The analysis also is not restricted to use cases that can achieved using PSI. We use PSI since the malicious detection of the web-page is based on blacklist. However, the analyis might be more sophisticated and involve machine learning. There is research focusing on secure computation for machine learning 35. Some of these techniques can also be used if the use case requires it.
Advantages of Our Proposed Anonymous Network. We summarize the advantages of proposed protocol as compared to Tor as follows: (i) Decreased Criminal Activity: Since all service providers who provide anonymous services have to be licensed, the incidence of criminal activity will be largely reduced. (ii) Difficult to Monitor: Any monitoring of the network will be very difficult, as it will involve monitoring the entire Internet. To a casual observer a packet traversing the proposed anonymous network will look like any other encrypted packet. (iii) Timing Attacks: Certain kinds of timing attacks that rely on the user choosing relays that belong to same AS that are possible in Tor, will not be possible in our proposed network since it is not an overlay network. The need for specialized exit relays is also eliminated. (iv) Easy to Use: Unlike Tor, no setup will be required on the part of the user. The user only needs to be aware of the service and they should be able to access it seamlessly from any browser.
In this paper, we propose a protocol that simultaneously addresses anonymity and analyzability of user data on the Internet. The main techniques used is the Secure Computation (i.e., the Private Set Intersection (PSI)), Onion routing and digital certificates. The protocol allows the ISP to forward web requests from its users to licensed web servers without knowing the web-page but still being able to analyze the web-page. The implementation shows that our protocol is practical. In this paper, we focus only on a use case of analysis which is a malicious detection based on blacklist by using PSI. However, there are many types analyses that can be carried out by the ISP with different data types, not just web browsing. This paper has opened the first step for thoughts on how to strike a balance between the user privacy and any analyzability of user data on the Internet. Future works will continue investigating the possibility of carrying out different kinds of analysis on the web data and more realistic simulations of the protocol.