Privacy you can measure. The anti-surveillance platform.
This document presents the Profile Blur Score methodology, platform architecture, and claim verification paths for dcoy. Every claim is backed by an external validator (EFF Cover Your Tracks, Optery) or an internal artifact (database log, dashboard event). Readers are encouraged to verify all claims using the free public tools cited throughout.
Surveillance companies aggregate every signal they can find about a person and convert that raw data into actionable intelligence profiles. Their clients are corporations, governments, political campaigns, and advertisers. The people being profiled have no visibility and no recourse.
dcoy is built in direct opposition to that infrastructure, running the same logic in reverse, on behalf of the individual.
| Surveillance industry | dcoy | |
|---|---|---|
| Approach | Aggregate signals, build profile, target | Corrupt signals, degrade profile, protect |
| Effect | Makes individuals legible to institutions | Makes individuals illegible to institutions |
| Client | Corporations, governments, campaigns | You |
| Scale effect | Intelligence improves with more data | Protection improves with more subscribers |
| Accountability | No visibility, no recourse | Monthly score proves what is working |
dcoy is a cross-platform consumer privacy platform that actively degrades commercial surveillance profiles across the devices a subscriber uses it on. Unlike tools that attempt to block or hide activity, dcoy corrupts the data surveillance systems already have, making it statistically unreliable for targeting, pricing, and political manipulation.
The Profile Blur Score (PBS) is the mechanism by which dcoy proves this is working. It is a 0 to 100 index that measures how well dcoy's three automated, independently verifiable protection layers are performing: behavioral noise injection (up to 40 points), tracker SDK blocking (up to 35), and browser fingerprint noise injection on Chrome (up to 25; canvas, WebGL, and audio confirmed randomized on EFF Cover Your Tracks). Data broker opt-out is an included assisted tool and does not contribute to the score, because completion depends on the subscriber filing the requests. Location signal noise is in design and contributes zero. Mobile advertising ID rotation is out of scope pending a user-guided flow design.
The following table summarises the status of each protection vector. Every "Live" status is a scored PBS layer backed by either an external validator (EFF) or a Tony-run dashboard log artifact. "Assisted" means dcoy provides the tool but the subscriber completes the action, so it is not scored.
| Status | Vector | PBS contribution |
|---|---|---|
| Live | Behavioral noise injection (up to 40 pts) | Score computation live in production since April 2026. Generates contradictory interest signals against a weekly delivery target. The signature layer; no commodity privacy product ships this at consumer scale. |
| Live | Tracker SDK blocking (up to 35 pts) | An extensive blocklist of tracker SDK domains. Block log visible in real time in dashboard. |
| Live | Browser fingerprint noise injection (up to 25 pts) | Page-level interception running in production on Chrome. Canvas, WebGL, and audio fingerprints confirmed randomized on EFF Cover Your Tracks, defeating cross-site fingerprint tracking. Does not make the browser non-unique in a single snapshot. |
| Assisted | Data broker opt-out (not scored) | Included assisted tool. dcoy prepares, routes, and tracks CCPA/GDPR opt-out requests; the subscriber files them. Not scored, because completion depends on the subscriber. Verify any broker yourself via Optery. |
| In design | Location signal noise (not scored) | Permission audit and IP variance components in design. Contributes zero to the PBS pending implementation. |
PBS ranges by tier, three live layers contributing out of 100: Personal at 90 days: 72 to 88. Ghost at 90 days: 72 to 88 (weekly audit cadence; the scored layers are the same, so the range is the same). Blackout: not yet shipping; ranges to be published when the tier is live. Browser fingerprint noise injection randomizes the canvas, WebGL, and audio surfaces on Chrome, confirmed on EFF Cover Your Tracks. Behavioral noise injection is live in production. Data broker opt-out is an included assisted tool and is not scored. Location signal noise is in design.
