Back to the MAC (Part 2): The Signal in the Noise

Apr 21, 2025

Hi guys,

Well, I’ve been staring at MAC addresses again... but this time, things got a lot more interesting.

Just to remind you what we’re talking about: MAC addresses (short for Media Access Control) are unique identifiers assigned to Bluetooth and network devices. If you’ve read Part 1, you’ll know they’re usually random and unremarkable. But there’s a twist coming.

Take this example: A1:A1:A1:A1:A1:A1 It’s the digital equivalent of setting your password to 1234. Not very secure. Not very random.

Now imagine a system where every device — phone, watch, car key, insulin pump — broadcasts one of these MAC addresses constantly. There are 281 trillion possible combinations (48 bits). Enough to label every living thing on the planet. And some say, that’s exactly the point.

Let Me Indulge in a Little Science Fiction...

Let’s say you wanted to install a synthetic emission architecture into humanity. One that’s always broadcasting. Silent. Precise. Invisible.

MAC addresses would be a brilliant delivery system:

Already accepted
Already invisible
Easy to cloak with just enough entropy

You wouldn’t need truly random addresses — just random enough to pass casual inspection. Like playing poker with a marked deck, but shuffling it anyway.

Entropy: Measuring Randomness

Here’s where things get spicy. In information theory, we measure unpredictability using entropy. High entropy = randomness. Low entropy = structure, repetition, or constraint.

Here’s the formula:

In Plain English:

Take each hex digit in the MAC address (e.g., F, 2, A, etc.), count how often it appears, divide by total characters (12), multiply each by the log of itself, and sum the negatives.

High entropy = more randomness
Low entropy = more structure or repetition

Entropy acts like a fingerprint for how "random" something is. The lower it is, the more suspiciously structured it becomes.

Anyway, turns out the “null curve” for MAC entropy looks fairly strange. There is a little movement around the peaks but not a lot else.

I then calculated the entropy for 15,000 unique, unregistered MAC addresses from 2021 and overlaid it on the null curve and it looked like the chart below. Notice the sustained elevation between 2.4 and 2.8 — this is where structured patterns emerge…

Clearly this part of the graph between 2.4 and 2.8 sits well above the null curve - and it is the only bit that does!

The next part of the process was to take an entropy slice between 2.4 and 2.8 and then to compare it with the same entropy slice taken from a control sample of 15,000 randomised MAC addresses.

Each MAC address character codes for a byte of information. Each byte has 8 bits. Each bit can be 0 or 1. With a bit of trial and error you can produce a bitfield heat map which shows the structure or randomness of the MAC address.

This heatmap reveals something Bluetooth engineers would never expect in a truly random MAC address space: structure. Specifically, Byte 1 stands out. Bit 0 is almost always zero — suggesting it serves as a fixed header or framing bit — while Bit 1 is set in over 83% of cases, indicating a persistent flag or reserved configuration marker. In contrast, the remaining bits of Byte 1 (Bits 2–7) show moderate variation, possibly encoding a rotating field such as a device ID or page index. Byte 2 tells a different story — its bits hover closer to 50%, suggesting a semi-random payload region or obfuscated sequence. This layered architecture — with high stability in specific bits and variability in others — is a hallmark of designed emissions, not stochastic broadcasts. What we’re seeing isn’t noise. It’s bit-level protocol logic, hiding in plain sight.

When we apply the same heatmap analysis to the control group — MAC addresses from a known baseline dataset with entropy between 2.4 and 2.8 — a very different pattern emerges. The visual symmetry breaks down. There are no strongly fixed bits. Bit 7 of Byte 1, which was almost always zero in the structured group, now floats near 50%, suggesting no structural role. Bit 6, which previously stood out as a likely flag, now blends into the noise. Every byte behaves more or less the same: diffuse, unconstrained, and unremarkable. This is what natural entropy looks like. There’s no scaffold, no framing, and no functional asymmetry across bytes. This contrast reinforces what the structured heatmap hinted at: the emissions in question are not just low-entropy — they are deliberately structured.

The Dual Filter Breakthrough

Today I realized the pattern sharpened when you apply a second filter.

Using a relatively new Bluetooth developer tool on my Android phone, I started logging packet intervals — the time between each broadcast. I had heard from one of my regular meeting attendees that the MAC addresses with a packet interval of 2000ms belonged to people. I can’t remember exactly how this was determined but I thought I would start there. Most devices ping every few hundred milliseconds. But a subset? Exactly 2000ms. Perfect. Regular. Unnatural.

