Machine Learning Use Cases in Identity and Access Management
simMachines supports a wide variety of use cases for Identity and Access Management, many uniquely enhanced by the R10 Similarity Engine in the XAI Analytic Workbench.
Probabilistic Identity Matching
Problem
CDP’s and DMP’s wanted to improve deterministic match rates to improve lookalike model performance and increase ad reach for their clients.
Challenges
Deterministic matching techniques optimized over years still left a large volume of records unmatched.
Probabilistic matching distance functions are computationally expensive and are typically too slow for large scale, real world usage.
Solution
simMachines’ high speed similarity engine was able to index the high dimension matching file with a high-accuracy character level distance function in hours and retrieve queries in milliseconds.
Result
For a CDP, an average match rate lift of 38% was achieved, and for a DMP we provided a match rate lift of 21%, both on client data sets being matched to national files.
Adaptive Authentication
Problem
Our partner needed an explainable anomaly detection capability to expand beyond its rule-based system to detect unauthorized access to its network.
Challenges
The existing rule-based system was not robust to emergent intrusion threats and required extensive labor to deploy and manage.
Solution
Our partner used simMachines’ Anomaly Detector to develop a system for identifying unusual usage patterns amongst their users supported by the specific anomalous factors triggering the detection.
Result
The resulting system is able to identify anomalous user behavior and respond to detected events with enhanced authentication requests and other remedial options.
Device Linking
Problem
Large DMP needed to improve linking of multiple devices from anonymous users.
Challenges
Deterministic matching techniques used to link multiple internet devices to the same user leave billions of devices unmatched and frequently drop or duplicate records as device parameters are changed. This presents a significant obstacle in identity management and ad delivery.
Solution
Ensemble approach using gradient boosting and similarity-based machine learning.
Result
70% increase over current fraud detection performance. Time to implement reduced from 6 weeks to 2 weeks. Continuous learning enabled and the Why factors reported/available behind every prediction.
Access Rights & Reconciliation
Problem
Our partner needed to be able to group together application access rights by role for its clients to assess existing access rights in terms of uniformity as well as assign the right level of access to new employees.
Challenges
Existing deterministic approaches left too many unmatched records. Typical probabilistic approaches are too computationally expensive to use at scale.
Solution
Utilizing character level distance functions, records with commonality can be automatically grouped and presented by distance of match to reduce reconciliation time and resource requirements.
Result
Leveraging this approach our partner is able to reduce the labor associated with the reconciliation process by 80%.
