#datasets — 1sec.ai

Multi-Source Cybersecurity Logs: An ATT&CK-Labeled Dataset and SLM Evaluation

Researchers introduce a new dataset and evaluation framework for detecting multi-stage cyberattacks using machine learning. The dataset provides labeled, multi-source logs from system, network, and browser activity. Existing datasets are limited, focusing on a single source or omitting key telemetry. This work enables more accurate detection of complex attacks.

Key takeaways

New dataset provides labeled, multi-source logs for cyberattack detection.
Existing datasets limited to single source or key telemetry.
Enables more accurate detection of complex attacks.

aarXiv#cybersecurity #machine-learning #datasets

researchDec 9

Open Preference Dataset for Text-to-Image Generation by the 🤗 Community

The Hugging Face community has released the Open Preference Dataset, a collection of human preferences for text-to-image generation. This dataset aims to improve the evaluation and development of text-to-image models by providing a standardized benchmark. You can access and contribute to the dataset through the Hugging Face platform. The dataset's release is expected to foster collaboration and advancements in the text-to-image generation field.

Key takeaways

Open Preference Dataset is now available on Hugging Face.
Aims to standardize evaluation for text-to-image models.
Community-driven dataset for collaborative development.

HHugging Face Blog#open-source #text-to-image #datasets