Skip to content

itsriodejaneiro/pegabot-back-end

BranchesTags

Repository files navigation

Available in Spanish Portuguese

spottingbot

Analyzing profiles on Twitter to detect bot behavior

spottingbot is an experimental and open-source project that needs you to evolve, do not hesitate to contribute on our GitHub repository by opening a pull request or to contact us at [email protected]. Documentation on how the current indexes are calculated is also available here

You can also join us on our Telegram group to freely talk about suggestions, improvement or simply ask us anything

Usage

spottingbot can be used both as a command-line interface application (cli) or as an independent module

Command-line interface

Installation

npm install

Settings file

Create a .env file that contains:

TWITTER_CONSUMER_KEY="Your application consumer key"
TWITTER_CONSUMER_SECRET="Your application consumer secret"
TWITTER_ACCESS_TOKEN_KEY="Your application access token key, only for user authentication"
TWITTER_ACCESS_TOKEN_SECRET="Your application access token secret, only for user authentication"

PORT="The port that the app will attach itself"

USE_CACHE="Flag that will control the cache usage. Must be either 0 or 1.
DEFAULT_CACHE_INTERVAL='The interval that the cache will be considered valid. Must be a valid Postgresql format, e.g: '30_days'."

DATABASE_HOST="Your Postgresql host"
DATABASE_USER="Your database user"
DATABASE_PASSWORD="Your database password"
DATABASE_NAME="Your database password"

BULK_SECURITY_TOKEN='Key that will be used to authorize calls to the bulk endpoint'

PUPPETER_SERVICE_ROOT_URL="Url to the service that will take screenshots from the Twitter profile"
PUPPETER_SECRET_TOKEN="Token for the screenshot service"
PSQL database

Pegabot uses a cache system to avoid remaking an analysis on the same user. This cache system uses a PostgreSQL database.

How to set-up the PostgreSQL database:

  1. Install and set PSQL on your machine
  2. Create a database for pegabots. If you want, you can create a new user with a password to own this database, or just use the default postgres user.
  3. Fill up the database data on the .env file. Example:
DATABASE_HOST="127.0.0.1"
DATABASE_USER="postgres"
DATABASE_PASSWORD=""
DATABASE_NAME="pegabot"
  1. Install the npm module sequelize-cli
  2. Run migrations with sequelize-cli db:migrate

Starting the API

You can run the command: npm run dev in order to start with nodemon and Babel. Or you can start with npm run start, which will build the application and run from a /build folder without Babel.

Endpoints

/botometer

Get analysis with only main indexes.

Method: GET

Example: curl --request GET --url 'http://127.0.0.1:5010/botometer?profile=twitter&search_for=profile&is_admin=true'

Required parameters

  • profile (STRING): user handle that will be analyzed;

  • search_for (STRING): only one value allowed profile

Optional parameters

  • is_admin (BOOLEAN): used to identify requests that are coming from the Pegabot admin interface; (1|0 OR true|false)

Response

{
  "metadata": {
    "count": 1
  },
  "profiles": [
    {
      "username": "twitter",
      "url": "https://twitter.com/twitter",
      "avatar": "http://pbs.twimg.com/profile_images/1354479643882004483/Btnfm47p_normal.jpg",
      "language_dependent": {
        "sentiment": {
          "value": 0.57
        }
      },
      "language_independent": {
        "friend": null,
        "temporal": 0.32269754443513865,
        "network": 0.4103015075376884,
        "user": 0
      },
      "bot_probability": {
        "all": 0.18614272171040389,
        "info": "<p>Um dos critérios que mais teve peso na análise foi o índice de Perfil</p>"
      },
      "user_profile_language": null
    }
  ]
}

/analyze

Get analysis with main indexes and subindexes.

