PolySwarm announced today the release of PolyScore™, a threat scoring algorithm that provides the probability that a given file contains malware in a single, authoritative number.
PolyScore™ has been designed to address some of the main shortcomings associated with crowdsourced models and existing multiscanners:
- Multiple and often conflicting binary opinions require additional, intuition-based work from analysts; which is time intensive, produces inconsistent results, and can not be automated.
- Scores found in solutions like VirusTotal use basic models that simply summarize results by aggregating opinions, a sub-optimal approach for identifying new and emergent threats.
PolyScore 's algorithm filters the noise and amplifies the signal by weighting the engine’s opinions based on recent past performance, strengths, confidence levels, and other rich contextual threat indicators built from millions of daily assertions generated inside PolySwarm.
“As the volume and complexity of cyber threats increase, contextualizing and prioritizing incidents becomes more critical. We developed PolyScore to enable SOC and CTI teams to make quick defensive decisions at scale, with unprecedented accuracy,” stated Paul Makowski, CTO of PolySwarm.
PolyScore uses a semi-supervised machine learning model to continuously improve over time and already outperforms any other methods by a significant margin, currently yielding a 97% accuracy rate.
See PolyScore in action!
