Computer Security SEED Labs 1: Transport Layer Security (TLS) Lab
SEED Labs
1
Transport Layer Security (TLS) Lab
Copyright © 2020 by Wenliang Du.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International
License. If you remix, transform, or build upon the material, this copyright notice must be left intact, or
reproduced in a way that is reasonable to the medium in which the work is being re-published.
1 Overview
Nowadays more and more data transmissions are done through the Internet. However, when data are trans- mitted over such a public network unprotected, they can be read or even modified by others. Applications worrying about the security of their communication need to encrypt their data and detect tampering. Cryp- tographic solutions can be used to achieve this goal. There are many cryptographic algorithms, and even for the same algorithm, there are many parameters that can be used. To achieve interoperability, i.e., allowing different applications to communicate with one another, these applications need to follow a common stan- dard. TLS, Transport Layer Security, is such a standard. Most web servers these days are using HTTPS, which is built on top of TLS.
The objective of this lab is to help students understand how the TLS works and how to use TLS in pro- gramming. The lab guides students to implement a pair of TLS client and server programs, based on which students will conduct a series of experiments, so they can understand the security principles underlying the TLS protocol. Students will also implement a simple HTTPS proxy program to understand the security impact if some trusted CAs are compromised. The lab covers the following topics:
• Public-Key Infrastructure (PKI)
• Transport Layer Security (TLS)
• TLS programming
• HTTPS proxy
• X.509 certificates with the Subject Alternative Name (SAN) extensions • Man-In-The-Middle attacks
Prerequisite. This lab depends on the PKI lab. Students should do the PKI lab before working on this lab. Readings. Detailed coverage of PKI and TLS can be found in the following:
• Chapters 24 and 25 of the SEED Book, Computer & Internet Security: A Hands-on Approach, 2nd Edition, by Wenliang Du. See details at https://www.handsonsecurity.net.
Lab Environment. This lab has been tested on the SEED Ubuntu 20.04 VM. You can download a pre- built image from the SEED website, and run the SEED VM on your own computer. However, most of the SEED labs can be conducted on the cloud, and you can follow our instruction to create a SEED VM on the cloud.
SEED Labs 2
2 Lab Environment
In this lab, we use three machines, one for client, one for server, and the other for proxy. We will use containers for these machines. Their IP addresses are listed in the following:
Container Setup and Commands. Please download the Labsetup.zip file to your VM from the lab’s website, unzip it, enter the Labsetup folder, and use the docker-compose.yml file to set up the lab environment. Detailed explanation of the content in this file and all the involved Dockerfile can be found from the user manual, which is linked to the website of this lab. If this is the first time you set up a SEED lab environment using containers, it is very important that you read the user manual.
In the following, we list some of the commonly used commands related to Docker and Compose. Since we are going to use these commands very frequently, we have created aliases for them in the .bashrc file (in our provided SEEDUbuntu 20.04 VM).
All the containers will be running in the background. To run commands on a container, we often need to get a shell on that container. We first need to use the "docker ps" command to find out the ID of thecontainer,andthenuse"docker exec"tostartashellonthatcontainer.Wehavecreatedaliasesfor them in the .bashrc file.
client: 10.9.0.5
server: 10.9.0.43
proxy: 10.9.0.143
$ docker-compose build # Build the container image
$ docker-compose up # Start the container
$ docker-compose down # Shut down the container
|
// Aliases for the Compose commands above
$ dcbuild
$ dcup
$ dcdown
# Alias for: docker-compose build
# Alias for: docker-compose up
# Alias for: docker-compose down
|
$ dockps // Alias for: docker ps --format "{{.ID}} {{.Names}}"
$ docksh <id> // Alias for: docker exec -it <id> /bin/bash
|
// The following example shows how to get a shell inside hostC
$ dockps
b1004832e275 hostA-10.9.0.5
0af4ea7a3e2e hostB-10.9.0.6
9652715c8e0a hostC-10.9.0.7
|
$ docksh 96
root@9652715c8e0a:/#
// Note: If a docker command requires a container ID, you do not need to
|
If you encounter problems when setting up the lab environment, please read the “Common Problems” section of the manual for potential solutions.
