> For the complete documentation index, see [llms.txt](https://deai.gitbook.io/deai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://deai.gitbook.io/deai/security/data-poisoning.md).

# Data Poisoning

Data poisoning involves introducing malicious data to manipulate model outputs towards the attacker's intention. It's effective in both general machine learning and federated learning. In DeAI environments, where data is provided by individuals, it's particularly vulnerable to such attacks.

For classification models, a prevalent technique in data poisoning is label-flipping, wherein honest training examples are systematically switched from one class label to another. Consequently, the affected model erroneously predicts these examples to the corresponding altered labels.

To defend data poisoning, several defense strategies have been proposed. Reject on Negative Impact (RONI) measures the impact of each training example on the error rate, and remove those with large nagatvie impact. Additionally, loss functions can be leveraged to detect and mitigate the influence of malicious data.

Data sanitization techniques offer another avenue for early identification of malicious data. Methods such as BERT embedding, activations analysis and provenance information analysis can aid in this detection process.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://deai.gitbook.io/deai/security/data-poisoning.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.