Use Case: Spam Detection

We provide a dataset intended for spam detection for quick testing. See lingspam for more information.

  • SecuML_server must be executed to launch the web server. http://<host>:<port>/SecuML/ gives access to SecuML menu.

    SecuML_server --secuml-conf <path_to_conf_file>
    
  • In the configuration file, input_data_dir must be set to input_data to test SecuML with the lingspam dataset we provide.

Note

The configuration file is required to run SecuML executables (e.g. SecuML_server, SecuML_DIADEM, SecuML_ILAB). It can be specified either with the parameter --secuml-conf for each execution, or globally with the environment variable SECUMLCONF.

Training and Diagnosing a Detection Model

SecuML_DIADEM SpamHam lingspam --secuml-conf <conf_file> LogisticRegression

Once the experiment has been completed, the following message is displayed:

Experiment <experiment_id> has been successfully completed.
See http://<host>:<port>/SecuML/<experiment_id>/ to display the results.
_images/performance.png

DIADEM Monitoring Interface

See DIADEM for more detail.

Annotating a Dataset with a Reduced Workload

SecuML_ILAB SpamHam lingspam --secuml-conf <conf_file> -a init_annotations.csv Ilab --auto --budget 500

Once the experiment has been completed, the following message is displayed:

Experiment <experiment_id> has been successfully completed.
See http://<host>:<port>/SecuML/<experiment_id>/ to display the results.
_images/monitoring.png

ILAB Monitoring Interface

See ILAB for more detail.