datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
xai-attack-detection-cifar10
XAI Attack Detection — CIFAR-10 PGD
This private research dataset contains balanced, paired clean and adversarial images for
studying whether an attack can be detected from a classifier explanation map.
Dataset construction
The source is the CIFAR-10 test split. A fine-tuned OpenCLIP ViT-B/16 classifies each
clean image. Clean-correct examples are attacked with untargeted L-infinity PGD using
epsilon 8/255, step size 2/255, 10 steps, and deterministic random… See the full description on the dataset page: https://huggingface.co/datasets/nimaeb/xai-attack-detection-cifar10.ebpf-k8s-attack-detection
eBPF Kubernetes Attack Detection — Syscall Aggregation Study
Dataset accompanying the paper:
In progress
Overview
Two experimental runs on a 4-node Kubernetes cluster (Kubernetes 1.35.4, Proxmox VMs, Intel Core i7-8700, 32 GB RAM) running the Online Boutique microservices application (11 services). Kernel events collected with Tracee 0.24 (21 event types). Traffic generated by Locust 2.43.
Two traffic profiles:
Seasonal (seasonal/) — diurnal load cycle, 20–200 virtual… See the full description on the dataset page: https://huggingface.co/datasets/jniecko/ebpf-k8s-attack-detection.presentation-attack-detection-2d-datasetThe dataset consists of photos of individuals and videos of him/her wearing printed 2D
mask with cut-out holes for eyes. Videos are filmed in different lightning conditions
and in different places (*indoors, outdoors*), a person moves his/her head left, right,
up and down. Each video in the dataset has an approximate duration of 15-17 seconds.prompt-injection-attack-detection-multilingual
Prompt Injection Attack Detection Multilingual Dataset
📌 Overview
This dataset is a merged and cleaned combination of two publicly available datasets for prompt injection detection:
PromptInjectionDataset/Injection-Attack-Detection-Dataset
rikka-snow/prompt-injection-multilingual
The goal of this merged dataset is to provide a larger and more diverse benchmark for binary classification of prompt injection attacks.
🎯 Task
Binary classification:
0 →… See the full description on the dataset page: https://huggingface.co/datasets/Octavio-Santana/prompt-injection-attack-detection-multilingual.2d-masks-presentation-attack-detectionThe dataset consists of videos of individuals wearing printed 2D masks or
printed 2D masks with cut-out eyes and directly looking at the camera.
Videos are filmed in different lightning conditions and in different places
(indoors, outdoors). Each video in the dataset has an approximate duration of 2
seconds.cut-2d-masks-presentation-attack-detectionThe dataset consists of videos of individuals wearing printed 2D masks or
printed 2D masks with cut-out eyes and directly looking at the camera.
Videos are filmed in different lightning conditions and in different places
(indoors, outdoors). Each video in the dataset has an approximate duration of 2
seconds.web-attack-detectionThe dataset contains 625,904 attack payload samples, with 294,771 labeled as 1 and 331,129 labeled as 0, including SQL injection, XSS, command injection, and other vulnerabilities.
xai-attack-detection-imagenette
XAI Attack Detection: Imagenette targeted BIM/PGD on ViT-B/16
Private research dataset of paired clean and targeted adversarial Imagenette images. It is
built to study how adversarial attacks change a Vision Transformer's explanation maps and to
support later work on attack detection. Each row is one source image with its clean and its
attacked version.
Summary
Pairs
12,420 (train 8,690 · validation 1,860 · test 1,870)
Source images
Imagenette v2… See the full description on the dataset page: https://huggingface.co/datasets/nimaeb/xai-attack-detection-imagenette.web-attack-detectionThe dataset contains 625,904 attack payload samples, with 294,771 labeled as 1 and 331,129 labeled as 0, including SQL injection, XSS, command injection, and other vulnerabilities.
Injection-Attack-Detection-Dataset
Dataset Details
Dataset Description
A proprietary dataset was constructed with examples in Portuguese, complemented by two public datasets in English: (Jaziri, 2024) and (Harelix, 2023). These sets were randomly mixed, resulting in a single dataset composed of diverse examples of legitimate and manipulated prompts, with the aim of simulating Prompt Injection attacks.