What is the SQL Tamper Framework? A context-aware SQL transformation framework for WAF bypass, plus SQLMap-compatible tamper scripts, that tokenizes SQL and transforms it deterministically while keeping it valid. Published on PyPI as sqlmap-tamper-framework.

How is it different from a normal SQLMap tamper script? Normal tamper scripts do string replacement and break on multi-character operators, shifting positions, and clause context. This framework lexes SQL into UUID-tracked tokens and tracks clause context, so output stays valid SQL.

How do I use it with SQLMap? Copy a script into SQLMap’s tamper directory (cp tamper_scripts/cloudflare2025.py /path/to/sqlmap/tamper/) and run sqlmap -u "https://target.com?id=1" --tamper=cloudflare2025.

Which WAFs are targeted? Cloudflare (cloudflare2025.py), AWS WAF v2 (awswaf2026.py), Azure WAF (azurewaf2026.py), ModSecurity CRS (modsec_crs2026.py), Imperva (imperva2026.py), Akamai Kona (akamai2026.py), plus an auto-select combiner (meta_tamper.py).

What database does it target? MySQL 5.7+ and MariaDB 10.x. It may not work against PostgreSQL, MSSQL, or Oracle, since techniques like /*!50000...*/ version comments are MySQL-specific.

What transformations does it include? Core: keyword wrapping, space replacement, case alternation, operator URL encoding. Advanced (v2.1.0+): homoglyphs, function wrapping, numeric obfuscation, comment chaos, logical-operator swap, hex encoding, and version-comment variation.

Why is the output deterministic instead of randomized? By design: deterministic output is reproducible for research and verification, and the absence of a randomization engine is a deliberate abuse-resistance choice.

What is the multi-character operator fix? A naive lexer breaks >= into > and =, encoding to the broken %3E=. The lexer checks multi-character operators first, so >= is one token and encodes atomically to %3E%3D.

What is UUID token tracking and why does it matter? Each token gets a UUID at lex time that never changes, so transformations that alter a token’s length (like wrapping SELECT) do not desynchronize tracking the way position indices would.

What does “context-aware” mean here? The framework tracks which SQL clause a token is in, so a rule can, for example, URL-encode operators only in WHERE/HAVING and leave the SELECT list untouched.

Can I write my own transformation? Yes. Create a TransformationRule with a transform_func(token, context), target_types (e.g. TokenType.OPERATOR), and allowed_clauses (e.g. [ClauseType.WHERE]), and add it to a SQLTransformer.

When should I use AST-based instead of token-based transformation? Use token-based (faster, simpler) for most queries; use AST-based (handles nesting and function calls) for complex or deeply nested queries.

How fast is it? Repository-reported benchmarks: ~1ms for a simple 10-token query, ~5ms for a 100-token query, ~10ms for a nested subquery.

How is it tested? The docs enumerate 33 tests (10 lexer, 10 transformer, 13 integration) covering operators, literals, comments, UUID tracking, context awareness, determinism, and real SQLMap payloads. (Note: the README badge says “49+ passing”; the itemized breakdown totals 33.)

How do I install it? git clone the repo and either copy a tamper script into SQLMap, or pip install -e . for framework/development use. It’s also on PyPI as sqlmap-tamper-framework.

What Python version is required? The README/PyPI badge says Python 3.8+; the PyPI metadata field says >=3.7. (The two disagree slightly.)

Does it guarantee a WAF bypass? No. Effectiveness varies by WAF configuration; there is no universal bypass guarantee. A bypass is always relative to a specific ruleset.

What is meta_tamper.py? An auto-select combiner that chains transformations based on environment variables, letting you drive the evasion chain from configuration rather than editing code.

What are the known limitations? MySQL/MariaDB focus, simplified (not full) SQL parsing with possible edge cases on deeply nested queries, and WAF-dependent effectiveness.

Is this legal to use? Only for authorized testing: systems you own, with written authorization, in authorized pentests or in-scope bug bounties, or against local vulnerable apps (DVWA, bWAPP, SQLi-labs). Unauthorized use is illegal under the CFAA, the UK Computer Misuse Act, India’s IT Act 2000 Section 66, and similar laws.