Detection Feedback

Submit detection misclassifications to Abnormal. Your input helps us improve our models and increase transparency.

In some cases, we may create temporary rules to detect similar messages while we work on updating our models. If you would like to permanently exempt certain messages from being flagged, you can do so from the Abnormal settings page.

2 Reports
Report Volume and Improvements (last 30 days)
25 Improvements
Detection improvements that were made based on your submission(s).
Report Type
Filter by
Apr 25, 2023 - Jul 23, 2023
SubmissionSubmitted byAnalysisInsights SummaryCase Status
Wojciech Iwański - Admin
Jun 1, 2023

The missed attack has been contained, and we have added a feature to detect similar messages in the future. This missed attack occurred due to a gap in sender and recipient frequency analysis. To address this, we will be deploying a new, unified machine learning model to enhance detection precision and reduce missed attacks by the end of Sep 2023. We will also be delivering key performance improvements to our wide multimodal models for greater detection accuracy and coverage across complex attack types by the end of Sep 2023.

RESOLVED
2 Emails Flagged
Read Emails: 1
Forwarded by User: 0
Replied to Message: 0
Sarah Chen - IT Manager
Jun 5, 2023

The flagged message was a legitimate business communication that was incorrectly identified as suspicious due to urgent language patterns. This occurred because our model was overly sensitive to financial urgency keywords in external communications. We have updated our legitimate business communication patterns and reduced sensitivity for established business relationships to prevent similar false positives while maintaining security coverage.

RESOLVED
12 Released + 23 Auto-Delivered
Total Reviewed: 38
Still Quarantined: 0
User Reports: 1
20 / page