Popular GitHub Repositories for Harmful Algal Blooms

Introducing DeToxify – the ultimate toxicity buster. Armed with NLP superpowers, it sniffs out harmful comments, neutralizes negativity, and rewrites toxicity into pure class. Think of it as Jarvis for your online spaces—making the internet a kinder, smarter place, one comment at a time.

This is an open-source version of the representation engineering framework for stopping harmful outputs or hallucinations on the level of activations. 100% free, self-hosted and open-source.

A survey on harmful fine-tuning attack for large language model

The official implementation of ECCV'24 paper "To Generate or Not? Safety-Driven Unlearned Diffusion Models Are Still Easy To Generate Unsafe Images ... For Now". This work introduces one fast and effective attack method to evaluate the harmful-content generation ability of safety-driven unlearned diffusion models.

This is the official code for the paper "Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation"

Resources (conference/journal publications, references to dataset) for harmful memes detection.

Task Complexity Classifier using Transformer-based NLP model based on Bloom's Taxonomy

Network related services, programs and applications are developing greatly, however, network security breaches are also developing with them. Network security is an evolving, challenging and a critical task. It is essential that there is a system in place to identify any harmful movement happening in network. An Intrusion detection system (IDS) has become the prerequisite software addressing cyber security in the modern era. Especially, with the greater complexity of advanced cyber-attacks and as such the uncertainty surrounding the detection of the types of attacks. This thesis proposes a novel approach using an ensemble of K-Means and Gaussian Mixture clustering combined with a deep neural network (DNN) algorithm. When compared with traditional artificial neural network’s (ANN’s) used within an IDS, our approach implements modern advanc

Guarding OpenStreetMap from harmful edits using machine learning

Official repository for the ICLR 2023 paper "A Learning Based Hypothesis Test for Harmful Covariate Shift"

Out of boredom, I was thinking about how to bypass the security on Claude3, which probably has the best protection "against harmful content". And I came up with the idea that when the AI itself knows my question, but cannot say it, there is a chance to bypass the security, and it worked.

Intercept AI - Empowering safe and secure online experiences through cutting-edge content filtering technology.

A Deep-learning model for classifying a meme as harmful or not

HMS - Harmful Brain Activity Classification || Classify seizures and other patterns of harmful brain activity in critically ill patients

In this work, we extend the FGSM method proposing multistep adversarial perturbation (MSAP) procedures to study the recommenders’ robustness under powerful methods. Letting fixed the perturbation magnitude, we illustrate that MSAP is much more harmful than FGSM in corrupting the recommendation performance of BPR-MF.

Our AI model, integrated with the Defend Her extension, uses advanced natural language processing (NLP) to predict and identify offensive or harmful language in real-time ???. By continuously analyzing content, it ensures accurate detection of bad words, enhancing the extension's ability to protect users and promote a safer online environment ?

This is "SHIELD" a System for Harmful explicit-content Identification and Evaluation through LLM-Driven approach. Its primary object is to score explicit content for 15 different categories of explicitness from 0 to 100, it leverages LLMs as the primary scoring tool.

RocketFood is a modern web application that helps users make healthier food choices by scanning and analyzing food ingredient labels. The app uses optical character recognition (OCR) to extract text from images of food labels, then analyzes the ingredients to identify potentially harmful, misleading, or toxic components.

*WIP* A Python-based desktop application for monitoring screen content and alerting parents about potentially harmful material using OpenAI or Gemini (Currently testing local vision LLMS with small footprint - probably will fine-tune my own)

Evaluating the impact of Gaussian Copula-based synthetic data augmentation on ML models for Harmful Algal Bloom detection