Data Security: Trust No One

Group 3: Safely Using Data Zero Trust Architecture Let's do that. Let's make a new directory called colab v2 and move the older colabs into the colab v1 directory in the colabs directory. That way we have the prior stuff and all the new stuff we're going to make in this. Ideally all those would end up having the same kind of structure. Does that make sense?

The Zero Trust Reality: Whether it's a sprawling ELT pipeline or a modest Postgres setup, the risk is the same: one slip—a password, a PII column (personally identifiable information like SSNs, names, IP addresses) column—and your company could be history. The network? Unreliable. The client? Suspect.

The Big 3: Data Security Disasters That Exposed America

The Big 3: Data Security Disasters That Exposed America When trust failed, millions paid the price 1. EQUIFAX (2017) The Identity Theft Catastrophe 148M Americans' SSNs exposed What they stole: • Social Security Numbers • Birth dates, addresses • Driver's license numbers The damage: • Data still sold on dark web • Victims got ~$125 each • Identity theft for decades 2. NATIONAL PUBLIC DATA (2024) The Data Broker Disaster 170M People across US/UK/Canada What they stole: • 2.9 billion records total • SSNs, phone numbers • Current & past addresses The damage: • Company went bankrupt • No help for victims • Half of US population hit 3. CHANGE HEALTHCARE (2024) The Medical Records Nightmare 193M Americans' medical data What they stole: • Medical records, SSNs • Diagnoses, prescriptions • Insurance information The damage: • Paid $22M, data still leaked • Pharmacies couldn't operate • Medical fraud risk forever 490% increase from 2023→2024 3-4x per year your data is exposed $4.88M average breach cost (IBM 2024) Your Database Could Be Next Unless you implement Zero Trust Security — where nothing and no one is trusted by default

Zero Trust, Always Verify: The 5 Attack Vectors

The Uncomfortable Truth

These 5 things WILL betray you. Plan accordingly.

Don't Trust the Network

Assume: Already compromised

• Man-in-the-middle attacks
• Packet sniffing
• DNS poisoning

Don't Trust Users

Assume: Already phished

• Credential theft
• Social engineering
• Malware on devices

Don't Trust Admins

Assume: Insider threat

• Disgruntled employees
• Compromised accounts
• Privilege abuse

Don't Trust Code

Assume: Injection attempts

• SQL injection
• Supply chain attacks
• Malicious libraries

Don't Trust Backups

Assume: Will be deleted

• Ransomware targets
• Corrupted restores
• Untested procedures

Zero Trust DB Checklist (Starter)

PriorityTimeZero Trust Action
CRITICAL5 minSecure Password Management
• Change default passwords
• Set password expiry
• Enforce strong passwords
CRITICAL1 hourSecure Network Configuration
• Enable SSL/TLS encryption
• Limit max connections
• Restrict connection sources
CRITICAL10 minLeast Privilege Access
• Read-only for analytics/data science
• Keep UPDATE/DELETE
HIGH5 minEnable audit logging
HIGH45 minComplete 3-2-1 Backup Strategy
• 3 copies: Primary + Secondary + Offsite
• 2 storage types: Local disk + Cloud (S3)
• 1 offsite: Different region/location
• Test restores: Verify backups actually work
• Immutable backups: Protect from ransomware

Zero Trust DB App Checklist (Starter)

Example: SQL Injection Defense

Demonstrates: Verify Everything (validate all inputs) • Layer Defenses (parameterized queries)

The Bobby Tables Attack (Still Works in 2025!) How a simple quote mark can destroy your database VULNERABLE CODE (What NOT to do) # Python - String concatenation (BAD!) def get_user (username): query = f"SELECT * FROM users WHERE name = ' {username} '" # User input directly in query! cursor.execute(query) return cursor.fetchall() Attacker Input: Robert'; DROP TABLE users; -- Resulting Query: SELECT * FROM users WHERE name = ' Robert'; DROP TABLE users; --' ↑ Query ends ↑ New command! ↑ Comment SECURE CODE (Always use prepared statements) # Python - Prepared statement (GOOD!) def get_user (username): query = "SELECT * FROM users WHERE name = %s" # Placeholder, not string! cursor.execute(query, (username,) ) return cursor.fetchall() Same Attack Input: Robert'; DROP TABLE users; -- Resulting Query: SELECT * FROM users WHERE name = 'Robert''; DROP TABLE users; --' ↑ Entire string treated as data, not code!