Course Objective & Scope
Every model API deployed to production exposes attack surfaces including unauthorised access, resource exhaustion (DoS), malicious model artefacts, data leakage and model extraction. This programme combines classic API and container security with OWASP ML/LLM security controls to harden the model serving layer.
Hands-on labs cover authentication with OAuth2/JWT and mTLS, input validation and rate limiting, safe serialization with safetensors and model signing against pickle exploits, PII masking and audit logging. The training concludes with a threat model and security hardening checklist tailored to the organisation's serving architecture.
Who Should Attend?
- ML, MLOps and backend engineers building and operating model APIs
- Application security (AppSec) and DevSecOps specialists auditing AI services
- Platform engineers responsible for data privacy and infrastructure security in regulated sectors
- Prerequisite: familiarity with REST APIs, Docker and basic networking/security concepts
Key Takeaways
- Conduct threat modelling (STRIDE and OWASP ML/LLM Top 10) for a model serving architecture
- Protect inference endpoints using OAuth2/JWT, mTLS, API gateways and rate limiting
- Prevent malicious model artefact attacks using safe formats, scanners and cryptographic signing
- Apply guardrails against adversarial inputs, model extraction and prompt injection attempts
- Implement PII masking and privacy-compliant audit trails in prediction logs
Curriculum
Threat Modelling at the Inference Layer
- Attack surfaces through the lens of OWASP Top 10 for ML and LLM Applications
- Threat analysis across data, model artefact, runtime and network layers
Authentication, Authorisation and Network Security
- API gateways, OAuth2/JWT, API key rotation and role-based access control
- Service-to-service mTLS, private endpoints and network isolation
- Token- and request-based rate limiting, quotas and payload size caps
Model Artefact and Supply Chain Security
- Pickle deserialization exploits, scanning with Fickling/ModelScan and migrating to safetensors
- Signing models and container images with Sigstore/Cosign
- Rootless containers, distroless base images and secret management
Model-Targeted Attacks and Guardrails
- Model extraction, membership inference and adversarial examples
- Output probability clipping, anomaly detection and query budgets
- Prompt injection filters and output validation for LLM services
Data Privacy, Logging and Security Checklist
- PII detection, masking in request/response logs and data protection compliance
- Audit logging and incident response readiness
- Security hardening review on the organisation's serving architecture
Other Courses in This Category
-
Exporting TensorFlow and Keras models as SavedModel, defining SignatureDefs and serving them over REST and gRPC with TensorFlow Serving. For data scientists and ML engineers starting to deploy TensorFlow models.
- Live Course
- 5 modules
-
Multi-model configurations, dynamic batching, XLA/TensorRT and mixed-precision acceleration, GPU memory tuning, Prometheus metrics and custom servables in TensorFlow Serving. For engineers operating high-traffic TF services.
- Live Course
- 5 modules
-
Operating TensorFlow Serving at scale on Docker and Kubernetes: pulling models from object storage, GPU scheduling, readiness/liveness probes, HPA/KEDA autoscaling and KServe integration. For MLOps and platform engineers.
- Live Course
- 5 modules
-
Deploying PyTorch models with TorchServe: writing custom Handlers, packaging .mar files with torch-model-archiver, using the Inference and Management APIs, and managing workers and versions. For data scientists and ML engineers using PyTorch.
- Live Course
- 5 modules
-
Multi-model DAG workflows in TorchServe, acceleration with torch.compile/ONNX/TensorRT, multi-GPU and large-model parallel serving, custom Prometheus metrics and production security. For senior ML and MLOps engineers.
- Live Course
- 5 modules
-
Dynamic batching and high-volume bulk scoring in TorchServe: calibrating batch_size and max_batch_delay, writing vectorised Handlers, padding variable-length inputs and integrating queues. For ML and data engineers.
- Live Course
- 5 modules
Get in touch
