shinyauthr is an R package providing module functions that can be used to add an authentication layer to your shiny apps.
It borrows some code from treysp's shiny_password template with the goal of making implementation simpler for end users and allowing the login/logout UIs to fit easily into any UI framework, including shinydashboard. To enable cookie-based authentification in browsers, it also borrows code from calligross's Shiny Cookie Based Authentication Example. See live example app here and code in the inst directory.
devtools::install_github("paulc91/shinyauthr")The package provides 2 module functions each with a UI and server element:
loginloginUIlogoutlogoutUI
Below is a minimal reproducible example of how to use the authentication modules in a shiny app. Note that you must initiate the use of the shinyjs package with shinyjs::useShinyjs() in your UI code for this to work appropriately.
library(shiny)
library(shinyauthr)
library(shinyjs)
# required javascript code
jsCode <- '
shinyjs.getcookie = function(params) {
var cookie = Cookies.get("id");
if (typeof cookie !== "undefined") {
Shiny.onInputChange("jscookie", cookie);
} else {
var cookie = "";
Shiny.onInputChange("jscookie", cookie);
}
}
shinyjs.setcookie = function(params) {
Cookies.set("id", escape(params), { expires: 0.5 });
Shiny.onInputChange("jscookie", params);
}
shinyjs.rmcookie = function(params) {
Cookies.remove("id");
Shiny.onInputChange("jscookie", "");
}
'
# dataframe that holds usernames, passwords and other user data
user_base <- data.frame(
user = c("user1", "user2"),
password = c("pass1", "pass2"),
permissions = c("admin", "standard"),
name = c("User One", "User Two"),
cookie = c("", ""),
stringsAsFactors = FALSE
)
ui <- fluidPage(
# enable cookie based authentification
tags$head(
tags$script(src = "https://cdn.jsdelivr.net/npm/js-cookie@2/src/js.cookie.min.js",
type = "text/javascript")),
# must turn shinyjs on and enable the js code
shinyjs::useShinyjs(),
shinyjs::extendShinyjs(text = jsCode),
# add logout button UI
div(class = "pull-right", shinyauthr::logoutUI(id = "logout")),
# add login panel UI function
shinyauthr::loginUI(id = "login"),
# status
textOutput("status"),
# setup table output to show user info after login
tableOutput("user_table")
)
server <- function(input, output, session) {
# call the logout module with reactive trigger to hide/show
logout_init <- callModule(shinyauthr::logout,
id = "logout",
active = reactive(credentials()$user_auth))
# call login module supplying data frame, user and password cols
# and reactive trigger
credentials <- callModule(shinyauthr::login,
id = "login",
data = user_base,
user_col = user,
pwd_col = password,
cookie_col = cookie,
log_out = reactive(logout_init()))
# pulls out the user information returned from login module
user_data <- reactive({credentials()$info})
output$status <- renderText({
if (credentials()$user_auth) {
paste0('You are logged on with a cookie value of:\n',
user_data()$cookie)
} else {
paste0('You are logged out')
}
})
output$user_table <- renderTable({
# use req to only render results when credentials()$user_auth is TRUE
req(credentials()$user_auth)
user_data()
})
}
shinyApp(ui = ui, server = server)
When the login module is called, it returns a reactive list containing 2 elements:
user_authinfo
The initial values of these variables are FALSE and NULL respectively. However,
given a data frame or tibble containing user names, passwords, cookies and other user data (optional), the login module will assign a user_auth value of TRUE if the user supplies a matching user name and password. The value of info then becomes the row of data associated with that user which can be used in the main to control content based on user permission variables etc.
The logout button will only show when user_auth is TRUE. Clicking the button will reset user_auth back to FALSE which will hide the button and show the login panel again.
You can set the code in your server functions to only run after a successful login through use of the req() function inside all reactives, renders and observers. In the example above, using req(credentials()$user_auth) inside the renderTable function ensures the table showing the returned user information is only rendered when user_auth is TRUE.
If you are hosting your user passwords on the internet, it is a good idea to first encrypt them with a hashing algorithm. You can use the digest package to do this. You can then tell the shinyauthr::login module that your passwords are hashed and what algorithm you used when hashing with digest. Your plain text passwords must be a character vector, not factors, when hashing for this to work as shiny inputs are passed as character strings.
For example, a sample user base like the following can be incorporated for use with shinyauthr:
# create a user base then hash passwords with md5 algorithm
# then save to an rds file in app directory
library(digest)
user_base <- data.frame(
user = c("user1", "user2"),
password = sapply(c("pass1", "pass2"), digest, "md5"),
permissions = c("admin", "standard"),
name = c("User One", "User Two"),
cookie = c("", ""),
stringsAsFactors = FALSE
)
saveRDS(user_base, "user_base.rds")# in your app code, read in the user base rds file
user_base <- readRDS("user_base.rds")# then when calling the module, set hashed = TRUE and algo = "md5"
credentials <- callModule(shinyauthr::login, "login",
data = user_base,
user_col = user,
pwd_col = password,
cookie_col = cookie,
hashed = TRUE,
algo = "md5",
log_out = reactive(logout_init()))I'm not a security professional so cannot guarantee this authentication procedure to be foolproof. It is ultimately the shiny app developer's responsibility not to expose any sensitive content to the client without the necessary login criteria being met.
I would welcome any feedback on any potential vulnerabilities in the process. I know that apps hosted on a server without an SSL certificate could be open to interception of usernames and passwords submitted by a user. As such I would not recommend the use of shinyauthr without an HTTPS connection.
For apps intended for use within commercial organisations, I would recommend one of RStudio's commercial shiny hosting options with built in authetication.
However, I hope that having an easy-to-implement open-source shiny authentication option like this will prove useful when alternative options are not feasible.
Paul Campbell
October 2018