| 1 |
<sect1 id="security"> |
| 2 |
|
| 3 |
<title>HTTP Client Security</title> |
| 4 |
|
| 5 |
<para>&neon; is intended to be secure against a specific threat |
| 6 |
model: use of a malicious HTTP server. Under this threat model, a |
| 7 |
range of attacks are possible against a client when the user (or |
| 8 |
application) can be tricked into accessing an HTTP server which is |
| 9 |
controlled by an attacker. This section documents various types of |
| 10 |
possible attack and describes what mitigation is used in |
| 11 |
&neon;.</para> |
| 12 |
|
| 13 |
<sect2> |
| 14 |
<title>CPU or memory consumption attacks</title> |
| 15 |
|
| 16 |
<para>&neon; uses fixed resource limits to prevent the following |
| 17 |
attacks:</para> |
| 18 |
|
| 19 |
<itemizedlist> |
| 20 |
<listitem> |
| 21 |
<para>memory/CPU consumption attack using an unbounded number |
| 22 |
of response header fields</para> |
| 23 |
</listitem> |
| 24 |
|
| 25 |
<listitem> |
| 26 |
<para>memory consumption attack using an unbounded length of |
| 27 |
individual response header lines (or continuation |
| 28 |
headers)</para> |
| 29 |
</listitem> |
| 30 |
|
| 31 |
<listitem> |
| 32 |
<para>memory consumption attack against the PROPFIND code |
| 33 |
using an unbounded number of properties (propstat elements) |
| 34 |
per resource</para> |
| 35 |
</listitem> |
| 36 |
|
| 37 |
<listitem> |
| 38 |
<para>memory consumption attack against the PROPFIND code |
| 39 |
using an unbounded CDATA element in a "flat property" |
| 40 |
value</para> |
| 41 |
</listitem> |
| 42 |
</itemizedlist> |
| 43 |
|
| 44 |
<para>Memory consumption attacks against applications based on |
| 45 |
&neon; by use of unbounded response length are also possible, but |
| 46 |
must be mitigated at application level. Memory consumption in |
| 47 |
&neon; itself is fixed and is not proportional to the response |
| 48 |
size.</para> |
| 49 |
|
| 50 |
<para>Test cases for all the above attacks are present in the |
| 51 |
&neon; test suite.</para> |
| 52 |
|
| 53 |
</sect2> |
| 54 |
|
| 55 |
<sect2> |
| 56 |
<title>SSL/TLS connection security</title> |
| 57 |
|
| 58 |
<para>When using a connection secured by SSL/TLS, it is necessary |
| 59 |
for clients to verify that the X.509 certificate presented by the |
| 60 |
server matches the server's expected identity. The algorithm |
| 61 |
required for this purpose is described in RFC 2818 and RFC 3280, |
| 62 |
and is implemented by &neon; in the following manner:</para> |
| 63 |
|
| 64 |
<itemizedlist> |
| 65 |
<listitem> |
| 66 |
<para>the hostname argument passed to <xref |
| 67 |
linkend="ne_session_create"/> is the expected identity of the |
| 68 |
server</para> |
| 69 |
</listitem> |
| 70 |
|
| 71 |
<listitem> |
| 72 |
<para>the subjectAltName extension of the certificate is used |
| 73 |
for comparision against the expected identity, in preference |
| 74 |
to the Subject name's commonName attribute.</para> |
| 75 |
</listitem> |
| 76 |
|
| 77 |
<listitem> |
| 78 |
<para>the dNSName, iPAddress, and uniformResourceIdentifier |
| 79 |
classes of GeneralName are supported in subjectAltName |
| 80 |
comparison.</para> |
| 81 |
</listitem> |
| 82 |
|
| 83 |
<listitem> |
| 84 |
<para>if no subjectAltName is present in the certificate, the |
| 85 |
most specific commonName attribute in the Subject name is used |
| 86 |
for comparison instead.</para> |
| 87 |
</listitem> |
| 88 |
</itemizedlist> |
| 89 |
|
| 90 |
<para>In the case where a server certificate is presented that |
| 91 |
does not match the expected identity (or is otherwise not |
| 92 |
trusted), &neon; will fail the request by default. This behaviour |
| 93 |
can be overridden by the use of a callback installed using <xref |
| 94 |
linkend="ne_ssl_set_verify"/>, which allows the application to |
| 95 |
present the certificate details to a user for manual/off-line |
| 96 |
verification, if possible.</para> |
| 97 |
|
| 98 |
<para>Test cases for the correctness of the implementation of the |
| 99 |
identity verification algorithm are present in the &neon; test |
| 100 |
suite.</para> |
| 101 |
|
| 102 |
</sect2> |
| 103 |
|
| 104 |
<sect2> |
| 105 |
<title>Control character insertion in error messages</title> |
| 106 |
|
| 107 |
<para>Where error messages (as returned by |
| 108 |
(<xref linkend="ne_get_error"/>) contain data supplied by the |
| 109 |
server, the untrusted data is sanitised to remove both control |
| 110 |
characters and non-ASCII characters. This prevents any attacks |
| 111 |
where such error messages are exposed to the user and can |
| 112 |
potentially distort the presentation of the interface (for |
| 113 |
example, through the use of a carriage return character in a text |
| 114 |
user interface).</para> |
| 115 |
</sect2> |
| 116 |
|
| 117 |
<sect2> |
| 118 |
<title>Attacks against authentication credentials</title> |
| 119 |
|
| 120 |
<para>Authentication credentials can be compromised by a |
| 121 |
"downgrade attack" by an active attacker; for example, where a |
| 122 |
MITM presents a Basic authentication challenge in place of the |
| 123 |
server's Digest challenge. &neon; mitigates these attacks by |
| 124 |
allowing the application (and hence, user) to specify that only a |
| 125 |
specific set of authentication protocols is permitted.</para> |
| 126 |
|
| 127 |
<para>&neon; supports the Digest and Negotiate authentication |
| 128 |
schemes, which both allow authentication of users without passing |
| 129 |
credentials in cleartext over the wire. The "domain" parameter is |
| 130 |
supported in Digest, allowing the server to restrict an |
| 131 |
authentication session to a particular set of URIs.</para> |
| 132 |
|
| 133 |
</sect2> |
| 134 |
|
| 135 |
</sect1> |