df4439186f
Trim trailing whitespace. It doesn't match OpenSSL coding standards, AFAICT, and it can cause problems with git tooling. Trailing whitespace remains in test data and external source. Reviewed-by: Kurt Roeckx <kurt@roeckx.be> Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com> Reviewed-by: Richard Levitte <levitte@openssl.org> (Merged from https://github.com/openssl/openssl/pull/8092)
319 lines
12 KiB
Text
319 lines
12 KiB
Text
HOWTO proxy certificates
|
|
|
|
0. WARNING
|
|
|
|
NONE OF THE CODE PRESENTED HERE HAS BEEN CHECKED! The code is just examples to
|
|
show you how things could be done. There might be typos or type conflicts, and
|
|
you will have to resolve them.
|
|
|
|
1. Introduction
|
|
|
|
Proxy certificates are defined in RFC 3820. They are really usual certificates
|
|
with the mandatory extension proxyCertInfo.
|
|
|
|
Proxy certificates are issued by an End Entity (typically a user), either
|
|
directly with the EE certificate as issuing certificate, or by extension through
|
|
an already issued proxy certificate. Proxy certificates are used to extend
|
|
rights to some other entity (a computer process, typically, or sometimes to the
|
|
user itself). This allows the entity to perform operations on behalf of the
|
|
owner of the EE certificate.
|
|
|
|
See https://www.ietf.org/rfc/rfc3820.txt for more information.
|
|
|
|
|
|
2. A warning about proxy certificates
|
|
|
|
No one seems to have tested proxy certificates with security in mind. To this
|
|
date, it seems that proxy certificates have only been used in a context highly
|
|
aware of them.
|
|
|
|
Existing applications might misbehave when trying to validate a chain of
|
|
certificates which use a proxy certificate. They might incorrectly consider the
|
|
leaf to be the certificate to check for authorisation data, which is controlled
|
|
by the EE certificate owner.
|
|
|
|
subjectAltName and issuerAltName are forbidden in proxy certificates, and this
|
|
is enforced in OpenSSL. The subject must be the same as the issuer, with one
|
|
commonName added on.
|
|
|
|
Possible threats we can think of at this time include:
|
|
|
|
- impersonation through commonName (think server certificates).
|
|
- use of additional extensions, possibly non-standard ones used in certain
|
|
environments, that would grant extra or different authorisation rights.
|
|
|
|
For these reasons, OpenSSL requires that the use of proxy certificates be
|
|
explicitly allowed. Currently, this can be done using the following methods:
|
|
|
|
- if the application directly calls X509_verify_cert(), it can first call:
|
|
|
|
X509_STORE_CTX_set_flags(ctx, X509_V_FLAG_ALLOW_PROXY_CERTS);
|
|
|
|
Where ctx is the pointer which then gets passed to X509_verify_cert().
|
|
|
|
- proxy certificate validation can be enabled before starting the application
|
|
by setting the environment variable OPENSSL_ALLOW_PROXY_CERTS.
|
|
|
|
In the future, it might be possible to enable proxy certificates by editing
|
|
openssl.cnf.
|
|
|
|
|
|
3. How to create proxy certificates
|
|
|
|
Creating proxy certificates is quite easy, by taking advantage of a lack of
|
|
checks in the 'openssl x509' application (*ahem*). You must first create a
|
|
configuration section that contains a definition of the proxyCertInfo extension,
|
|
for example:
|
|
|
|
[ v3_proxy ]
|
|
# A proxy certificate MUST NEVER be a CA certificate.
|
|
basicConstraints=CA:FALSE
|
|
|
|
# Usual authority key ID
|
|
authorityKeyIdentifier=keyid,issuer:always
|
|
|
|
# The extension which marks this certificate as a proxy
|
|
proxyCertInfo=critical,language:id-ppl-anyLanguage,pathlen:1,policy:text:AB
|
|
|
|
It's also possible to specify the proxy extension in a separate section:
|
|
|
|
proxyCertInfo=critical,@proxy_ext
|
|
|
|
[ proxy_ext ]
|
|
language=id-ppl-anyLanguage
|
|
pathlen=0
|
|
policy=text:BC
|
|
|
|
The policy value has a specific syntax, {syntag}:{string}, where the syntag
|
|
determines what will be done with the string. The following syntags are
|
|
recognised:
|
|
|
|
text indicates that the string is simply bytes, without any encoding:
|
|
|
|
policy=text:räksmörgås
|
|
|
|
Previous versions of this design had a specific tag for UTF-8 text.
|
|
However, since the bytes are copied as-is anyway, there is no need for
|
|
such a specific tag.
