Skip to content

Conversation

@mazhurin
Copy link
Collaborator

No description provided.

mazhurin and others added 30 commits June 4, 2020 18:34
Unit tests and linting fixes.
… configuration istead of full python model names.
…e.training.model refers to the full module path of the Model class(not the Enum key).
Country and Host features. Stratified sampling(parameter 'max_samples_per_host'). Support for nested features in JSON parser(geoip feature fix).
* Debug log line removed from AttackDetection

* Organization creating is optional. No classifier in labeler.
* Classifier model. 'classifier_score' column. DBReader fix.

* Classifier pipeline. Incident loader class. Train classifier task.
* Incident detector. Fix for delayed stop.

* Unit tests are working now.

* Linting fixes.
…aly score rather than on both anomaly and classifier score. The reason is that classifier is biased towards historical incidents and is not good in detecting previously unseen patterns. (#113)
* KSQL added to the k8s deployment.
* ed_retrivier removed
parsing weblog in spark
* postprocessing streamin to s3. Tested.
* Readme update. NA fix in postprocessing
* new json weblogs format
* Lostash deployment. LoadBalancer in Kafka.
* Logstash, prediction_behave column, timescaledb.
* Elastic search added to the deployment.
* Host as a key in send_challenge. First testing version of utm_source, utm_medium, utm_capmpaign cstats columns.

* Kafka field names fixes.
* Whitelisting solved challenge IPs

* 20m stats topic size increase. 80GB kafka storage increase.

* domain whitelisting fix

* Grace period for fresh sessions. Warmup period for sessions.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants