Support client bootstrapping to receive OSCORE security information #950

rikard-sics · 2020-12-17T12:53:15Z

This commit enables the client to receive OSCORE security context information from the bootstrap server during the bootstrapping process. The bootstrap server web UI was updated to accept settings for OSCORE under the LWM2M Server tab. Some changes were done in the client side code to support this functionality. Finally, an OSCORE server identity is now generated and set.

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

This commit adds early support for using OSCORE to communicate between a Leshan client and server. The server can register and perform its request towards the server using OSCORE. Likewise the server can perform requests to the client using OSCORE. The web interface of the server has been extended to allow adding clients that use OSCORE for security including their OSCORE security context information. On the client side further work is needed to allow setting OSCORE security context information on it. Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

…eation Code to implement the points listed here: eclipse-leshan#718 (comment) The OSCORE object is generated and filled in the Leshan client demo. It will then be used to create a ServerInfo object using the ServersInfoExtractor. Next the ServerInfo is used in the CaliforniumEndpointsManager to generate endpoints towards servers that OSCORE should be used with. See also issue regarding OSCORE client integration: eclipse-leshan#726 Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

An identity string based on OSCORE parameters is now set for connected clients that use OSCORE. The identity string contains the OSCORE Sender and Recipient ID the client and server are using. These parameters can be retrieved from the source EndpointContext. The identity for OSCORE is also set as a SecurityInfo based on the OSCORE context configured on the server to use with a specific client. The identity of a connected client and the configured SecurityInfo are then matched in the SecurityChecker. Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

This commit adds support for the client communicating to the bootstrap server using OSCORE. The bootstrap server web UI has been extended to accept settings for OSCORE. Some modifications were also done on the client to support it bootstrapping using OSCORE. It supports OSCORE settings as command line parameters since before. Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

The possibility to input an ID Context for OSCORE on the client-demo, bsserver-demo and server demo has now been removed. Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

This commit enables the client to receive OSCORE security context information from the bootstrap server during the bootstrapping process. The bootstrap server web UI was updated to accept settings for OSCORE under the LWM2M Server tab. Some changes were done in the client side code to support this functionality. Finally, an OSCORE server identity is now generated and set. Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2020-12-17T12:54:56Z

This is a resubmission of the old PR #919.

Setting the server identity for OSCORE should solve the problem recently reported in #725 (comment).

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2020-12-18T11:11:51Z

I have taken into account the feedback from the last PR with regards to excessive trace printing.

Also I just added code from the suggestion regarding retrieval of the OSCORE object from the old PR:
#919 (comment)
I realized I had not added that yet.

sbernard31 · 2020-12-18T13:59:15Z

Also I just added code from the suggestion regarding retrieval of the OSCORE object from the old PR:
#919 (comment)
I realized I had not added that yet.

I written this quickly without testing just to give the idea, so it could contain bugs 😁

sbernard31

Testing this I faced this NPE which is not related to this PR :

java.lang.NullPointerException: null
	at org.eclipse.leshan.server.security.SecurityChecker.checkSecurityInfos(SecurityChecker.java:63)
	at org.eclipse.leshan.server.bootstrap.DefaultBootstrapSessionManager.begin(DefaultBootstrapSessionManager.java:71)
	at org.eclipse.leshan.server.bootstrap.DefaultBootstrapHandler.bootstrap(DefaultBootstrapHandler.java:88)
	at org.eclipse.leshan.server.californium.bootstrap.BootstrapResource.handlePOST(BootstrapResource.java:104)
	at org.eclipse.californium.core.CoapResource.handleRequest(CoapResource.java:227)
	at org.eclipse.leshan.core.californium.LwM2mCoapResource.handleRequest(LwM2mCoapResource.java:51)
	at org.eclipse.californium.core.server.ServerMessageDeliverer.deliverRequest(ServerMessageDeliverer.java:106)
	at org.eclipse.californium.core.network.stack.BaseCoapStack$StackTopAdapter.receiveRequest(BaseCoapStack.java:204)

When connecting to a bootstrap server using oscore with an endpoint name which should not use oscore.

sbernard31 · 2020-12-18T09:55:21Z

leshan-server-core/src/main/java/org/eclipse/leshan/server/bootstrap/BootstrapUtil.java

@@ -145,6 +179,10 @@ public static BootstrapWriteRequest toWriteRequest(int instanceId, ACLConfig acl
        for (Entry<Integer, ACLConfig> acl : bootstrapConfig.acls.entrySet()) {
            requests.add(toWriteRequest(acl.getKey(), acl.getValue(), contentFormat));
        }
+        // handle oscore //TODO: Avoid to write also bs oscore object back to client?


