Skip to content
AI Security

Lesson 5 of 13 · 6 min read

The Inference Stack as Attack Surface

An LLM inference service is a stack of components that all handle untrusted input: an API gateway, authentication and quotas, request and prompt processing, tokenizers, the model server, GPU drivers, caches, logging, and the management plane. Each layer has classic risks such as parsing bugs, exposed dashboards and broken tenancy, plus AI-specific ones: cost exhaustion, model extraction through the API, prompt leakage through shared caches, and data retained in logs. Threat-modeling the stack layer by layer turns 'secure the model' into concrete controls.

Premium lessonReviewed By the AISecurityInterviews editorial team
Get full Premium access · $25

Every worked answer in all 17 topics, every lesson in every course, the full concept curriculum, the complete practice-test bank, and the downloadable security field guides.

6 months · One payment · No auto-renewal

Practise it as an interviewer asks it

Cloud, Container & AI Infrastructure SecurityContainers, gVisor, microVMs and VMs: where is each isolation boundary?Cloud, Container & AI Infrastructure SecurityWhy are cloud metadata endpoints an SSRF target, and how do modern defences work?Adversarial ML & Model PrivacyWhat is membership inference, and why does overfitting raise the risk?Cloud, Container & AI Infrastructure SecurityWhat does the shared responsibility model mean for a managed AI service?Adversarial ML & Model PrivacyWhat is model extraction, what does the attacker gain, and how do you detect and limit it?Cloud, Container & AI Infrastructure SecurityThreat-model an LLM inference API that receives untrusted input at scale.

Related lessons

Protecting Model WeightsWhat Containers Isolate, and What They Do NotCloud IAM and Privilege Escalation PathsSensitive Information DisclosureSandboxing Untrusted Code