|
|
|
|
hex indicates the string is encoded in hex, with colons between each byte
|
|
(every second hex digit):
|
|
|
|
policy=hex:72:E4:6B:73:6D:F6:72:67:E5:73
|
|
|
|
Previous versions of this design had a tag to insert a complete DER
|
|
blob. However, the only legal use for this would be to surround the
|
|
bytes that would go with the hex: tag with whatever is needed to
|
|
construct a correct OCTET STRING. The DER tag therefore felt
|
|
superfluous, and was removed.
|
|
|
|
file indicates that the text of the policy should really be taken from a
|
|
file. The string is then really a file name. This is useful for
|
|
policies that are large (more than a few lines, e.g. XML documents).
|
|
|
|
The 'policy' setting can be split up in multiple lines like this:
|
|
|
|
0.policy=This is
|
|
1.policy= a multi-
|
|
2.policy=line policy.
|
|
|
|
NOTE: the proxy policy value is the part which determines the rights granted to
|
|
the process using the proxy certificate. The value is completely dependent on
|
|
the application reading and interpreting it!
|
|
|
|
Now that you have created an extension section for your proxy certificate, you
|
|
can easily create a proxy certificate by doing:
|
|
|
|
openssl req -new -config openssl.cnf -out proxy.req -keyout proxy.key
|
|
openssl x509 -req -CAcreateserial -in proxy.req -days 7 -out proxy.crt \
|
|
-CA user.crt -CAkey user.key -extfile openssl.cnf -extensions v3_proxy
|
|
|
|
You can also create a proxy certificate using another proxy certificate as
|
|
issuer (note: I'm using a different configuration section for it):
|
|
|
|
openssl req -new -config openssl.cnf -out proxy2.req -keyout proxy2.key
|
|
openssl x509 -req -CAcreateserial -in proxy2.req -days 7 -out proxy2.crt \
|
|
-CA proxy.crt -CAkey proxy.key -extfile openssl.cnf -extensions v3_proxy2
|
|
|
|
|
|
4. How to have your application interpret the policy?
|
|
|
|
The basic way to interpret proxy policies is to start with some default rights,
|
|
then compute the resulting rights by checking the proxy certificate against
|
|
the chain of proxy certificates, user certificate and CA certificates. You then
|
|
use the final computed rights. Sounds easy, huh? It almost is.
|
|
|
|
The slightly complicated part is figuring out how to pass data between your
|
|
application and the certificate validation procedure.
|
|
|
|
You need the following ingredients:
|
|
|
|
- a callback function that will be called for every certificate being
|
|
validated. The callback be called several times for each certificate,
|
|
so you must be careful to do the proxy policy interpretation at the right
|
|
time. You also need to fill in the defaults when the EE certificate is
|
|
checked.
|
|
|
|
- a data structure that is shared between your application code and the
|
|
callback.
|
|
|
|
- a wrapper function that sets it all up.
|
|
|
|
- an ex_data index function that creates an index into the generic ex_data
|
|
store that is attached to an X509 validation context.
|
|
|
|
Here is some skeleton code you can fill in:
|
|
|
|
#include <string.h>
|
|
#include <netdb.h>
|
|
#include <openssl/x509.h>
|
|
#include <openssl/x509v3.h>
|
|
|
|
#define total_rights 25
|
|
|
|
/*
|
|
* In this example, I will use a view of granted rights as a bit
|
|
* array, one bit for each possible right.
|
|
*/
|
|
typedef struct your_rights {
|
|
unsigned char rights[(total_rights + 7) / 8];
|
|
} YOUR_RIGHTS;
|
|
|
|
/*
|
|
* The following procedure will create an index for the ex_data
|
|
* store in the X509 validation context the first time it's called.
|
|
* Subsequent calls will return the same index. */
|
|
static int get_proxy_auth_ex_data_idx(X509_STORE_CTX *ctx)
|
|
{
|
|
static volatile int idx = -1;
|
|
if (idx < 0) {
|
|
X509_STORE_lock(X509_STORE_CTX_get0_store(ctx));
|
|
if (idx < 0) {
|
|
idx = X509_STORE_CTX_get_ex_new_index(0,
|
|
"for verify callback",
|
|
NULL,NULL,NULL);
|
|
}
|
|
X509_STORE_unlock(X509_STORE_CTX_get0_store(ctx));
|
|
}
|
|
return idx;
|
|
}
|
|
|
|
/* Callback to be given to the X509 validation procedure. */
|
|
static int verify_callback(int ok, X509_STORE_CTX *ctx)
|
|
{
|
|
if (ok == 1) {
|
|
/*
|
|
* It's REALLY important you keep the proxy policy
|
|
* check within this section. It's important to know
|
|
* that when ok is 1, the certificates are checked
|
|
* from top to bottom. You get the CA root first,
|
|
* followed by the possible chain of intermediate
|
|
* CAs, followed by the EE certificate, followed by
|
|
* the possible proxy certificates.
