FloodMask
Collection
FloodMask is a phishing detection pipeline inspired by PhishIntention • 3 items • Updated
FloodMask Logo Classifier
Part of the FloodMask phishing detection pipeline. Takes logo regions cropped by the AWL detector and classifies them against a known brand reference list for phishing attribution.
Pipeline Role
Screenshot -> [AWL] -> logo region -> [Logo Classifier] -> brand match -> [CRP Classifier] -> phishing verdict
Model Details
- Architecture: DeiT (Data-efficient Image Transformer), distilled variant, fine-tuned for brand logo classification
- Base feature extractor:
facebook/deit-base-distilled-patch16-384 - Format: ONNX (optimized for CUDA via
optimum.onnxruntime) - Input: Cropped logo region (RGB image)
- Output: Brand class probabilities
Usage
from optimum.onnxruntime import ORTModel
from transformers import AutoImageProcessor
from PIL import Image
processor = AutoImageProcessor.from_pretrained("ashim/floodmask-logo-classifier-2026-03-06")
model = ORTModel.from_pretrained("ashim/floodmask-logo-classifier-2026-03-06")
image = Image.open("logo_crop.png").convert("RGB")
inputs = processor(images=image, return_tensors="pt")
outputs = model(**inputs)
CUDA Usage
CUDA_OPTIONS = {
"device_id": 0,
"arena_extend_strategy": "kNextPowerOfTwo",
"gpu_mem_limit": 2 * 1024 * 1024 * 1024,
"cudnn_conv_algo_search": "EXHAUSTIVE",
"do_copy_in_default_stream": True,
}
model = ORTModel.from_pretrained(
"ashim/floodmask-logo-classifier-2026-03-06",
provider="CUDAExecutionProvider",
provider_options=CUDA_OPTIONS,
)
Architecture Decisions
google/mobilenet_v2_1.0_224 was evaluated as an alternative. DeiT (distilled, patch16-384) was selected for stronger feature representations on small, visually diverse logo crops where MobileNet's inductive biases were less effective.
Related Models
Citation
@inproceedings{liu2022phishintention,
title={PhishIntention: Toward Explainable Content-based Phishing Detection through Intent Identification},
author={Liu, Ruofan and Lin, Yun and Yang, Xianglin and Ng, Siang Hwee and Divakaran, Dinil Mon and Dong, Jin Song},
booktitle={USENIX Security Symposium},
year={2022}
}