SEED Labs 3
Volumes. Code editing is more convenient inside the VM than in containers, because we can use our favorite editors. In order for the VM and container to share files, we have created a shared folder between the VM and the container using the Docker volumes. If you look at the Docker Compose file, you will find out that we have added the following entry to some of the containers. It indicates mounting the ./volumes folder on the host machine (i.e., the VM) to the /volumes folder inside the container. We will write our code in the ./volumes folder (on the VM), so they can be used inside the containers.
3 Task 1: TLS Client
In this task, we will incrementally build a simple TLS client program. Through the process, students will understand the essential elements and security considerations in TLS programming. We will run this client program on the client container.
3.1 Task 1.a: TLS handshake
Before a client and a server can communicate securely, several things need to be set up first, including what encryption algorithm and key will be used, what MAC algorithm will be used, what algorithm should be used for the key exchange, etc. These cryptographic parameters need to be agreed upon by the client and the server. That is the primary purpose of the TLS Handshake Protocol. In this task, we focus on the TLS handshake protocol. The following sample code initiates a TLS handshake with a TLS server (the name of the server needs to be specified as the first command line argument).
Listing 1: handshake.py (in Labsetup/volumes)
|
volumes: |
- ./volumes:/volumes
|
#!/usr/bin/env python3
|
import socket, ssl, sys, pprint hostname = sys.argv[1] port = 443 cadir = ’/etc/ssl/certs’ |
# Set up the TLS context
context = ssl.SSLContext(ssl.PROTOCOL_TLS_CLIENT)
context.load_verify_locations(capath=cadir)
context.verify_mode = ssl.CERT_REQUIRED
context.check_hostname = True
|
# Create TCP connection
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM) sock.connect((hostname, port)) input("After making TCP connection. Press any key to continue ...") |
|
# Add the TLS ssock = context.wrap_socket(sock, server_hostname=hostname,
do_handshake_on_connect=False)
ssock.do_handshake() # Start the handshake pprint.pprint(ssock.getpeercert()) |
SEED Labs 4
input("After handshake. Press any key to continue ...") |
# Close the TLS Connection
ssock.shutdown(socket.SHUT_RDWR)
ssock.close()
|
Tasks. Use the code above to communicate with a real HTTPS-based web server. Additional code may need to be added to complete the tasks. Students can find the manual for Python’s SSL module online. Please report the following:
3.2
• • • •
What is the cipher used between the client and the server? Please print out the server certificate in the program. Explain the purpose of /etc/ssl/certs.
Use Wireshark to capture the network traffics during the execution of the program, and explain your observation. In particular, explain which step triggers the TCP handshake, and which step triggers the TLS handshake. Explain the relationship between the TLS handshake and the TCP handshake.
Task 1.b: CA’s Certificate
In the previous task, we use the certificates in the /etc/ssl/certs folder to verify server’s certificates. In this task, we will create our own certificate folder, and place the corresponding certificates in the folder to do the verification.
Please create a folder called client-certs, and change the cadir line in the client program to the following. Run the client program and report your observation.
cadir = ’./client-certs’
To solve this problem, you need to place the corresponding CA’s certificate into your client-certs
folder. Please use your client program to find out what CA certificate is needed to verify the www.example.com
server’s certificate, and then copy the certificate from the /etc/ssl/certs to your own folder. Run your
client program again. If you have done everything correctly, your client program should be able to talk to
the server.
It should be noted that copying CA’s certificate to the "./client-certs" folder is not enough. When TLS tries to verify a server certificate, it will generate a hash value from the issuer’s identify infor- mation, use this hash value as part of the file name, and then use this name to find the issuer’s certificate in the "./client-certs" folder. Therefore, we need to rename each CA’s certificate using the hash value generated from its subject field, or we can make a symbolic link out of the hash value. In the following command, we use opensslto generate a hash value, which is then used to create a symbolic link.
$ openssl x509 -in someCA.crt -noout -subject_hash
4a6481c9 $ ln -s someCA.crt 4a6481c9.0 $ ls -l |
total 4
lrwxrwxrwx 1 ... 4a6481c9.0 -> someCA.crt
-rw-r--r-- 1 ... someCA.crt
|
SEED Labs 5
Additional requirement: Please conduct this task for two different web servers that use different CA certificates.
