A collection of UIs and interactive reports to assist BD model administration, and some data exploration.

MODEL MANAGEMENT

prune models

SPARKOV PRUNER

manage eval collections

MANAGE MODEL EVAL COLLECTIONS to truncate collections/delete models after evaluation

False Negative reader

false negative UI displays newly reported false negatives, with an option to automatically sample the reported behaviour

AUTOMATED MODEL CREATION SUPERVISION

a set of uis to inspect auto-created matrices

how to evaluate new models:

  1. find (in Kibana or Graylog) a query that describes the bot data

  2. use the bot session_id store UI to get the corresponding session_ids

  3. run ./botSampler (or wait for daily cron)

  4. run ./matrixEvaluator (or wait for daily cron)

  5. VISUALISE MATRIX EVALUATION RESULTS and set inner thresholds

  6. run ./submitWithEvalMatrices (or wait for daily cron)

  7. VISUALISE EVALUATED MODEL RESULTS and set thresholds or discard model