established, thoroughly validated approach for turning a noisy physical signal into a trustworthy cryptographic key, grounded in solid theory and years of practical implementation. For PUFs like SRAM or RO PUF that read out an analog threshold, that stabilization layer remains genuinely necessary.
VIA PUF™'s difference isn't in how it solves that problem — it's in where the problem starts. Because it builds on Connection and Disconnection, two physically distinct states rather than a fine-grained comparison, it can reach the same goal — a key that reconstructs reliably every time — through a considerably lighter architecture. Stability can come from careful correction after the fact, or from choosing a physical foundation that doesn't drift much in the first place. Both are valid paths, and VIA PUF™ happens to take the second one.
Frequently Asked Questions
What does it actually mean to say VIA PUF™ has essentially no BER?
It doesn't mean the error probability is mathematically zero. Based on ICTK's published datasheet, it means no bit-value changes have been observed across mass-production data covering more than 9 billion vPUF cells — a result validated against the Steadiness requirement in ISO/IEC 20897-1 and -2.
Doesn't skipping ECC and Helper Data create a security weakness?
If anything, it's closer to the opposite. Without a separate error-correction layer and Helper Data storage/reconstruction pipeline, there's less added circuitry, less stored auxiliary data, and fewer processes to manage overall. A simpler architecture generally means fewer places that need to be verified.
Could other PUFs be built without ECC too?
It depends on the physical mechanism behind the PUF. PUFs that derive their values from analog threshold comparisons tend to see more environmental drift, and for those architectures, ECC and Helper Data-based stabilization remains a valid, well-proven solution.
Are ECC and Helper Data bad technology, then?
Not at all. They're a widely validated, standard method — used across academia and industry — for turning a noisy physical signal into a trustworthy key. VIA PUF™ simply needs less of that layer because of its underlying physical characteristics; that says nothing about the value of the technology itself.
References
See How VIA PUF™ Works
Curious about VIA PUF™'s Steadiness data, or how a noisy physical signal normally gets turned into a stable key in the first place?
☑️ Why Does a PUF Give a Different Reading Every Time?
☑️ Explore the VIA PUF™ Security IP
Read more