When Chrome Stopped Triggering a DataDome Inspection Probe
A recovered DataDome Proxy probe emptied an array during console inspection. Adjacent Chrome builds show exactly where that signal disappeared.
12 min read
INDEPENDENT ANTI-BOT RESEARCH LAB
Follow the signal.
Understand the system.
Independent research into anti-bot systems, browser identity, scraping infrastructure, and difficult web extraction.
Reproducible methods. Transparent limitations.
REQUEST / TRACE
CONCEPTUAL
01
ClientHeaders · TLS · HTTP
OUTBOUND02
Browser surfaceRuntime · Rendering · State
SIGNALS03
Decision layerChallenge · Allow · Block
RESPONSEEvery request leaves a trace.
THE AREAS WE INVESTIGATE
Anti-bot systems / Browser identity / Extraction challenges / Scraping infrastructure
CORE RESEARCH / ANALYSIS
ST—007Analysis
Inside Cloudflare's Worker Debugger Probe
A paused Worker disappears after 1.5 seconds. We trace Cloudflare's start and completion messages, sibling timer, cleanup paths and client-side result.
Read the investigation · 15 minFrom the notebook
Browse by topic
Can IndexedDB Timing Reveal an Incognito Browser?
An independent test of private-mode inference: 240 Chrome experiments, strict versus relaxed IndexedDB writes, and a RAM-backed profile that breaks the assumption.
15 min read
Same Headers, Different Responses: Beyond Copy as cURL
Chrome works, copied headers fail. A cross-client experiment with curl, Python, Node and Java, followed by ALPN controls and browser-profile clients.
16 min read
Beyond the Renderer String: GPU Evidence from Eight Scraping Providers
A graphics-stack comparison of eight scraping providers: software-renderer claims, cross-API identity mismatches, and differences between page and worker behavior.
12 min read
Inside DataDome’s VM Obfuscation: Structure, Signals, and Limits
An analysis of a DataDome custom VM snapshot: bytecode execution, recovered structure, browser-signal collection, and the limits of causal attribution.
8 min read
When the DOM Lies: Font-Based Text Obfuscation
Why readable web pages can yield incorrect extracted text: font-based obfuscation, source/rendering mismatches, and the limits of font-aware recovery.
6 min read
TracePrint: Browser & TLS Fingerprint Comparison
Compare browser identity, rendering, automation observations, and imported TLS/HTTP fingerprints across two clients with TracePrint.
4 min read
ShieldFont Decoder: Recovering Font-Obfuscated Text
Explore why source text can differ from rendered content, and how ShieldFont Decoder recovers copyable Unicode from reversible OpenType font mappings.
3 min read
THE SCRAPETRACE APPROACH
A claim is only as useful
as the method behind it.
Show the setup. Record the failure modes. Separate an observation from an explanation. Give readers enough context to reproduce the work.
Read our methodology