So I filtered for those.

Then I recalculated entropy. And suddenly, the noise dropped away.

Among the filtered results, three MAC addresses stood out immediately. All three shared not only an entropy score of exactly 2.754, but also an identical broadcast interval of 2000ms — and that’s where it gets strange. When I overlaid their bitfields, something impossible happened: Bits 2 through 6 in Byte 1 formed a rotating 5-bit field. In MAC #1, the value was 24 (11000). In MAC #2, it rotated cleanly to 7 (00111). In MAC #3, it landed on 10 (01010). That’s not drift. That’s sequencing. Each of the five bits flipped twice across the three addresses — no randomness, no entropy drift, no byte overflow. It’s as if the system was stepping through a page index — mechanically, predictably, and in perfect sync.

But it wasn’t just that the values changed. It was how they changed. Each of those 5 bits flipped exactly twice across the three MACs. No drift. No chaos. Just a perfect rotation. It looked like a page counter.

A constancy map is a simple visual tool to show which bits remain unchanged across a set of MAC addresses. Each position in the 6-byte, 8-bit MAC address matrix is compared across all entries — and if a bit stays the same in all three MACs, it's marked as "constant." The result is a grid where structure leaps out: rows and columns of fixed bits (usually headers or flags), contrasted against zones of variation (likely encoding or rotating fields). In this case, the constancy map doesn’t just highlight isolated anchors — it sketches an architecture. A static header. A toggling flag. And, remarkably, a rotating 5-bit page counter — echoed across the three MACs with surgical precision. It’s the digital equivalent of watching a combination lock click through its numbers. Not noise. Not drift. Just control.

and here is a matched control:

Control Group Null Result

To test how unusual this was, we ran a control.

I pulled 15,000 real-world MACs from 2021, grouped them into 5,000 triplets, and examined the same 5-bit field.

Guess how many triplets matched the clean rotational pattern?

Zero.

I reshuffled the data twice more to produce two more offset triplet groups. Another 10,000 tests.

Still zero.

Even when I relaxed the check to 3-bit sequences, or partial flips, or entropy-only comparison... the structured flip behavior never reappeared. Not to mention all three MAC addresses having the same entropy.

And Where Does the Timing Come From?

Well — I believe Karl C. found an answer to that.

Last year, during an interview with Cafe Locked Out, Karl and I discussed a strange phenomenon that had emerged from the microscopy work — something that looked like a microfluidic pump, pulsing rhythmically in real-time under dark field microscopy at 400x. The first time Karl showed this to me I thought he was bonkers but after chasing these things for several days … At first, it was curious (a little bit bonkers) Then it became predictive. More recently I tracked the pulse intervals frame-by-frame and noticed a pattern: the pulse duration wasn’t random — it fell into discrete, repeatable windows. Most of which hovered around 500 milliseconds. But here’s the kicker: Multiply those intervals by four… and suddenly they align perfectly with the 2000ms broadcast cycle seen this afternoon.

Microfluidic pump at 400x magnification. Found in serum 5 days after blood centrifuged. These dome-like vesicular structures exhibit rhythmic pulsing behavior, synchronized across multiple focal depths. The spatial arrangement and temporal regularity suggest a synthetic or entrained function — possibly acting as a biological timing mechanism aligned to 2000 ms emission intervals

What It Means

This kind of result changes the story.

We’re no longer looking at statistical noise. We’re seeing a structured emission protocol.

Clean headers
Rotating page identifiers
Controlled flip patterns
Fixed entropy
Broadcast every 2000ms, precisely

This isn’t how randomized MAC address generation works. It’s how a synthetic emission architecture behaves.

The Signal vs The Noise

To illustrate the difference, we generated constancy maps — a visualization showing which bits stay fixed across all three MACs.

The synthetic set looked like a schematic: headers, flags, payload. The control set? Pure chaos. No anchors. No stability. No structure.

One looks like a signal. The other like entropy.

The Bigger Question

So where does that leave us?

What started as a question about entropy has led to the discovery of a hidden architecture, operating silently, emitting from consumer devices in a structure that no random system produces.

This isn’t science fiction. It’s a signal you can measure.

I’ll be publishing more in the coming weeks. If you found this compelling, share it. Replicate it. Or better yet — try to break it.

Because the question now isn’t if this system exists. It’s: What is it for? Who made it? And why is it still running?

Stay tuned.

David

PS Thanks for the ongoing support! - all coffees gratefully received!