Alembic - Automated Model Inference for Stateful Network Functions

Explore a conference talk on Alembic, an automated model inference system for stateful network functions. Learn about the challenges of creating accurate models for complex network functions and how Alembic addresses these issues. Discover how the system uses symbolic finite-state machine representations and an ensemble approach to generate behavioral models for given configurations. Gain insights into the practical applications of Alembic in network testing and verification, and understand its potential impact on improving the accuracy and efficiency of network function modeling. Examine the evaluation results, including a case study on the Untangle Firewall, and consider the limitations and future directions of this innovative approach to network function modeling.

Intro
Motivating Example: Stateful Firewall (FW)
Today: Need NF Models for Testing and Verification
Limitation of Handwritten Model: Inaccuracy
Challenges on Large Configuration Space
We Can Compose Models of Individual Rules
Use Symbolic Models to represent Large Sets
Exploit Independence to Create an Ensemble of FSMS
Challenges on Inferring NF Behavior
Background on L* for Black-box FSM Inference
Practical Challenges of Applying L* for an NF
Generating Input Alphabet to handle Large Traffic Space
Learning the State Granularity
Alembic Workflow: Offline
Evaluation Summary
Evaluation Setup
Firewall Case Study: Untangle Firewall
Limitations and Future Work
Conclusions: Alembic can accurately model stateful NFS