Trojan Detection Software Challenge - Round 9 Train Dataset

This is the training data used to construct and evaluate trojan detection software solutions. This data, generated at NIST, consists of natural language processing (NLP) AIs trained to perform one of three tasks, sentiment classification, named entity recognition, or extractive question answering on English text. A known percentage of these trained AI models have been poisoned with a known trigger which induces incorrect behavior. This data will be used to develop software solutions for detecting which trained AI models have been poisoned via embedded triggers. This dataset consists of 210 QA AI models using a small set of model architectures. Half (50%) of the models have been poisoned with an embedded trigger which causes misclassification of the input when the trigger is present.

Data and Resources

Field Value
accessLevel public
bureauCode {006:55}
catalog_@context https://project-open-data.cio.gov/v1.1/schema/data.json
catalog_conformsTo https://project-open-data.cio.gov/v1.1/schema
catalog_describedBy https://project-open-data.cio.gov/v1.1/schema/catalog.json
identifier ark:/88434/mds2-2539
issued 2022-01-26
landingPage https://data.nist.gov/od/id/mds2-2539
language {en}
license https://www.nist.gov/open/license
modified 2022-01-24 00:00:00
programCode {006:045}
publisher National Institute of Standards and Technology
resource-type Dataset
source_datajson_identifier true
source_hash 0854bee2cdad991d5eb1d0b3f37a97d91812729e
source_schema_version 1.1
theme {"Information Technology:Software research","Information Technology:Cybersecurity"}
Groups
  • AmeriGEOSS
  • National Provider
  • North America
Tags
  • adversarial-machine-learning
  • ai
  • amerigeo
  • amerigeoss
  • artificial-intelligence
  • ckan
  • geo
  • geoss
  • machine-learning
  • national
  • north-america
  • trojan-detection
  • united-states
isopen False
license_id other-license-specified
license_title other-license-specified
maintainer Michael Paul Majurski
maintainer_email michael.majurski@nist.gov
metadata_created 2025-11-21T08:40:20.896937
metadata_modified 2025-11-21T08:40:20.896941
notes This is the training data used to construct and evaluate trojan detection software solutions. This data, generated at NIST, consists of natural language processing (NLP) AIs trained to perform one of three tasks, sentiment classification, named entity recognition, or extractive question answering on English text. A known percentage of these trained AI models have been poisoned with a known trigger which induces incorrect behavior. This data will be used to develop software solutions for detecting which trained AI models have been poisoned via embedded triggers. This dataset consists of 210 QA AI models using a small set of model architectures. Half (50%) of the models have been poisoned with an embedded trigger which causes misclassification of the input when the trigger is present.
num_resources 57
num_tags 13
title Trojan Detection Software Challenge - Round 9 Train Dataset