I don't get this comment ? if you add an oscore object in your Bootstrap Config that means that you want to write the OSCORE object or I missed something ?

My point here was that let's say a client communicates to the bootstrap server using OSCORE, and it gets OSCORE configuration to use towards the device manager. In such case as things are now 2 oscore objects will be written to the client. One for the bootstrap server and one for device manager. But the client already has the OSCORE security configuration to use towards the bootstrap server. So I was not sure how this should be handled (perhaps it's fine).

Your concern makes sense but this is maybe not the right place for this TODO, because this is more a bsserver-demo question. I will try to explain.

There are 2 main concepts about bootstrap server.

The SecurityStore which contains information about how a device should connect to server (using psk, rpk , x509, or oscore ).

And ConfigStore which contains configuration which should be written/deleted on device during bootstrap session
Since a recent refactoring(#883), there is 2 way to create this kind of config :

the old BootstrapConfig more easier to write but less flexible.

the new BootstrapConfiguration which is more complex but also more flexible.

BootstrapUtil is just an utility class which help to go from the old way to the new.

All of those ☝️ above is leshan concept and there is no problem here.

the client already has the OSCORE security configuration to use towards the bootstrap server.

This is up to the user to create a config where it delete then rewrite the oscore object or do not change anything.

But for demo we use a "special"(not so good?) SecurityStore which uses the content of ConfigStore to provide securityInfo. (see BootstrapConfigSecurityStore javadocs)

In that case, in you want to allow client to connect using oscore you need a bootstrap config with oscore object and so you will write an oscore object. This is imposible to accept oscore connection and to not write an oscore object. (This is a bsserver-demo limitation only)

This works for other objects (security 0 and server 1) because we delete it before to rewrite it.
(see bsconfigstore.js)

We should probably also delete OSCORE ? (but maybe only when we write oscore object to avoid to disturb client which does not support OSCORE object ?)

I see, thanks for the explanation. As for deleting OSCORE, would it be good to add that to this PR also?

or in another PR as you prefer :)

(if you do it in another PR move the TODO)

Thanks! I'll try to do it here already.

I have now added this functionality in commit 8034e0f

I found it a bit difficult to determine if the OSCORE object is being written, in a compact way. I didn't want to rely on checking hardcoded indexes in it. But perhaps there is a briefer way.

I think it's OK.

sbernard31 · 2020-12-18T11:02:29Z

leshan-client-core/src/main/java/org/eclipse/leshan/client/servers/ServersInfoExtractor.java

+                        ObjectLink oscoreObjLink = (ObjectLink) security.getResource(SEC_OSCORE_SECURITY_MODE)
+                                .getValue();
+                        if (oscoreObjLink != null) {
+                            int oscoreObjectInstanceId = oscoreObjLink.getObjectInstanceId();
+                            if (!oscoreObjLink.isNullLink() && oscoreObjLink.getObjectId() != OSCORE) {
+                                LOG.warn(
+                                        "The security object's 'OSCORE Security Mode' links to an incorrect object type.");
+                            } else {
+                                oscoreInstance = oscores.getInstance(oscoreObjectInstanceId);
+                                if (oscoreInstance == null) {
+                                    // maybe we should use same log level for this one and the other just above
+                                    LOG.error("Failed to retrieve OSCORE object linked from BS security object");
+                                }
+                            }


I think we should maybe consider oscore null link as not an error ?
And maybe have more help full log ?
Maybe something like this ?

