1. Confidentiality: Encrypted data cannot be read without the user's password and Core Secret
2. Integrity: Tampered tokens are detected and rejected (GCM authentication)
3. Memory Safety: Sensitive data is protected from swap files and core dumps

• Weak passwords: Library cannot compensate for "password123"
• Core Secret compromise: If server environment is compromised, all data is at risk
• Client-side attacks: Keyloggers, screen capture, etc.
• Quantum attacks on KDF: Argon2id/PBKDF2 are not post-quantum secure (but AES-256 is)

DO NOT create a public GitHub issue for security vulnerabilities.

1. Email: Send details to security@flash512-vanguard.dev

   • Include: Vulnerability description, steps to reproduce, potential impact
   • PGP key available at PGP.txt

2. Acknowledgment: We will acknowledge receipt within 48 hours

3. Assessment: We will assess the vulnerability and provide a timeline within 7 days

4. Fix: We will develop and test a fix

5. Disclosure: We will coordinate public disclosure with you

• Buffer overflows or memory corruption
• Side-channel attacks (timing, cache, etc.)
• Weak random number generation
• Authentication bypass
• Token forgery
• Any deviation from documented security properties

• Theoretical attacks without proof of concept
• Attacks requiring physical access to the server
• Attacks on deprecated algorithms (we don't support any)
• Issues in third-party libraries (report to them directly)

# Generate a strong Core Secret
openssl rand -base64 64 > core_secret.txt
chmod 600 core_secret.txt

# Load into environment (do NOT commit to git!)
export FLASH512_VANGUARD_CORE=$(cat core_secret.txt)

Never:

• ❌ Hardcode the Core Secret in source code
• ❌ Commit it to version control
• ❌ Share it over insecure channels
• ❌ Use a weak/short Core Secret (< 64 chars)

Enforce strong passwords in your application:

• Minimum 12 characters
• Mix of uppercase, lowercase, numbers, symbols
• Check against breached password databases (e.g., Have I Been Pwned)

Always use secure_open() for sensitive data:

# ✅ Good: Memory is auto-wiped
with secure_open(token, password) as buffer:
    process(buffer.data)

# ❌ Bad: Plaintext stays in memory
plaintext = Flash512Vanguard.open(token, password)
process(plaintext)
# plaintext is still in memory until garbage collected

Monitor the audit logger for suspicious activity:

import logging

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("flash512.audit")

# Logs will show:
# DATA PROTECTED | KDF: Argon2id | 2026-06-29T...
# DATA OPENED | KDF: Argon2id | 2026-06-29T...
# OPEN FAILED: Decryption failed | 2026-06-29T...

Keep dependencies updated:

# Check for vulnerabilities
pip-audit

# Update dependencies
pip install --upgrade flash512-vanguard cryptography argon2-cffi

1. Python Strings are Immutable: Once you call .decode('utf-8') on bytes, the string cannot be wiped from memory. Use SecureBuffer to keep data as bytes.

2. No Forward Secrecy: If the Core Secret is compromised, all past tokens can be decrypted (if the attacker also has the user passwords).

3. No Post-Quantum KDF: Argon2id and PBKDF2 are not quantum-resistant. However, AES-256 is post-quantum secure.

4. mlock() May Fail: On systems with low ulimit -l, mlock() may fail silently. The library continues without memory locking (best-effort).

We welcome third-party security audits. If you conduct one, please share the results.

• General questions: GitHub Discussions