Threats

##Threat Model

The threat model that this is being assumed here is as follows:

• The cheapest attack of Tor will be to look up bridges directly. Employing dozens of people to scour the web for Bridge address will be simpler and more effective than attempting to develop a filter for a hidden protocol.
• Stateful DPI is expensive, and not likely used. Reconstructing entire HTTP streams to verify that it is proper is expensive, but valid.
• Connecting to a generic web server for a long period of time may be considered unusual, while connecting to a image hosting website for a long time is less unusual.
• Fingerprinting of HTTP traffic is difficult. This needs to be revisited with regards to the headers attached to data.
• Image files transferred over HTTP can be verified to be valid images. Manual checking of Image data will not actually happen however, due to cost reasons.
• HTTPS will be blocked.

##Mitigation