LwM2mObjectInstance oscoreInstance = null; ObjectLink oscoreObjLink = (ObjectLink) security.getResource(SEC_OSCORE_SECURITY_MODE).getValue(); if (oscoreObjLink != null && !oscoreObjLink.isNullLink()) { if (oscoreObjLink.getObjectId() != OSCORE) { LOG.warn("Invalid Security info for bootstrap server : 'OSCORE Security Mode' does not link to OSCORE Object but to {} object.", oscoreObjLink.getObjectId()); } else { oscoreInstance = oscores.getInstance(oscoreObjLink.getObjectInstanceId()); if (oscoreInstance == null) { LOG.warn("Invalid Security info for bootstrap server : OSCORE instance {} does not exist.",oscoreObjLink.getObjectInstanceId() ); } } }

If it sounds good to you, this should also be adapted to Server by adapting the log (talking about LWM2M server instead of bootstrap server and given some information like server id maybe or url to know which server is not correctly configured)

About considering oscore null link as "normal", this is currently the default behavior of our current implementation of Security Object. See :
https://github.com/eclipse/leshan/blob/4488653d712bb917da4a6af1f691eb5ce0cd2f69/leshan-client-core/src/main/java/org/eclipse/leshan/client/object/Security.java#L252

We should also check is oscores is null

LwM2mObjectInstance oscoreInstance = null; ObjectLink oscoreObjLink = (ObjectLink) security.getResource(SEC_OSCORE_SECURITY_MODE).getValue(); if (oscoreObjLink != null && !oscoreObjLink.isNullLink()) { if (oscoreObjLink.getObjectId() != OSCORE) { LOG.warn("Invalid Security info for bootstrap server : 'OSCORE Security Mode' does not link to OSCORE Object but to {} object.", oscoreObjLink.getObjectId()); } else { if (oscores == null) { LOG.warn("Invalid Security info for bootstrap server : OSCORE object enabler is not available." ); } else { oscoreInstance = oscores.getInstance(oscoreObjLink.getObjectInstanceId()); if (oscoreInstance == null) { LOG.warn("Invalid Security info for bootstrap server : OSCORE instance {} does not exist.",oscoreObjLink.getObjectInstanceId() ); } } } }

(:warning: I didn't test that code and write log message quickly so maybe there is still some bug and warning message could be better ?)

Thank you for the suggestions regarding this. I will make a commit fairly soon to try and solve this point.

I have now made a commit in a811182 to update this. Next week I will be back full time and can take a look at this and the other open points in more detail.

sbernard31 · 2020-12-18T11:09:09Z

...emo/src/main/java/org/eclipse/leshan/server/bootstrap/demo/BootstrapConfigSecurityStore.java


            // First find the context for this endpoint
-            for (Map.Entry<Integer, BootstrapConfig.OscoreObject> oscoreEntry : bsConfig.oscore.entrySet()) {
-                OscoreObject value = oscoreEntry.getValue();
+            for (Map.Entry<Integer, BootstrapConfig.ServerSecurity> bsEntry : bsConfig.security.entrySet()) {


here it seems you don't use entry.getKey(), so maybe you should use bsConfig.security.values() instead ?

True, I believe it was like that since the code below does things in a similar way.

I have now updated to use .values() instead in commit 465f82c

sbernard31 · 2020-12-18T14:06:57Z

No linked to this PR too but I think this attribute need some javadoc :
https://github.com/eclipse/leshan/blob/4488653d712bb917da4a6af1f691eb5ce0cd2f69/leshan-server-core/src/main/java/org/eclipse/leshan/server/bootstrap/BootstrapConfig.java#L228

sbernard31 · 2020-12-18T15:15:46Z

(Just to let you know, I will be unavailable next 2 weeks)

rikard-sics · 2020-12-18T15:41:55Z

(Just to let you know, I will be unavailable next 2 weeks)

Yes, I understand. I will be much less available in the next 2-3 weeks also.

rikard-sics · 2020-12-18T15:42:13Z

I written this quickly without testing just to give the idea, so it could contain bugs

Sure :) I did try to test it for combinations of OSCORE communication & OSCORE bootstrapping and it seemed solid.

rikard-sics · 2020-12-18T15:45:36Z

Testing this I faced this NPE which is not related to this PR :

Hmm, thanks a lot for pointing this out. I will try to figure out what is happening there.

