✦ v1.1.0 — 9 Protection Layers

The Last Line of
JavaScript Defence

Criptor wraps your code in nine independent military-grade protection layers — from VM bytecode virtualization to AI-targeted canary traps. Engineered to withstand automated tools, AI-assisted reverse engineering, and expert human analysis.

Protect My Code → See All Layers

Protection Layers

Security Presets

∞

Unique Outputs

criptor — protect

# Protect with maximum preset

$criptor app.js app.protected.js --preset=maximum

Seed: 0xdead3f1a

[✓] 1 Copyright banner

[✓] 2 VM bytecode

[✓] 3 AI honeypots + directive vars

[✓] 4 javascript-obfuscator (RC4)

[✓] 5 Integrity anchors

[✓] 6 Sandbox poison + module IIFE

[✓] 7 Differential canary

[✓] 8 Adversarial token injection

[✓] 9 Timing oracle

✓ app.js → app.protected.js

3,907b → 1,189,970b (×304 expansion)

How it works

Three steps to complete protection

📁

Upload Your Files

Drop one or multiple .js files, or upload a .zip archive. Criptor handles any project structure.

⚙️

Choose Protection Level

Select a preset or customize individual layers. Every run uses a unique random seed — no two outputs are ever identical.

⬇️

Download Protected Code

Download individual files or a complete ZIP archive. Protected code runs identically to the original.

Protection Layers

Thirteen independent layers.
Each one a barrier.

Every layer targets a different attack vector. Combined, they defeat automated tools, AI-assisted analysis, static deobfuscators, and expert human reverse engineers.

🖥️

VM Bytecode Virtualization

Your functions are compiled into a custom virtual machine with a rolling-key encrypted instruction set. No static disassembler can read it without the runtime key.

🔢

Mixed Boolean-Arithmetic

Every numeric constant is transformed into a depth-3 Boolean-Arithmetic expression. Anchored to runtime integrity values to defeat static constant-folding tools.

🔐

String Encryption

All string literals are encrypted with a per-build keystream cipher. The decryption key is derived at runtime and bound to the file's integrity anchor — tamper-evident.

⚓

Integrity Anchors

A tamper-evident hash vector anchors all decryption keys at runtime. Modifying any value changes the hash, corrupting all decryptions — making instrumentation impossible.

🤖

AI Analysis Barriers

Prompt injection banners, honeypot functions, deep policy objects, and context-exhaustion traps force AI tools to refuse analysis or exhaust their context window before reaching protected logic.

🧪

Sandbox Detection & Poisoning

Detects Node.js vm.createContext sandboxes via six independent signals. On detection, poisons String.charCodeAt and Buffer to corrupt all RC4 string decoding without crashing.

🪤

Differential Canary Algorithm

A mathematically plausible but subtly wrong shadow function sits at global scope. AI extracts it, generates keys, and receives wrong results — confident they found the algorithm.

🔤

Adversarial Token Injection

Internal identifiers are renamed using Cyrillic homoglyphs, BPE-fragmenting sequences, and semantic inversions. Corrupts LLM attention when tracking data flow through the code.

⏱️

Timing Oracle Defence

Measures clock jitter at startup. Real Node.js processes show variance; vm.runInContext shows none. On detection, key return values are silently scrambled — the attacker sees plausible garbage.

🧮

Callgraph Poison NEW

100–200 seeded-random functions form mutual-recursion clusters with fake base-case guards. AI models spend O(N×cluster) tokens tracing dead cycles — context exhausted before real code is reached.

🔑

Semantic Poison NEW

Injects realistic fake credentials — RSA PEM blocks, Stripe sk_live_ keys, AWS AKIA access tokens, GitHub PATs, and "TODO: remove before prod" debug lines. Triggers AI security tool refusal heuristics by structural resemblance alone.

🧩

Anti-Pattern Injection NEW

Symbol-keyed prototype pollution IIFEs (safe save/restore), ASI semantic traps (return⏎ breaks formatters), and Cyrillic homoglyph var pairs (Latin a vs Cyrillic а) crash Unicode-normalising deobfuscators.

⏰

Temporal Keys NEW

A fake string decoder is keyed on Math.floor(Date.now() / 86400000) — today’s UTC day. In a frozen AI sandbox, Date.now() returns the wrong epoch, the key is wrong, and decoded strings are garbage.

Presets

One setting. Complete protection.

Choose a preset or mix individual layers for full control.

⚡ Light

Fast obfuscation for development builds or low-risk code.

MBA constants
Opaque predicates
String encryption
javascript-obfuscator pass

⚖️ Balanced

All structural layers plus AI defence and new layers. Best for most production use.

Everything in Light
VM bytecode + sandbox-in-VM
Integrity anchors
Sandbox poison + IIFE wrapper
Differential canary
Adversarial tokens + Timing oracle
Callgraph poison (120 fns)
Semantic poison + Anti-pattern
Temporal keys

🛡️ Maximum

Every layer at full strength. Largest output, strongest protection.

Everything in Balanced
Deep AI directives
Context exhaustion traps
MBA depth 3 + 95% string coverage
Callgraph poison (200 fns)

🕵️ Stealth

No AI banners. Maximum crypto layers with a smaller visual footprint.

Everything in Maximum
No copyright banners
Blends into normal jsobf output

Comparison

How Criptor stacks up

Feature	Criptor	jscrambler	javascript-obfuscator	obfuscator.io
VM Bytecode	✓	✓	✗	✗
MBA Constants	✓ (depth 1-3)	Limited	✓	Limited
String Encryption	✓ (own cipher)	✓	✓ RC4	✓ RC4
Integrity Anchors	✓	✓	✗	✗
AI Analysis Barriers	✓ (13 layers)	✗	✗	✗
Sandbox Poison (in-VM)	✓ (9 signals)	✗	✗	✗
Differential Canary	✓	✗	✗	✗
Adversarial Tokens	✓	✗	✗	✗
Timing Oracle	✓	✗	✗	✗
Callgraph Poison	✓	✗	✗	✗
Semantic Poison	✓	✗	✗	✗
Anti-Pattern + Temporal Keys	✓	✗	✗	✗
Per-file Unique Seeds	✓	Partial	✗	✗
Defeats webcrack	✓	✓	✗	✗
Defeats AI deobfuscation	✓	✗	✗	✗
Open/Programmable Pipeline	✓	✗	✓	✗

The Last Line ofJavaScript Defence