Method: GET

Example: curl --request GET --url 'http://127.0.0.1:5010/analyze?profile=twitter&search_for=profile'

Required parameters

  • profile (STRING): user handle that will be analyzed;

  • search_for (STRING): only one value allowed profile

Response

{
  "root": {
    "profile": {
      "handle": "twitter",
      "link": "https://twitter.com/twitter",
      "label": "Perfil",
      "description": "<p>Algumas das informações públicas dos perfis consideradas na análise do PEGABOT são o nome do perfil do usuário, e quantos caracteres ele possui, quantidade de perfis seguidos (following) e seguidores (followers), texto da descrição do perfil, número de postagens (tweets) e favoritos. Após coletar as informações, os algoritmos do PEGABOT processam e transformam os dados recebidos em variáveis que compõem o cálculo final de probabilidade.</p>",
      "figure": "",
      "analyses": [
        {
          "title": "SELO DE VERIFICAÇÃO",
          "description": "A presença do selo de verificação oferecido pelo Twitter influencia positivamente nos resultados, uma vez que a plataforma possui um procedimento manual para validar a identidade desses usuários.",
          "summary": "<p>Usuário verificado</p>",
          "conclusion": 0
        }
      ]
    },
    "network": {
      "description": "<p>O algoritmo do PegaBot coleta uma amostra da linha do tempo do usuário, identificando hashtags utilizadas e menções ao perfil para realizar suas análises. O objetivo é identificar características de distribuição de informação na rede da conta analisada.</p>O índice de rede avalia se o perfil possui uma frequência alta de repetições de menções e hashtags. No caso de um bot de spams, geralmente se usam as mesmas hashtags/menções, e é isso que esse índice observa. Por exemplo, se 50 hashtags são usadas e são 50 hashtags diferentes, não é suspeito, mas se só uma hashtag é usada 100% das vezes, então é muito suspeito.</p>",
      "label": "Rede",
      "analyses": [
        {
          "title": "DISTRIBUIÇÃO DAS HASHTAGS",
          "description": "<p>Calcula o tamanho da distribuição dessas hashtags na rede. Ou seja, avalia se a utilização de hashtags do perfil apresenta uma frequência anormal.</p><p>Quanto mais próximo de 0% menor a probabilidade de ser um comportamento de bot.</p>",
          "summary": "<p>Total: 10. Únicas: 3</p>",
          "conclusion": "0.70",
          "hashtags": []
        },
        {
          "title": "DISTRIBUIÇÃO DAS MENÇÕES",
          "description": "<p>Calcula o tamanho da distribuição de menções ao perfil do usuário na rede. Ou seja, avalia as menções realizadas pelo perfil com base em sua frequência.</p><p>Quanto mais próximo de 0% menor a probabilidade de ser um comportamento de bot.</p>",
          "summary": "<p>Total: 14. Únicas: 14</p>",
          "conclusion": "0.00",
          "mentions": []
        },
        {
          "title": "HASHTAGS E MENÇÕES",
          "description": "HASHTAGS E MENÇÕES",
          "summary": "<p>Calculamos o score distríbuido (0.35) e o tamanho da rede (0.06)</p>",
          "conclusion": "0.41",
          "stats": []
        }
      ]
    },
    "emotions": {
      "description": "<p>Após coletar os dados, os algoritmos do PEGABOT fornecem uma pontuação, em uma escada de -5 a 5m de cada uma das palavras dos tweets coletados. A classificação se baseia em uma biblioteca, onde, cada uma das palavras possui uma pontuação, sendo considerada mais ou menos negativa, positiva ou neutra. Assim, ao final da classificação, calcula-se a pontuação média para a quantidade de palavras positivas, negativas e neutras utilizadas pelo usuário.</p>",
      "label": "Sentimentos",
      "analyses": [
        {
          "summary": "<p>o perfil tem pontuação de 0.57, classificando-se como positivo.</p>",
          "conclusion": 0.57,
          "samples": {
            "title": "VEJA AQUI O EXEMPLO DE 3 TWEETS DO USUÁRIO",
            "list": []
          }
        }
      ]
    }
  }
}

/botometer-bulk

Bulk process for first endpoint. The params for this request must be sent as JSON

Method: GET

Example:

curl --request GET \
  --url http://127.0.0.1:5010/botometer-bulk \
  --header 'Content-Type: application/json' \
  --header 'X-API-KEY: $ENV{BULK_API_KEY}' \
  --data '{
	"is_admin": true,
	"profiles": [
		"twitter"
	] 
}'