3.3 Task 1.c: Experiment with the hostname check
The objective of this task is to help students understand the importance of hostname checks at the client side. Please conduct the following steps using the client program.
-
Step 1: Get the IP address of www.example.com using the dig command, such as the following (you may want to run this command in your VM or host computers, because the dig command is not installed inside the container):
-
Step 2: Modify the /etc/hosts file (inside the container), add the following entry at the end of the file (the IP address is what you get from the dig command).
93.184.216.34 www.example2020.com -
Step 3: Switch the following line in the client program between True and False, and then connect your client program to www.example2020.com. Describe and explain your observation.
context.check_hostname = False # try both True and FalseThe importance of hostname check: Based on this experiment, please explain the importance of host- name check. If the client program does not perform the hostname check, what is the security consequence? Please explain.
3.4 Task 1.d: Sending and getting Data
In this task, we will send data to the server and get its response. Since we choose to use HTTPS servers, we need to send HTTP requests to the server; otherwise, the server will not understand our request. The following code example shows how to send HTTP requests and how to read the response.
$ dig www.example.com
...
;; ANSWER SECTION:
www.example.com. 403 IN A 93.184.216.34
# Send HTTP Request to Server
request = b"GET / HTTP/1.0\r\nHost: " + \ hostname.encode(’utf-8’) + b"\r\n\r\n" |
ssock.sendall(request)
# Read HTTP Response from Server
response = ssock.recv(2048) while response: |
pprint.pprint(response.split(b"\r\n")) response = ssock.recv(2048) |
SEED Labs 6
Tasks. (1) Please add the data sending/receiving code to your client program, and report your observation. (2) Please modify the HTTP request, so you can fetch an image file of your choice from an HTTPS server (there is no need to display the image).
4 Task 2: TLS Server
Before working on this task, students need to create a certificate authority (CA), and use this CA’s private key to create a server certificate for this task. How to do these is already covered in another SEED lab (the PKI lab), which is the prerequisite for this lab. In this task, we assume that all the required certificates have already been created, including CA’s public-key certificate and private key (ca.crt and ca.key), and the server’s public-key certificate and private key (server.crt and server.key). It should be noted that the common name used in the server certificate must contain the student’s last name and the current year.
We will use the server container to run this TLS server program. Make sure you set up the DNS mapping accordingly, so the name of your TLS server points to the IP address of the server container.
4.1 Task 2.a. Implement a simple TLS server
In this task, we will implement a simple TLS server. We will use the client program from Task 1 to test this server program. A sample server code is provided in the following.
Listing 2: server.py (in Labsetup/volumes)
#!/usr/bin/env python3
import socket |
|
import ssl html = """ HTTP/1.1 200 OK\r\nContent-Type: text/html\r\n\r\n <!DOCTYPE html><html><body><h1>Hello, world!</h1></body></html> """ |
SERVER_CERT = ’./server-certs/server.crt’ SERVER_PRIVATE = ’./server-certs/server.key’ context = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER)
|
context.load_cert_chain(SERVER_CERT, SERVER_PRIVATE)
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM, 0) sock.bind((’0.0.0.0’, 443)) sock.listen(5) |
while True: newsock, fromaddr = sock.accept() ssock = context.wrap_socket(newsock, server_side=True) data = ssock.recv(1024) # Read data over TLS ssock.sendall(html.encode(’utf-8’)) # Send data over TLS |
ssock.shutdown(socket.SHUT_RDWR) # Close the TLS connection ssock.close() |
SEED Labs 7
Testing. We will use the client program developed in Task 1 to test this server program. In Task 1, the client program loads the trusted certificates from the /etc/ssl/certs folder. In this task, the CA is created by us, and its certificate is not stored in that folder. We do not recommend students to add this CA to that folder, because that will affect the entire system. Students should store the CA’s certificate in the "./client-certs" folder, and then follow the instruction in Task 1 to set up the folder. Please test your program using the /etc/ssl/certs folder and the ./client-certs folder, respectively. Please describe your observation and explain why.
