httpscreenshot

httpscreenshot

Installation via Docker

docker pull jesseosiecki/httpscreenshot docker run jesseosiecki/httpscreenshot

Installation on Ubuntu

Via Script

Run install-dependencies.sh script as root.

This script has been tested on Ubuntu 20.04 as root (sudo).

Manually

apt-get install swig swig2.0 libssl-dev python-dev python-pip
pip install -r requirements.txt

If you run into: 'module' object has no attribute 'PhantomJS' then pip install selenium (or pip install --upgrade selenium).

If installing on Kali Linux, PhantomJS might not be in the repositories, you can download from https://bitbucket.org/ariya/phantomjs/downloads/phantomjs-1.9.8-linux-x86_64.tar.bz2 and symlink to /usr/bin like so:

sudo ln -s /path/to/phantomjs /usr/bin/phantomjs

README and Use Cases

HTTPScreenshot is a tool for grabbing screenshots and HTML of large numbers of websites. The goal is for it to be both thorough and fast which can sometimes oppose each other.

Before getting into documentation - this is what I USUALLY use for options if I want to screenshot a bunch of sites:

./httpscreenshot.py -i \<gnmapFile\> -p -w 40 -a -vH

Notice there are a ton of worker threads (40). This can be problematic, I make up for failures that could have been a result of too many threads with a second run:

./httpscreenshot.py -i \<gnmapFile\> -p -w 5 -a -vH

YMMV

The options are as follows:

-h, --help show this help message and exit -l LIST, --list LIST List of input URLs -i INPUT, --input INPUT nmap gnmap output file -p, --headless Run in headless mode (using phantomjs) -w WORKERS, --workers WORKERS number of threads -t TIMEOUT, --timeout TIMEOUT time to wait for pageload before killing the browser -v, --verbose turn on verbose debugging -a, --autodetect Automatically detect if listening services are HTTP or HTTPS. Ignores NMAP service detction and URL schemes. -vH, --vhosts Attempt to scrape hostnames from SSL certificates and add these to the URL queue -dB DNS_BRUTE, --dns_brute DNS_BRUTE Specify a DNS subdomain wordlist for bruteforcing on wildcard SSL certs -r RETRIES, --retries RETRIES Number of retries if a URL fails or timesout -tG, --trygui Try to fetch the page with FireFox when headless fails -sF, --smartfetch Enables smart fetching to reduce network traffic, also increases speed if certain conditions are met. -pX PROXY, --proxy PROXY SOCKS5 Proxy in host:port format

Some of the above options have non-obvious use-cases, so the following provides some more detail:

-l, --list -> Takes as input a file with a simple list of input URLs in the format "http(s)://<URL>"

-i, --input -> Takes a gnmap file as input. This includes masscan gnmap output.

-p, --headless -> I find myself using this option more and more. By default the script "drives" Firefox. As the number of threads increases this becomes really ugly - 20,30 Firefox windows open at once. This options uses "phantomjs" which doesn't have a GUI but will still do a decent job parsing javascript.

-w, --workers -> The number of threads to use. Increase for more speed. The list of input URL's is automatically shuffled to avoid hammering at IP addresses that are close to each other when possible. If you add too many threads, you might start seeing timeouts in responses - adjust for your network and machine.

-t TIMEOUT, --timeout -> How long to wait for a response from the server before calling it quits

-v, --verbose -> Will spit out some extra debugging output.

-a, --autodetect -> Without this option enabled, HTTPScreenshot will behave as follows:

If a LIST of urls is specified as input, sites with scheme "http://" are treated as non-ssl and sites with scheme "https://" are treated as ssl-enabled

For GNMAP input the script will scrape input and try to use any SSL detection performed by nmap. Unfortunately this is unreliable, nmap doesn't always like to tell you that something is SSL enabled. Further, masscan doesn't do any version or service detection.

The -a or --autodetect option throws away all SSL hints from the input file and tries to detect on its own.

-vH, --vhosts -> Often when visiting websites by their IP address (e.g: https://192.168.1.30), we will receive a different page than expected or an error. This is because the site is expecting a certain "virtual host" or hostname instead of the IP address, sometimes a single HTTP server will respond with many different pages for different hostnames.

For plaintext "http" websites, we can use reverse DNS, BING reverse IP search etc... to try and find the hostnames associated with an IP address. This is not currently a feature in HTTPScreenshot, but may be implemented later.

For SSL enabled "https" sites, this can be a little easier. The SSL certificate will provide us with a hint at the domain name in the CN field. In the "subject alt names" field of the certificate, when it exists, we may get a whole list of other domain names potentially associated with this IP. Often these are in the form "*.google.com" (wildcard certificate) but sometimes will be linked to a single hostname only like "www.google.com"

The -vH or --vhosts flag will, for each SSL enabled website extract the hostnames from the CN and subject alt names field, and add them to the list of URL's to be screenshotted. For wildcard certificates, the "*." part of the name is dropped.

-dB, --dns_brute -> Must use with -vH for it to make sense. This flag specifies a file containing a list of potential subdomains. For any wildcard certificate e.g: "*.google.com", HTTPScreenshot will try to bruteforce valid subdomains and add them to the list of URLs to be screenshotted.

-r, --retries -> Sometimes Firefox or ghostscript timeout when fetching a page. This could be due to a number of factors, sometimes you just have too many threads going, a network hiccup, etc. This specifies the number of times to "retry" a given host when it fails.

-tG, --trygui -> Upon failure to fetch with the headless browser phantomJS, will pop open FireFox and try again.

-sF, --smartfetch -> Enables smart fetching to reduce network traffic, also increases speed if certain conditions are met.