The consumer privacy market is populated by tools that each solve one piece of the problem, charge a premium for it, and provide no way to verify the result. The following comparison reflects three live dcoy components against the competition's complete feature sets.
| Feature | DeleteMe | Blur | Privacy Bee | NordVPN | dcoy |
|---|---|---|---|---|---|
| Behavioral noise injection | ✗ | ✗ | ✗ | ✗ | ✓ live |
| Tracker SDK blocking | ✗ | ✗ | ✗ | limited | ✓ |
| Browser fingerprint noise injection | ✗ | ✗ | ✗ | ✗ | ✓ live |
| Monthly proof / audit score | ✗ | ✗ | ✗ | ✗ | ✓ Profile Blur Score |
Competitor prices reflect publicly listed annual rates as of May 2026. dcoy: Personal $108/year, Ghost $228/year. Blackout coming soon.
A commercial surveillance profile is the aggregate of data points that surveillance pricing intermediaries, data brokers, and behavioral advertising platforms maintain about an individual. The FTC's January 2025 surveillance pricing study found that intermediary firms worked with at least 250 client businesses (grocers, apparel, health and beauty, home goods, convenience, and hardware retailers among them) using personal data to set targeted prices. Sources contributing to those profiles include:
The existing consumer privacy toolkit was designed for a desktop-first threat model built in the early 2010s. Modern surveillance infrastructure has evolved beyond it:
dcoy attacks the surveillance profile through several coordinated layers. Three are automated, verifiable, and contribute to the headline PBS: behavioral noise injection, tracker blocking, and browser fingerprint noise. A fourth, data broker opt-out, is an included assisted tool that the subscriber completes and is not scored. A measurement layer audits all of it.
On desktop, the Chrome extension performs page-level interception of canvas, WebGL, audio context, font enumeration, navigator properties, and WebRTC signals via main-world script injection. The canvas, WebGL, and audio fingerprint surfaces all read as randomized on EFF Cover Your Tracks, which defeats the cross-site fingerprint tracking these surfaces enable. dcoy randomizes these surfaces rather than forcing them to a uniform value, so it does not make the browser non-unique in a single EFF snapshot; the residual entropy lives in surfaces dcoy deliberately leaves alone to avoid breaking sites, such as the system font list and timezone. Firefox is on the roadmap.
Mobile device-level identity noise is on the roadmap. Programmatic rotation of the Google Advertising ID is not possible from a third-party app; only the user (via device Settings) or the OS can reset it. A user-initiated rotation guidance flow with PBS attribution on confirmation is in design. An iOS App Tracking Transparency audit would form part of an iOS build, which is not committed.
The Chrome extension is live on the Chrome Web Store, reporting tracker and fingerprint blocks to the dcoy API every 5 minutes. Firefox port is on the roadmap. The Android app is live on the Google Play Store. iOS is scaffold only and not committed; a full NetworkExtension and App Groups integration would be substantial development work.
On Android, dcoy runs as a VpnService intercepting all device traffic and filtering it against a bundled blocklist of tracker SDK domains. Filtering uses DNS-over-HTTPS with a local resolver. No traffic is routed through dcoy servers. No root access is required.
On desktop, the Chrome extension blocks the same category of domains at the request level. iOS NetworkExtension filtering would form part of an iOS build, which is not committed. The current Android blocklist is a curated set focused on high-frequency consumer tracker SDKs; expansion via EasyPrivacy and Disconnect.me sourcing is on the roadmap.
dcoy prepares CCPA deletion requests and GDPR erasure requests for a regularly updated list of consumer-facing data brokers, routes them to the subscriber, and tracks their status. The subscriber files the requests. Automated headless submission is not feasible at consumer scale: broker sites defend their forms with CAPTCHAs, interstitials, and SSO walls, and empirical testing across brokers returned zero accepted headless submissions. The assisted model is the architecture, not a fallback. Because completion depends on the subscriber, this layer does not contribute to the PBS.
Broker coverage is tiered: Personal covers a list of 30 data brokers and people-search sites, Ghost an expanded list of 55 adding long-tail people-search and public-records sites. For each, dcoy prepares the opt-out request and routes it to the subscriber to file, then tracks completion in the dashboard checklist. There is no automated submission step; the empirical recon that drove this design found no broker accepting headless form submissions. The tiered coverage and the checklist are live in production.
