Statistical coherence

The most common weakness in anti-detect browsers is not individual parameter values — it is the combination. A random fingerprint generator can produce a profile that claims to be a Windows 10 laptop with an Apple Silicon GPU, a 4K screen with a display scale of 1, and a font list that has never appeared on any real device. Detection systems do not need to find a single suspicious value; they look for combinations that have never been seen in the wild.

Maskifi's fingerprint generation is statistically coherent: attributes are sampled together from distributions that reflect how those attributes co-occur on real devices, so the resulting combination is plausible as a whole, not just plausible in each field in isolation.

Why random values fail

Imagine assigning each fingerprint attribute independently at random:

• Pick a random OS from {Windows, macOS, Linux}.
• Pick a random GPU from a list of all known GPUs.
• Pick a random screen resolution from all observed resolutions.
• Pick a random set of fonts.

The result can easily produce a macOS browser reporting a GPU that has never shipped in a Mac, or an entry-level CPU core count paired with a high-end 4K screen. Each individual value looks normal; the combination does not. Platforms that track fingerprint distributions can score the likelihood of a combination and flag anything that falls outside observed ranges.

How coherent generation works

When Maskifi generates a fingerprint, it accepts optional constraints (OS family, screen size preference, locale) and produces a complete fingerprint that satisfies them.

The key properties of the generated fingerprint:

• OS consistency: The browser identity string, platform, operating system, request headers, and available fonts all agree on the same operating system and version.
• GPU plausibility: The graphics vendor and renderer, together with supported features, reflect a GPU model that is actually shipped with or commonly used alongside the chosen OS.
• Screen coherence: Resolution, display scale, color depth, and window dimensions are internally consistent — available height is less than screen height by a taskbar margin; inner and outer window sizes account for browser chrome.
• Locale alignment: Language preferences and the corresponding request header all express the same locale.
• Battery state: Charging state, level, and timing fields are consistent with each other.
• Audio: Sample rate, output latency, and channel count reflect values typical for the chosen platform.

The attributes in a generated fingerprint are sampled together, so the resulting combination matches combinations observed on real devices rather than being an arbitrary mix of individually valid values.

Constraints you can set

When requesting a new fingerprint (via Regenerate in the editor, or when creating a profile), you can optionally pin:

Constraint	Effect
Operating system	Pins the OS: Windows, macOS, or Linux.
Screen size range	Bounds the minimum and maximum screen width and height.
Locale	Pins the language, aligning all language-related fields.

Any constraint not specified is filled in coherently around the ones that are.

What coherence does not cover

Coherence is about plausible static attributes. It does not make a profile behave like a human:

• Behavioral signals (mouse movement patterns, typing cadence, scroll rhythm) are a separate layer. Maskifi's humanization option, when enabled for a browser launch, applies human-like delays and cursor movement to automation actions.
• History signals (cookie age, browsing history length) depend on how the profile is used over time, not on the initial fingerprint.
• Canvas and audio noise (the micro-differences in pixel and audio output that form per-device hashes) are handled by the Maskifox engine at the rendering layer.

Coherence ensures that a fresh profile cannot be immediately disqualified by a statistical check. Sustained believability requires using the profile in a natural way over time.

Comparison with random spoofing

Approach	Single-value plausibility	Cross-attribute consistency	Detection resistance
No spoofing (real device)	Excellent	Excellent	Poor (repeated)
Random per-attribute	Good	Poor	Low
Maskifi coherent generation	Good	Good	High

The core insight is that detection systems have access to large fingerprint databases. Any combination of attributes that does not appear in those databases — or appears rarely and only on accounts that later turned out to be fraudulent — is a risk signal. Generating combinations that are statistically normal eliminates this class of signal.