Prompt Injection: The Trust-Boundary Bug in LLM Applications
How direct and indirect prompt injection cross trust boundaries, and how to limit what a manipulated model can expose or do.
Security work, written down.
I'm Sudhir, a CISSP-certified Cyber Security Professional working on enterprise security, governance, compliance, and validation. I write practical notes on how security controls are designed, assessed, operated, and evidenced in large enterprise environments.
Security architecture · Essay
Why Theoretical Security Fails in Production
How to choose security controls that reduce risk without driving users toward workarounds.
A control can be secure on paper and still make the system less safe.
Force people to create a new complex password every month and some will write it down. Block every file-sharing service without providing an approved alternative and project data will move through personal accounts. Add six approval steps to emergency access and engineers will look for a permanent shortcut.
"The control has failed to account for the work surrounding it."
Selected writing
Security Architecture: Principles for Designing Secure Systems and Networks
How defense in depth, identity, segmentation, threat modeling, and recovery fit into a security architecture.
More from the writing
All 12 posts- Passing the CISSP at 100 Questions Certification 6 min read
- Cisco IOS XE Web UI Vulnerabilities: How to Check Your Exposure Vulnerability management 4 min read
- iHack CTF 2025: Three Challenges I Solved Offense 3 min read
- Zero Trust: Why the Perimeter Model Is Dead Security architecture 5 min read
- Is Zscaler the Right Move for a Large IT Company? Network security 7 min read
- Firewall Compliance Without the Checkbox Theatre Compliance 6 min read
6 posts
Network security
Firewall policy, segmentation, and hardening at scale.
Latest Cisco IOS XE Web UI Vulnerabilities: How to Check Your Exposure
5 posts
Security architecture
Layering, blast radius, and designing for the exception.
Latest Why Theoretical Security Fails in Production
2 posts
Offense
CTFs, red teaming, and offensive security.
Latest Prompt Injection: The Trust-Boundary Bug in LLM Applications
2 posts
AI security
Prompt injection, model risk, and securing LLM systems.
Latest Prompt Injection: The Trust-Boundary Bug in LLM Applications
1 post
Compliance
Controls people actually follow, and audits you can live with.
Latest Firewall Compliance Without the Checkbox Theatre
1 post
Vulnerability management
Triage, prioritisation, and reading a report like an attacker.
Latest Cisco IOS XE Web UI Vulnerabilities: How to Check Your Exposure
1 post
Certification
Exam preparation, study notes, and what the paper is worth.
Latest Passing the CISSP at 100 Questions
I'm a Cyber Security Professional. I write about how security decisions play out in practice. More about me