[tor-talk] Perceived safety of Tor Browser and onion services

[Philipp Winter]

This is the third part of our preliminary analysis of how Tor users
interact with onion services [0].  In this part, we look at the
subjective feeling of safety that people experience when using Tor
Browser and onion services, respectively.

Question 6.6 in our survey asked:

> Please tell us about how safe you feel when browing onion sites as
> compared to normal websites?

We deliberately did not define "safe," to leave the interpretation up to
our participants.  Here's the breakdown:

                  Safer  Less safe  Neutral
    ---------------------------------------
    Experts      72.24%      6.31%   21.45%
    Non-experts  58.48%     10.71%   30.80%

Experts [1] tend to feel safer on onion services than non-experts [2].

One explanation is that experts' understanding of the underlying
technology gives them more confidence in identifying and fighting off
phishing attacks etc.  Besides, experts may be less influenced by media
reports that focus on the shadier aspects of the "Dark Web."

Another issue is that non-technical users often don't distinguish
between more nuanced aspects of anonymity.  Some of our interview
participants expressed that there's no point in logging in to services
over Tor because "if I log in they know who I am."  Concepts such as
location anonymity, self-authenticating names, and end-to-end encryption
elude them.

The next question in our survey asked our participants to explain their
choice.  Here are the most prevalent themes:

- The underlying technology of onion services (self-authenticating
  names, end-to-end encryption) overwhelmingly made our participants
  feel safe.  People expect a clear improvement in security compared
  to the use of normal websites.

- Orthogonal to the technology, many participants voiced concern about
  illegal, sketchy, and questionable content on onion services.  The
  term "wild west" was used occasionally.  Phishing sites, honeypots,
  and compromised onion sites were also a concern.

- The lack of advertising companies on onion services was mentioned as a
  good thing by several participants.

- A handful of participants complained that it is difficult to know if
  you ended up at the right onion site or a phishing site.

Now let's look at Tor Browser.  Question 6.4 asked:

> How safe do you feel when using Tor Browser compared to another
> browser?

                  Safer  Less safe  Neutral
    ---------------------------------------
    Experts      86.16%      5.97%    7.86%
    Non-experts  83.33%      5.26%   11.40%

Interestingly, the difference between both demographics is significantly
smaller here, presumably in part because Tor Browser is widely seen as a
content-agnostic tool while onion services are frequently associated
with the content that they are perceived to host.

Again, the next question asked our participants to explain their choice.
Here are the most prevalent themes:

- Non-experts lack the ability to evaluate or understand Tor's
  design which is why they defer to expert opinion, their gut feeling,
  or the trust they have in Tor developers.  The Tor Project is
  perceived to focus more on privacy and security than any other browser
  vendor, which many participants appreciated.  Also, its transparency
  further contributes to the trust people have.

- Most of the security criticism focused not on Tor Browser but on the
  underlying Firefox code base.  Many participants were unhappy with the
  exploit mitigation techniques, lack of sandboxing, and the complex
  code base.  Chrome was sometimes brought up as the golden standard for
  browser security.

- Malicious exit relays were a concern for a handful of participants.

- A couple of participants feel safer when using Tor Browser but are
  concerned that their use of Tor makes them stick out and turn into a
  target for government agencies.

- Some participants weren't sure if their Tor setup works properly.
  This is a common theme that we also noticed in our interviews.
  Non-technical users want visual feedback that shows that their network
  traffic comes out "somewhere else."

The above was joint work with my colleagues Marshini Chetty, Annie
Edmundson, Nick Feamster, and Laura M. Roberts.

[0] <https://nymity.ch/onion-services/>

[1] Participants who stated that they are either "highly knowledgeable"
    or an "expert" in Internet security.

[2] Participants who stated that they have either "no knowledge," are
    "mildly knowledgeable," or "moderately knowledgable" in Internet
    security.