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2021-01-20T15:06:03Z

No linked to this PR too but I think this attribute need some javadoc :
https://github.com/eclipse/leshan/blob/4488653d712bb917da4a6af1f691eb5ce0cd2f69/leshan-server-core/src/main/java/org/eclipse/leshan/server/bootstrap/BootstrapConfig.java#L228

Javadoc for this has now been added in commit 3784bd8

rikard-sics · 2021-01-20T15:26:59Z

Testing this I faced this NPE which is not related to this PR :

java.lang.NullPointerException: null
	at org.eclipse.leshan.server.security.SecurityChecker.checkSecurityInfos(SecurityChecker.java:63)
	at org.eclipse.leshan.server.bootstrap.DefaultBootstrapSessionManager.begin(DefaultBootstrapSessionManager.java:71)
	at org.eclipse.leshan.server.bootstrap.DefaultBootstrapHandler.bootstrap(DefaultBootstrapHandler.java:88)
	at org.eclipse.leshan.server.californium.bootstrap.BootstrapResource.handlePOST(BootstrapResource.java:104)
	at org.eclipse.californium.core.CoapResource.handleRequest(CoapResource.java:227)
	at org.eclipse.leshan.core.californium.LwM2mCoapResource.handleRequest(LwM2mCoapResource.java:51)
	at org.eclipse.californium.core.server.ServerMessageDeliverer.deliverRequest(ServerMessageDeliverer.java:106)
	at org.eclipse.californium.core.network.stack.BaseCoapStack$StackTopAdapter.receiveRequest(BaseCoapStack.java:204)

When connecting to a bootstrap server using oscore with an endpoint name which should not use oscore.

This NPE seems to be fine now, I am not experiencing this.

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2021-01-21T10:59:20Z

I think that covers all the open points. Please tell me if you feel something more should be addressed.

sbernard31 · 2021-01-21T16:42:15Z

This NPE seems to be fine now, I am not experiencing this.

I still get the NPE, if I try to connect with a client with oscore setting but didn't add any bootstrapconfig at server side.

rikard-sics · 2021-01-21T17:12:51Z

I still get the NPE, if I try to connect with a client with oscore setting but didn't add any bootstrapconfig at server side.

Thanks! I will check it out further trying to replicate your setup. I was testing with some bootstrap config added.

rikard-sics · 2021-01-25T15:45:48Z

I still get the NPE, if I try to connect with a client with oscore setting but didn't add any bootstrapconfig at server side.

Hmm, maybe I am still testing with different settings but I cannot replicate this.

This is my output on the bootstrap server:

2021-01-25 16:39:25,173 INFO LeshanBootstrapServer - Bootstrap server started at coap://0.0.0.0/0.0.0.0:5683 coaps://0.0.0.0/0.0.0.0:5684
2021-01-25 16:39:25,461 INFO LeshanBootstrapServerDemo - Web server started at http://0.0.0.0:8080/.
2021-01-25 16:40:03,375 ERROR RequestDecryptor - Security context not found
2021-01-25 16:40:03,378 ERROR ObjectSecurityLayer - Error while receiving OSCore request: Security context not found

I tried having a bootstrap server with no configuration at all, and bootstrapping with OSCORE. I also tried having a bootstrap server with a matching endpoint name but which shouldn't use OSCORE, and bootstrapping with OSCORE.

rikard-sics · 2021-01-25T16:06:28Z

I still get the NPE, if I try to connect with a client with oscore setting but didn't add any bootstrapconfig at server side.

Actually testing a bit more I could replicate it. If I start the bootstrap server, add a client with OSCORE configuration and then delete that client again. The problem is that the OSCORE context will live on and be valid even after the client has been deleted in the bootstrap server web interface. Then for the following request from the client with OSCORE a context will be found and it will decrypt correctly.

sbernard31 · 2021-01-25T16:09:33Z

In my case, I think I was using an existing OSCORE context from another config but with an endpoint name without any bootstrap config.

rikard-sics · 2021-01-25T16:11:34Z

In my case, I think I was using an existing OSCORE context from another config but with an endpoint name without any bootstrap config.

I see, yes it seems I can replicate it. I will add a commit to fix this.

sbernard31 · 2021-01-25T16:12:21Z

The issue is in SecurityChecker.checkSecurityInfos. I guess "isOscore" if block must be more like "isSecure" one ?

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics · 2021-01-25T16:23:03Z

The issue is in SecurityChecker.checkSecurityInfos. I guess "isOscore" if block must be more like "isSecure" one ?

Yes, indeed. This should now have been fixed in commit 6f26f46

sbernard31 · 2021-01-26T14:46:45Z

I integrated this in oscore banch (squashed in 1 commit 0d45552)

I did some minor changes :

Rewite a little bit BootstrapConfigSecurityStore.getAllByEndpoint() (rename some variable & remove continue)
Remove unwanted space in bsconfigstore.js
In ServersInfoExtractor.getInfo() add serverURI in logs.
In BootstrapUtil remove //TODO: Avoid to write also bs oscore object back to client?