Required headers

  • apiKey (string): Must be set on the .env file, with the BULK_SECURITY_TOKEN name;

Required parameters

  • profiles (STRING ARRAY): Array containing the profiles that will be analyzed.
    • MAX ARRAY SIZE: 50

Optional parameters

  • is_admin (BOOLEAN): used to identify requests that are coming from the Pegabot admin interface; (1|0 OR true|false)
  • twitter_api_consumer_key (STRING): Consumer key that will be used to instantiate a Twitter API client;
    • Must be sent alongside twitter_api_consumer_secret
  • twitter_api_consumer_secret (STRING): Consumer key that will be used to instantiate a Twitter API client;
    • Must be sent alongside twitter_api_consumer_key

Response

{
  "analyses_count": 1,
  "analyses": [
    {
      "twitter_user_data": {
        "id": "111111111111111111111111111",
        "handle": "@twitter",
        "user_name": "Twitter",
        "url": "https://twitter.com/twitter",
        "avatar": "https://twitter.com",
        "created_at": "Thu Apr 09 21:32:54 +0000 2020"
      },
      "twitter_user_meta_data": {
        "tweet_count": 2176,
        "follower_count": 304,
        "following_count": 537,
        "hashtags": [],
        "mentions": []
      },
      "pegabot_analysis": {
        "user_index": 0.2755072802592182,
        "temporal_index": 0.09934108527131784,
        "network_index": 0.7706060606060606,
        "sentiment_index": 0.6,
        "bot_probability": 0.34909088522731935
      },
      "metadata": {
          "used_cache": true
      }
    }
  ]
}

Using as a module

Both User and App-only authentication are supported, for App-only, the Bearer token will be automatically requested

Start

npm start username

or

source/cli.js username

username has to be replaced by the profile to analyze

Install bin locally on your system

npm link sudo might be necessary

Then

spottingbot username

Module

Call

const spottingbot = require('spottingbot');

spottingbot(username, index);

username is a string that contains the screen name of the Twitter profile to analyze.

twitter_config is an object that contains Twitter credentials, both User and App-only authentication are supported, for App-only, the Bearer token will be automatically requested, the twitter_config object should be like this:

{
  consumer_key: "Your application consumer key",
  consumer_secret: "Your application consumer secret",
  access_token_key: "Your application access token key", // Only for User authentication
  access_token_secret: "Your application access token secret" // Only for User authentication
}

index is used to disable some indexes, it's an object that looks like this:

{
  user: true,
  friend: true,
  temporal: true,
  network: true
}

By default, or if omitted, everything is true.

To disable only one index, it isn't necessary to add the others keys in the object, {friend: false}, will work.

Return value

spottingbot handles both callback style and node promise style

Callback
spottingbot(username, twitter_config, index, function(error, result) {
  if (error) {
    // Handle error
    return;
  }
  // Do something with result
})
Promise
spottingbot(username, twitter_config, index)
  .then(result => {
    // Do something with result
  })
  .catch(error => {
    // Handle error
  })
Value

The return value is an object that contains

{
  metadata: {
    count: 1 // Always 1 for now
  },
  profiles: [
     {
       username: 'screen_name',
       url: 'https://twitter.com/screen_name',
       avatar: 'image link',
       language_dependent: {
         sentiment: {
           value: 0.65
         }
       },
       language_independent: {
         friend: 0.19,
         temporal: 0.37,
         network: 0.95,
         user: 0
       },
       bot_probability: {
         all: 0.37
       },
       user_profile_language: 'en',
     }
  ]
}

spottingbot is a project inspired by Botometer, an OSoMe project.

This project is part of the PegaBot initiative.

PegaBot is a project of the Institute of Technology and Society of Rio de Janeiro (ITS Rio), Instituto Equidade & Tecnologia and AppCívico.

spottingbot is an experimental and open-source project that needs you to evolve, do not hesitate to contribute on our GitHub repository by opening a pull request or to contact us at [email protected]. Documentation on how the current indexes are calculated is also available here_

About

No description, website, or topics provided.

Resources

Readme

License

GPL-3.0 license
Activity
Custom properties

Stars

0 stars

Watchers

4 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages