Quarantyne - Modern Web Firewall: Stop Account Takeovers, Weak Passwords, Cloud IPs, DoS Attacks, Disposable Emails

Automated web security made simple
Quarantyne is a reverse-proxy that protects web applications and APIs from fraudulent behavior, misuse, bots and cyber-attacks in real-time.

Requirements

• Java 8

Presentation
Quarantyne is a reverse-proxy written in java. It fronts a web application or API and protects it from fraudulent behavior, misuse, bots and cyber-attacks. It cannot stop them all, but it will definitely make it harder and more expensive to perform.
It's like a firewall but smarter, because it does not just block traffic because the user-agent is not in a whitelist. Quarantyne also performs deep request inspection to detect if, for example, the password used has been compromised before, or if the email is disposable, with minimal configuration and no changes in your application. Our coverage section precisely lists what Quarantyne can identify.

Features

Wide coverage of common HTTP threats and misuse
See coverage for a complete list of the threats and misuse Quarantyne can identify and stop.

Deep traffic analysis
Quarantyne performs deep inspection of web traffic going to your application to verify that the data being sent is not compromised or junk.

Generic integration
Quarantyne adds extra HTTP headers to the request it proxies to your service. For example, an HTTP request coming from AWS will bear the following headers:

• X-Quarantyne-Labels: PCX
• X-Quarantyne-RequestId: 08a0e31a-f1a5-4660-9316-0fdf5d2a959d

Active protection
Quarantyne can be configured to stop malicious requests from reaching your servers, avoiding wasting computing/DB/cache resources, metrics skew, junk data... See (Passive vs Active)[#passivevsactive].

Metrics & health reporting
Quarantyne binds to an internal adminPort, where metrics (latencies, success rate...) as well as the health of the proxy are reported.

Privacy friendly / GDPR compliance
Quarantyne is offline software. It runs inside your private network and does not communicate over the Internet with anyone to share data about your traffic, your business, or your users.

Ops Friendly.
Single jar with 0 dependencies. Metrics are available on [proxyHost]:[adminPort]/metrics. Service health is available on [proxyHost]:[adminPort]/health

Coverage
Quarantyne is able to detect the following threats and misuse.

Label	Definition	Behavior	Implemented
LBD	Large Body Data	Overload target's form processor with POST/PUT request with body > 1MB	yes
FAS	Fast Browsing	Request rate faster than regular human browsing	yes
CPW	Compromised Password	Password used is known from previous data breach. Possible account takeover	yes
DMX	Disposable Email	Email used is a disposable emails service	yes
IPR	IP Address Rotation	Same visitor is rotating its IP addresses	no
SHD	Suspicious Request Headers	Abnormal HTTP Request headers	yes
SUA	Suspicious User-Agent	User Agent not from a regular web browser	yes
PCX	Public Cloud Execution	IP address belongs to a public cloud service like AWS or GCP	no
IPD	IP/Country discrepancy	Country inferred from visitor IP is different from country field in submitted request	no
SGE	Suscpicious Geolocation	This request is not usually received from this geolocation. Possible account takeover.	no

Passive vs. Active

Passive mode
Quarantyne lets you decide how you want to handle requests it flags. Quarantyne's default configuration is to NOT block tainted traffic. This traffic will make its way to your server and will be labelled as such via HTTP headers.
Passive mode is the recommended way to get familiar with Quarantyne and to get a sense of what's going on inside your web traffic. In your application, log or plot the incoming Quarantyne labels and you might be surprised (or not) by what you find!

Active Mode
In active mode, Quarantyne prevents tainted traffic from reaching your application. Blocking happens only you configure explicitely Quarantyne to do so. The configuration section explains how traffic blocking can be enabled.

Configuration
Two complementary configuration systems are used: command-line arguments and an external (local or remote) JSON configuration file.

Command-line arguments
Run the following command to display the help and what arguments are available

$ java -jar quarantyne -h
Usage: <main class> [options]
  Options:
    --admin
      internal ip:port where to access admin, UI and metrics. Optional
    --config-file
      Optional URL or local path to a Quarantyne JSON configuration file
    --egress
      HTTP destination where Quarantyne forwards annotated web traffic.
      Default: http://httpbin.org
    --help, -help, --h, -h
      Display help about available configuration arguments
      Default: false
    --ingress
      ip:port of inbound web traffic.
      Default: 0.0.0.0:8080

The --config-file is an optional JSON configuration file that tells Quarantyne how requests to your service are structured. It enables deep traffic analysis and increase coverage.

Traffic config JSON file
The traffic config file is optional and can either be an absolute local path or a remote HTTP(S) URL to a JSON file containing a single JSON object with the following structure. Describing the structure of your HTTP requests helps Quarantyne perform deep inspection of critical data such as password, emails or countries.
Quarantyne is able to parse payloads submitted via POST/PUT with a Content-Type of application/json or application/x-www-form-urlencoded.
Root properties are optional.

Property	Definition	Notes
*_action	A POST/PUT data payload	login_action describes the data structure sent when logging in. register_action defines the data structure sent when registering / creating an account.
*_action.path	Path where data is submitted	Must start by /
*_action.identifier_param	Form/JSON key name where the user identifier is sent
*_action.secret_param	Form/JSON key where the user password is sent
email_param_keys	Form/JSON key where email addresses are sent
country_iso_code_param_keys	Form/JSON key where country iso codes are sent
blocked_request_page	HTTP response to return when blocking a request	It's better when this looks like a legit page/error as to not tip off the attack. Even better if you can inject fake data :)
blocked_classes	An array of attack classes to block.	[] is equivalent to passive mode. ['all'] stops every class of attack Quarantyne can detect. See coverage

Quick run

Hosted demo
Available at https://demo.quarantyne.com/. In this scenario, Quarantyne is fronting httpbin.org in passive mode. Threats and misuse sent will be flagged via HTTP headers so querying https://demo.quarantyne.com/headers or https://demo.quarantyne.com/anything is a good start to see what's going on. Hint: start simple, start with curl.

Run the jar
Quarantyne ships as a single 0-dependencies executable jar. Download a release and run:

$ java -jar quarantyne.jar

Build from source
Clone this repo or and run the following

$ ./gradlew run

You should see the following:
You are all set! By default, Quarantyne starts on 127.0.0.1:8080, and proxies traffic to http://httpbin.org.
Send a few requests to http://127.0.0.1:8080/headers via various means. If fraudulent behavior is detected, you should see X-Quarantyne-Label HTTP headers in the request receive by your application. Hint: try with curl.

Distributions

Heroku Buildpack
https://github.com/quarantyne/heroku-buildpack-quarantyne

Docker image
Coming soon

Mailing List
https://groups.google.com/forum/#!forum/quarantyne-users

License
Apache 2