PipeIntegrity

Image and Text Validation

PipeIntegrity is a PipeCodes internal package written in (and to be used with) JavaScript (Node.js). It's function is to validate certain user inputs, returning back a result regarding the validation of that said input, thus protecting the integrity of any web application.

Features

Here's what PipeIntegrity, as of the current version, detects as invalid:

• Images of nudity and pornographic content;
• Profane and/or inappropriate text.

Technology

Here are some technologies PipeIntegrity depends on:

• TensorFlow.js - Machine learning library;
• TensorFlow.js via Node.js - TensorFlow adapter to Node.js;
• NSFWJS - Evaluates image content;
• bad-words - Evaluates text content.

Installation

PipeIntegrity requires Node.js v14+ to run. To install PipeIntegrity, please run the following commands on your terminal:

cd [your project folder]
npm i @pipecodes/pipeintegrity

Usage

The package includes two functions you can use for each of these inputs.

Images

The "validateImage" function will first validate if the image extension matches the extension provided in the second argument, and then will validate the image's content.

const pipeintegrity = require('@pipecodes/pipeintegrity');

const myImage = "[this must be a string with either an URL or a local path to that image]";
const myExtension = 'png';

// Default: 0.4 - this is the "sensitivity" the validator is when encountering any inappropriate content in the image. 
// The value must be between 0 and 1 (excluding).
// The higher the value, the more difficult will be for any inappropriate image to be invalidated.
const myThreshold = 0.7; 

pipeintegrity.validateImage(myImage, myExtension, myThreshold)
    .then((validation) => {
        // Here you can check "validation.isValid" for a decision on wether 
        // the image is appropriate or not.
        // Also, "validation.predictions" contains an array with the types of image it might fall into.
    })
    .catch((error) => {
        // Handle errors.
    });

Text

The "validateText" function will take a text string and validate if the text has any inappropriate words of phrases that are included in the default list of profane content. As a second argument to this function, you can provide an object containing the properties "blacklist" and "whitelist". Both of them must be an array with strings, otherwise an error will be thrown.

• options.blacklist adds extra words to be marked as "forbidden";
• options.whitelist removes words that were otherwise "forbidden" and are now "allowed".

const pipeintegrity = require('@pipecodes/pipeintegrity');

const myText = "[any kind of text in a string]";
const options = {
    blacklist: ['This MUST be an array', 'containing', 'only strings'],
    whitelist: ['This also MUST be an array', 'containing', 'only strings'],
}

try {
    const validation = pipeintegrity.validateText(myText, options);
    // "validation.isValid" has the decision of
    // wether the text has any inappropriate content or not.
    // You can also access "validation.textClean" for a
    // censored version of the text provided.
} catch (error) {
    // Handle errors.
}