The behavioral noise component generates human-realistic browsing sessions whose interest categories are semantically inverse to the subscriber's real profile. The goal is to introduce enough contradictory signal that the ad platform's interest graph loses coherence, reducing its accuracy for targeting and pricing.
The architecture is as follows: a profile inversion engine computes the semantic opposite of the subscriber's Google Ad Center interest categories. A session simulator (Playwright-based headless browser) generates browsing sessions through residential proxy infrastructure, with variable scroll depth, dwell time, and interaction patterns designed to pass platform bot detection. A feedback loop monitors interest graph coherence to determine whether sessions are being counted or filtered.
The current PBS contribution measures signals delivered against weekly target rate. Interest-graph coherence delta against a baseline is the target methodology and an ongoing engineering refinement. The injection architecture is live in production, no commodity privacy product ships this at consumer scale.
Score computation is live in production since April 2026. Sessions and signals persist to the noise_injections table on every agent cycle. PBS contribution is non-zero and folded into the headline ranges.
Why we publish methodology: The measurement layer is the product. The way we calculate the behavioral contribution is open for inspection. The refinement path from signal-delivery to interest-graph coherence delta is engineering work, not a gating credibility question.
Regularly, dcoy runs a full profile audit: checking Google My Ad Center and Meta Ad Preferences for assigned interest categories, analysing tracker block log data, and scoring browser fingerprint randomization confirmed via EFF Cover Your Tracks. The three scored components feed directly into the PBS calculation and are displayed in the subscriber dashboard with source links. The assisted broker opt-out checklist is tracked alongside but is not part of the score.
The three scored components contribute their full calculated value. Assisted broker opt-out and in-design location signal noise are not scored.
The vectors below reflect dcoy's own synthesis of which surveillance signals matter most, informed by the FTC 6(b) Surveillance Pricing Study (January 2025), Princeton Web Transparency and Accountability Project research, EFF Cover Your Tracks, Dubé and Misra "Personalized Pricing and Consumer Welfare" (Journal of Political Economy, 2023), and Privacy Rights Clearinghouse data broker research. The Profile Blur Score is built from the three of these that dcoy can run automatically and verify independently, scaled so the three sum to 100. The remaining vectors are handled by an assisted tool or are in design, and are not scored. The underlying sources informed but did not directly produce these figures.
| Vector | PBS weight | Platform | Status | Method |
|---|---|---|---|---|
| Behavioral noise injection | 40 pts | All | Live | Cloud agent; signals delivered against weekly target |
| Tracker SDK interception | 35 pts | All | Live | Extension and on-device DNS domain blocking |
| Browser fingerprint | 25 pts | Desktop | Live | Main-world JS noise injection; canvas, WebGL, audio randomized on EFF |
| Data broker opt-out | not scored | All | Assisted | Prepared, routed, and tracked; subscriber files |
| Location data | not scored | All | In design | Permission audit + IP variance |
The Profile Blur Score (PBS) is a 0 to 100 index expressing how well dcoy's three automated, independently verifiable protection layers are performing for a subscriber. A PBS of 0 means none of the three layers is contributing; a PBS of 100 means all three are at their maximum measured value. The score does not claim that total surveillance has been neutralized; it reports the measured performance of the three layers dcoy can run automatically and prove. Assisted broker opt-out, location signal noise, and mobile advertising ID are real parts of the surveillance landscape that fall outside the score.
The three scored components contribute their calculated value to the headline PBS. Assisted broker opt-out and in-design location signal noise are not scored.
Three components contribute to the headline PBS: behavioral noise injection (40 pts max), tracker SDK blocking (35 pts max), and browser fingerprint noise injection on Chrome (25 pts max). The three maximums sum to 100, which is the ceiling shown in the subscriber dashboard. Data broker opt-out, location signal noise, and mobile advertising ID are not part of the score.
