MAPID thesis defense
Contribution 3 · Adaptive adversary

One differentiable graph

How the joint-loss attack backpropagates through four separate models. The three places the chain breaks, the repair for each, and a live Gumbel-Softmax temperature demo.

Contribution 1 · Multi-turn defense

Evidence, turn by turn

The log-odds accumulator. Build a session by firing detectors, watch evidence stack against the gate, and switch the intent primitive to see why the swap mattered.

Contribution 1 · The diagnosis

Fitted to the wrong thing

A boundary around one class versus a boundary between two. Four stages from the 30-of-30 failure through the Pyrrhic corner to the classifier.

Methods · Hyperparameters

Which knobs mattered

All 23 settings sorted by how each value was arrived at, with the one real sensitivity sweep and the loss trajectories that answer the budget question.

Methods · Evaluation

Rare by nature

Class imbalance at three ratios pointing two directions, and a base-rate slider showing what a 0.033 false-positive rate actually costs in precision.

Architecture · Coordinator

On the path, not in the loss

Why gradients traverse the Coordinator although it has no loss term, and a switchboard showing the dispatch plan answers to configuration rather than to anything the attacker sends.

Indirect injection · Map-Reduce

Quarantine by fragment

Step an injected instruction through chunking, six isolated workers and the Reduce step, then see the category crossover where this defense wins exactly where the Guards lose.

Defense in depth · Capability gate

Required minus granted

The gate as set arithmetic. Pick a response, set what the deployment grants, and watch allow or deny fall out, including the emoji result that works by accident.

Perimeter · Input and Output Guards

Bread, filling, bread

Build the sandwich prompt for either Guard, then see why they swap roles between benchmarks and why that makes one faster than no defense at all.

Contribution 2 · Security-aware KV cache

What the cache forgets first

A hyperparameter-driven simulation of twenty-five invocations under pressure. Move the threshold and the protected floor and watch the canary survive or vanish.