4.2 Task 2.b. Testing the server program using browsers
In this task, we will test our TLS server program using the browser on the host VM. First, point your browser to the server, report what you see from the browser and explain why. The server listens to port 443, which is the default port for HTTPS.
In order for browsers to communicate with your TLS server, the browser needs to verify your server’s certificate. It has to use the certificate issuer CA’s certificate to do the verification, but since this CA is created in our lab, the browser does not have it on its trusted certificate list. We need to manually add our CA’s certificate to it. To do that, type the following URL in the address bar, and click the View Certificates button on the page (scroll to the bottom).
about:preferences#privacy
In the Authorities tab, you will see a list of certificates that are already accepted by Firefox. From here, we can import our own certificates. After choosing the certificate file, please select the following option: “Trust this CA to identify web sites”. You will see that our certificate is now in Firefox’s list of accepted certificates.
Please demonstrate that your browser can successfully communicate with your TLS server, and can display the content returned by the server.
4.3 Task 2.c. Certificate with multiple names
Many websites have different URLs. For example, www.example.com, www.example.org, example. com all point to the same web server. Due to the hostname matching policy enforced by most TLS client programs, the common name in a certificate must match with the server’s hostname, or TLS clients will refuse to communicate with the server.
To allow a certificate to have multiple names, the X.509 specification defines extensions to be attached to a certificate. This extension is called Subject Alternative Name (SAN). Using the SAN extension, it’s possible to specify several hostnames in the subjectAltName field of a certificate.
To generate a certificate signing request with such a field, we can use a configuration file, and put all the necessary information in this file (the PKI lab shows how you can do everything in the command line). The following configuration file gives an example. It specifies the content for the subject field and add a subjectAltName field in the extension. The field specifies several alternative names, including a wildcard name *.bank32.com. It should be noted that the field must also include the one from the common name field; otherwise, the common name will not be accepted as a valid name.
Listing 3: server openssl.cnf
[ req ]
prompt = no
|
distinguished_name = req_distinguished_name
req_extensions = req_ext
|
SEED Labs 8
[ req_distinguished_name ]
|
|
C = US |
[ req_ext ]
subjectAltName = @alt_names
[alt_names]
|
DNS.3 = *.bank32.com |
We can use the following "openssl req" command to generate a pair of public/private keys and a certificate signing request:
When the CA signs a certificate, for the security reason, by default, it does not copy the extension field from the certificate signing request into the final certificate. In order to allow the copying, we need to change the openssl’s configuration file. By default, openssl uses the configuration file openssl.cnf from the /usr/lib/ssl directory. Inside this file, the copy extensions option is disabled (commented out). We do not want to modify this system-wide configuration file. Let us copy it file to our own folder, and rename it as myopenssl.cnf. We then uncomment the following line from this file:
Now, we can use the following program to generate the certificate (server.crt) for the server from the certificate signing request (server.csr), and all the extension fields from the request will be copied to the final certificate.
Students need to demonstrate that their server can support multiple hostnames, including any hostname in their respective domains.
5 Task 3: A Simple HTTPS Proxy
TLS can protect against the Man-In-The-Middle attack, but only if the underlying public-key infrastructure is secured. In this task, we will demonstrate the Man-In-The-Middle attack against TLS servers if the PKI infrastructure is compromised, i.e., some trusted CA is compromised or the server’s private key is stolen.
We will implement a simple HTTPS proxy called mHTTPSproxy (m stands for mini). The proxy program simply integrates the client and server programs from Task 1 and 2 together. How it works is illustrated in Figure 1. We will run the proxy on the proxy container.