Running in production since April 2026. The score reflects signals delivered against a weekly target rate. The interest-graph coherence delta formula is the target methodology and an ongoing engineering refinement. This is the signature layer and carries the largest weight in the score.
Scaled on the count of known tracker SDK domain requests blocked over the prior 30-day period, drawn from the extension and on-device DNS logs. A logarithmic curve rewards early blocking heavily and reaches the maximum at sustained high block volume.
Running in production on Chrome via main-world JS injection. The canvas, WebGL, and audio fingerprint surfaces all read as randomized on EFF Cover Your Tracks, which defeats the cross-site fingerprint tracking these surfaces enable. dcoy randomizes these surfaces rather than forcing them to a uniform value, so it does not make the browser non-unique in a single EFF snapshot.
In design. The permission audit and IP variance components described elsewhere are the target architecture. This vector contributes zero to the PBS pending implementation and is not part of the ceiling.
An included assisted tool, covered in the architecture section. dcoy prepares, routes, and tracks opt-out requests; the subscriber files them. Because completion depends on the subscriber, it is not scored and does not appear in the PBS formula.
The following ranges reflect the three live, scored protection layers contributing to the headline PBS at Day 90, out of 100. The scored layers are the same on Personal and Ghost, so the range is the same on both. Assisted broker opt-out and in-design location signal noise are not scored.
| Tier | 90-day PBS | Primary driver | Audit cadence |
|---|---|---|---|
| Personal · $9/mo | 72 to 88 | Three scored layers | Monthly |
| Ghost · $19/mo | 72 to 88 | Same three scored layers; assisted broker coverage expanded to 55 sites (not scored) | Weekly |
| Blackout · coming soon | To be published when tier ships | - | - |
Every scored PBS component is verifiable from outside dcoy's systems using free public tools: behavioral noise injection via the dashboard event log, tracker blocking via the dashboard block log, and browser fingerprint via EFF Cover Your Tracks (canvas, WebGL, and audio confirmed randomized). Data broker presence is independently checkable via Optery, though broker opt-out is an assisted tool and is not part of the score.
dcoy is actively seeking credentialed journalists, academic researchers, and regulators to run controlled before-and-after audits of the behavioral noise component. We will provide free Ghost-tier access for the duration of the study. Results will be published in full: positive, negative, or inconclusive.
Yes. Every action dcoy takes operates on the user's own device, within their own accounts, and exercises rights they already hold:
PBS ranges published in dcoy's marketing (72 to 88 at Day 90, the same on Personal and Ghost because the scored layers are identical) reflect the three live, scored protection layers: behavioral noise injection (up to 40), tracker SDK blocking (up to 35), and browser fingerprint noise injection on Chrome (up to 25; canvas, WebGL, and audio confirmed randomized on EFF Cover Your Tracks). Data broker opt-out is an included assisted tool and is not scored. Location signal noise is in design and contributes zero. Mobile advertising ID rotation is not currently shipped and is excluded from PBS.
dcoy does not claim to eliminate surveillance, defeat all tracking, or guarantee any specific outcome. The PBS measures what is measurable. It discloses what is not yet proven. That honesty is the product.
The surveillance industry built a machine to make individuals legible to institutions. dcoy runs the same logic in the opposite direction, making individuals illegible, continuously. Three components are live, automated, and contributing to the headline PBS today: behavioral noise injection, tracker SDK blocking, and browser fingerprint noise injection on Chrome (canvas, WebGL, and audio confirmed randomized on EFF Cover Your Tracks). Data broker opt-out is an included assisted tool and is not scored. Location signal noise is in design. Mobile advertising ID rotation is excluded from PBS until a user-guided flow ships. The roadmap is clear.
The Profile Blur Score is how we prove it. Not with assertions. With numbers. With auditable formulas. With verification tools anyone can run. With a dashboard that shows the before and the after, every month, linked to the public sources that feed it. Every claim has a verification path, and the methodology is published in full.
Every other privacy product asks you to trust them. We built the score so you don't have to.
The methodology is public. The architecture is described in full above. We invite independent replication.