Data Security & Privacy
The "Vault". Moving beyond compliance checkboxes to applied engineering: Cryptography, Key Management, and Privacy Computing.
The Data Security Lifecycle (DSMM)
Security controls must adapt as data changes state. Encryption at Rest is useless if the data is stolen during Processing.
Collection Phase
Critical security controls must be applied when data is in the Collection state.
Key Challenge
Balancing business desire for data vs. legal minimization requirements.
Controls (Must Have)
- Privacy Notice (PIPL)
- Minimal Collection
- Data Classification
Top Risks
- Unauthorized Collection
- Hidden Trackers
Privacy Computing & The Future
MPC
Multi-Party ComputationMultiple parties compute a function over their inputs while keeping those inputs private.
Use Case: Two banks finding common fraudulent customers without sharing customer lists.
TEE
Trusted Execution EnvironmentHardware-based isolation (Intel SGX, ARM TrustZone) ensuring code and data loaded inside are protected from the OS.
Use Case: Secure biometric matching on mobile devices.
FHE
Fully Homomorphic EncryptionThe "Holy Grail". Processing data while it remains encrypted.
Use Case: Outsourcing DNA analysis to the cloud without the cloud provider ever seeing the DNA data.
The Crypto "Hall of Shame"
Cryptography is binary: it's either correct or broken. There is no "mostly secure" encryption. These are the most common implementation errors seen in the wild.
Fixed IVs
Using the same Initialization Vector for every encryption. Breaks AES-GCM completely.
ECB Mode
Using Electronic Codebook mode. Patterns in plaintext remain visible in ciphertext (The Tux Penguin).
Hardcoded Keys
Embedding 'secret_key' in the source code or git repo. 'Security by Hope'.
No Integrity Check
Using AES-CBC without an HMAC. Vulnerable to Padding Oracle attacks.
Weak Randomness
Using `Math.random()` for key generation instead of a CSPRNG.
Homebrew Crypto
Inventing your own 'simple XOR' algorithm. Don't roll your own crypto.