The proxy is actually a combination of the TLS client and server programs. To the browser, the TLS
openssl req -newkey rsa:2048 -config ./server_openssl.cnf -batch \
-sha256 -keyout server.key -out server.csr
# Extension copying option: use with caution.
|
copy_extensions = copy
|
openssl ca -md sha256 -days 3650 -config ./myopenssl.cnf -batch \
-in server.csr -out server.crt \
|
-cert ca.crt -keyfile ca.key
|
SEED Labs
9
Browser (Client)
HTTPS Proxy
HTTPS Server
HTTP Request
Figure 1: How mHTTPSproxy works
Forward
Forward
proxy is just a server program, which takes the HTTP requests from the browser (the client), and return HTTP responses to it. The proxy does not generate any HTTP responses; instead, it forwards the HTTP requests to the actual web server, and then get the HTTP responses from the web server. To the actual web server, the TLS proxy is just a client program. After getting the response, the proxy forwards the response to the browser, the real client. Therefore, by integrating the client and server programs implemented in the previous two tasks, students should be able to get a basic proxy working.
It should be noted that the purpose of this task is to use this simple proxy to understand how the Man- In-The-Middle attack works when the PKI infrastructure is compromised. It is not intended to implement a product-quality HTTPS proxy, because making the proxy work for every web server is not an easy job, as many aspects of the HTTP protocol need to be considered. Since the focus of this lab is on TLS, students can choose two different servers, and demonstrate that their proxy works for those servers. Students who are interested in product-quality HTTPS proxy, can find that from the Internet, such as the open-source mitmproxy.
Handling multiple HTTP requests. A browser may simultaneously send multiple HTTP requests to the server, so after receiving an HTTP request from the browser, it is better to spawn a thread to process that request, so the proxy program can handle multiple simultaneous requests. The following code snippet shows how to create a thread to handle each TLS connection.
The thread will execute the code in the process request function, which forwards the HTTP re- quest from the browser to the server, and then forward the HTTP response from the server to the browser. A code skeleton is provided in the following:
HTTP Response
import threading while True: sock_for_browser, fromaddr = sock_listen.accept() ssock_for_browser = context_srv.wrap_socket(sock_for_browser, server_side=True)
|
x = threading.Thread(target=process_request, args=(ssock_for_browser,))
x.start()
|
def process_request(ssock_for_browser): |
hostname = ’www.example.com’ |
SEED Labs 10
# Make a connection to the real server
|
sock_for_server = socket.create_connection((hostname, 443)) ssock_for_server = ... # [Code omitted]: Wrap the socket using TLS request = ssock_for_browser.recv(2048) if request: |
# Forward request to server
ssock_for_server.sendall(request)
# Get response from server, and forward it to browser
response = ssock_for_server.recv(2048) while response: |
ssock_for_browser.sendall(response) # Forward to browser response = ssock_for_server.recv(2048) ssock_for_browser.shutdown(socket.SHUT_RDWR)
ssock_for_browser.close()
|
The client setup. For this task, since we will use a browser, we will use the hosting VM as the client/vic- tim, instead of using the client container. In the real-world attack, when the victim tries to visit a web server (say www.example.com), we will launch attacks to redirect the victim to our proxy. This is usually done by DNS attacks, BGP attacks, or other redirection attacks. We will not actually do such attacks. We simply add the following entry to the /etc/hosts file on the host VM (10.9.0.143 is the IP address of the mitm-proxy container in our setup).
10.9.0.143 www.example.com
By doing the above, we simulate the result of redirection attacks: the victim’s traffic to the web server www.example.com will be redirected to the attacker’s machine, where your mHTTPSproxy is waiting for HTTP requests.
The proxy setup. It should be noted that changes on the VM’s /etc/hosts file will affect all the containers, because containers uses Docker’s embedded DNS server, which forwards external DNS lookups to the DNS servers configured on the host (i.e., the VM), and the host’s DNS server do use the /etc/hosts file. Therefore, due to the change above, on the proxy container, the IP address to www.example.com is also mapped to 10.9.0.143. This is a problem, because the proxy needs to communicate with the actual web server.
To do so, we can ask the container to use an external DNS server, instead of using its built-in one. This can be done by manually changing the /etc/resolv.conf file on the proxy container (not the one on the VM). The file has one or multiple nameserver entry. Change the first one to 8.8.8.8, which is the public DNS server provided by Google. You can also use other public DNS servers.
nameserver 8.8.8.8
A more permanent way to solve this problem is to add the following DNS entry to the MITM-Proxy’s configuration in the docker-compose.yml file. This entry tells the container’s built-in DNS server to forward DNS requests to 8.8.8.8.