|
|
*/
|
|
X509 *xs = X509_STORE_CTX_get_current_cert(ctx);
|
|
|
|
if (X509_get_extension_flags(xs) & EXFLAG_PROXY) {
|
|
YOUR_RIGHTS *rights =
|
|
(YOUR_RIGHTS *)X509_STORE_CTX_get_ex_data(ctx,
|
|
get_proxy_auth_ex_data_idx(ctx));
|
|
PROXY_CERT_INFO_EXTENSION *pci =
|
|
X509_get_ext_d2i(xs, NID_proxyCertInfo, NULL, NULL);
|
|
|
|
switch (OBJ_obj2nid(pci->proxyPolicy->policyLanguage)) {
|
|
case NID_Independent:
|
|
/*
|
|
* Do whatever you need to grant explicit rights to
|
|
* this particular proxy certificate, usually by
|
|
* pulling them from some database. If there are none
|
|
* to be found, clear all rights (making this and any
|
|
* subsequent proxy certificate void of any rights).
|
|
*/
|
|
memset(rights->rights, 0, sizeof(rights->rights));
|
|
break;
|
|
case NID_id_ppl_inheritAll:
|
|
/*
|
|
* This is basically a NOP, we simply let the current
|
|
* rights stand as they are.
|
|
*/
|
|
break;
|
|
default:
|
|
/* This is usually the most complex section of code.
|
|
* You really do whatever you want as long as you
|
|
* follow RFC 3820. In the example we use here, the
|
|
* simplest thing to do is to build another, temporary
|
|
* bit array and fill it with the rights granted by
|
|
* the current proxy certificate, then use it as a
|
|
* mask on the accumulated rights bit array, and
|
|
* voilà, you now have a new accumulated rights bit
|
|
* array.
|
|
*/
|
|
{
|
|
int i;
|
|
YOUR_RIGHTS tmp_rights;
|
|
memset(tmp_rights.rights, 0, sizeof(tmp_rights.rights));
|
|
|
|
/*
|
|
* process_rights() is supposed to be a procedure
|
|
* that takes a string and it's length, interprets
|
|
* it and sets the bits in the YOUR_RIGHTS pointed
|
|
* at by the third argument.
|
|
*/
|
|
process_rights((char *) pci->proxyPolicy->policy->data,
|
|
pci->proxyPolicy->policy->length,
|
|
&tmp_rights);
|
|
|
|
for(i = 0; i < total_rights / 8; i++)
|
|
rights->rights[i] &= tmp_rights.rights[i];
|
|
}
|
|
break;
|
|
}
|
|
PROXY_CERT_INFO_EXTENSION_free(pci);
|
|
} else if (!(X509_get_extension_flags(xs) & EXFLAG_CA)) {
|
|
/* We have an EE certificate, let's use it to set default! */
|
|
YOUR_RIGHTS *rights =
|
|
(YOUR_RIGHTS *)X509_STORE_CTX_get_ex_data(ctx,
|
|
get_proxy_auth_ex_data_idx(ctx));
|
|
|
|
/* The following procedure finds out what rights the owner
|
|
* of the current certificate has, and sets them in the
|
|
* YOUR_RIGHTS structure pointed at by the second
|
|
* argument.
|
|
*/
|
|
set_default_rights(xs, rights);
|
|
}
|
|
}
|
|
return ok;
|
|
}
|
|
|
|
static int my_X509_verify_cert(X509_STORE_CTX *ctx,
|
|
YOUR_RIGHTS *needed_rights)
|
|
{
|
|
int ok;
|
|
int (*save_verify_cb)(int ok,X509_STORE_CTX *ctx) =
|
|
X509_STORE_CTX_get_verify_cb(ctx);
|
|
YOUR_RIGHTS rights;
|
|
|
|
X509_STORE_CTX_set_verify_cb(ctx, verify_callback);
|
|
X509_STORE_CTX_set_ex_data(ctx, get_proxy_auth_ex_data_idx(ctx), &rights);
|
|
X509_STORE_CTX_set_flags(ctx, X509_V_FLAG_ALLOW_PROXY_CERTS);
|
|
ok = X509_verify_cert(ctx);
|
|
|
|
if (ok == 1) {
|
|
ok = check_needed_rights(rights, needed_rights);
|
|
}
|
|
|
|
X509_STORE_CTX_set_verify_cb(ctx, save_verify_cb);
|
|
|
|
return ok;
|
|
}
|
|
|
|
|
|
If you use SSL or TLS, you can easily set up a callback to have the
|
|
certificates checked properly, using the code above:
|
|
|
|
SSL_CTX_set_cert_verify_callback(s_ctx, my_X509_verify_cert, &needed_rights);
|
|
|
|
|
|
--
|
|
Richard Levitte
|