FastChat

A simple WhatsApp-like chat service. Supports groups and DMs with secure end-to-end encryption. Server architecture supports even load distribution. Was made as part of the CS 251 course project in Autumn 2022, by team Konigsberg.

Detailed documentation can be found here.

Running the Code

To run the Client, execute python client.py <server IP address> To run the Servers, execute python system.py

Change globals to change IP address of server and loadbalancer, number of servers, etc.

Dependencies:

• psycopg2
• cryptography, binascii, base64
• pwnlib
• Sphinx

Here is a brief overview of the code:

Clients

Implement DMs, creating groups, and all other client-side features, like

• Clients can login with single factor authentication from any device.
• Restore their chat history as far back as they wish, on any device.
• Messages can be sent to recipients who are offline, and pending messages can be restored on demand. Program also supports a QuickChat feature where history is not imported and the client can directly message without having to wait for the messages to arrive.
• Groups support message broadcasting to a select set of users.

The Client module handles all client-side requirements like IO with the select module, and personal database management with PostgreSQL. Sockets are handled by the MessageHandler module, which takes care of the message transmission protocol. All cryptography is handled by the crypto module.

End-to-End Encryption

Each user generates his own private-public key pair (RSA). The keys are encrypted with the password and stored on the server database for exporting to other devices. The server doesn't know the passwords since it stores only the SHA256 hash value. Public keys are made available.

The message encryption protocol first generates a symmetric-encryption Fernet key. The message plaintext is encrypted with this, and the key itself is encrypted with the RSA token of the recipient. This encrypted key is appended to the message and sent to all intended recipients, including DMs and Groups.

Messages on the client's personal database are stored with password encryption, and can be securely exported to any device.

Client Database Management

Each client owns a personal database on the 'cloud'. This database is inaccesible to the server architecture, but simultaneously enables easy export of messaging history and contacts to clients across many devices. Messages are stored password encrypted. The database stores a chats table, and a message history table.

Servers

The server architecture consists of a set of servers, each connected to one another. These server programs are launched when the Load Balancer is launched. The load balancer is a special server which knows the host and ports of each server. The client sends a login request to the Load Balancer. Upon valid authentication/new account creation, the generated RSA tokens are preserved, and the client gets added to the servers' database of users. The load balancer returns the least busy server to the client, which the client then connects to.

• The main server loop engages the clients through multiple sockets, and contains sockets to other servers.
• The server knows where each online client is connected, and sends the message to the recipients server.
• Pending messages are maintained on a database, and deleted once popped by request.

Server message protocol is handled by the MessageHandler class, and databasing managed by PostgreSQL.

Server Database Management

• The server architecture maintains a database of FastChat users and their RSA tokens, and their password hashes for authentication.
• It also maintains a database of groups, where the Group members are in a python-list-formatted string.
• Further, it maintains a database of pending messages which were sent when a given client was offline, and can restore them on prompt, then remove them from the pending table.

The Load Balancer system runs the PostgreSQL server for database management.

Load Balancing

The Load Balancer is a special type of Server program that

• Initializes the servers, the (local) PostgreSQL server and the database.
• Authenticates clients, and exports their keys whenever required. It also handles new account creation requests.
• Monitors the traffic on each of the main servers.
• When a valid authentication is complete, the Load Balancer sends to the client the address of the least busy server (three algorithms - least-load, round-robin and naive - are implemented to balance the load), and all further client-server interactions are handled there.
• The load balancer is never under stress of receiving actual content.