Maybe we should rebase again OSCORE branch on master, I will look if there is not too much conflict.

rikard-sics · 2021-01-26T15:42:13Z

I integrated this in oscore banch (squashed in 1 commit 0d45552)

I did some minor changes :

Rewite a little bit BootstrapConfigSecurityStore.getAllByEndpoint() (rename some variable & remove continue)

Remove unwanted space in bsconfigstore.js

In ServersInfoExtractor.getInfo() add serverURI in logs.

In BootstrapUtil remove //TODO: Avoid to write also bs oscore object back to client?

Great, thanks a lot!

Maybe we should rebase again OSCORE branch on master, I will look if there is not too much conflict.

Yes that may be good to do again. Hopefully there are not too many conflicts.

sbernard31 · 2021-01-26T16:02:20Z

I created a oscore2 branch and tested quickly and it seems OK.
But there was several conflict, I hope I didn't add any issue by resolving it.

Maybe you can test/look at it and if you think it's OK I made it the oscore branch ?

rikard-sics · 2021-01-26T17:13:44Z

I created a oscore2 branch and tested quickly and it seems OK.
But there was several conflict, I hope I didn't add any issue by resolving it.

Maybe you can test/look at it and if you think it's OK I made it the oscore branch ?

Thanks. Yes I can have a look and test things through.

rikard-sics · 2021-01-27T10:34:59Z

I created a oscore2 branch and tested quickly and it seems OK.
But there was several conflict, I hope I didn't add any issue by resolving it.

Maybe you can test/look at it and if you think it's OK I made it the oscore branch ?

I went over and tested a number of things including:

Registration with OSCORE
- Then getting a resource on the client from the server web interface
Bootstrapping with OSCORE security
Bootstrapping for OSCORE material
Bootstrapping with OSCORE security & for OSCORE material
Other various combinations with DTLS and CoAP

Everything I tested worked so this looks good to me.

sbernard31 · 2021-01-27T10:54:09Z

Ok so now oscore branch is now rebased on master.
Thx @rikard-sics for the contribution 🙏

rikard-sics · 2021-01-27T12:58:10Z

Ok so now oscore branch is now rebased on master.
Thx @rikard-sics for the contribution

Thanks for all the assistance!

rikard-sics added 9 commits November 2, 2020 11:37

eclipse-leshan#674: Initial support for OSCORE security object

815300b

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Workaround about missing Token of observe request for oscore.

c48d54d

Allow setting OSCORE security settings in client command line arguments

3e0c2db

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Remove ID Context inputs for OSCORE

4488653

The possibility to input an ID Context for OSCORE on the client-demo, bsserver-demo and server demo has now been removed. Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Updated code for finding associated oscore instance from security object

af75452

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 reviewed Dec 18, 2020

View reviewed changes

sbernard31 mentioned this pull request Jan 4, 2021

OSCORE support in Leshan #725

Open

rikard-sics added 2 commits January 7, 2021 18:10

Further updates to code for finding oscore instance from security object

a811182

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Added javadoc for oscoreSecurityMode

3784bd8

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

rikard-sics added 2 commits January 20, 2021 16:35

Modified OSCORE context retrieval for endpoint

465f82c

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

First delete OSCORE object on client when writing to it

8034e0f

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

Fix NPE when bootstrapping with OSCORE to server without matching config

6f26f46

Signed-off-by: Rikard Höglund <rikard.hoglund@ri.se>

sbernard31 force-pushed the oscore branch from 0d45552 to 8b77ac2 Compare January 27, 2021 10:52

sbernard31 closed this Jan 27, 2021

sbernard31 mentioned this pull request Jan 28, 2021

How to add Config when client request? #963

Closed

Support client bootstrapping to receive OSCORE security information #950

Support client bootstrapping to receive OSCORE security information #950

Conversation

rikard-sics commented Dec 17, 2020

rikard-sics commented Dec 17, 2020

rikard-sics commented Dec 18, 2020

sbernard31 commented Dec 18, 2020

sbernard31 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rikard-sics Jan 20, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

rikard-sics Jan 4, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

sbernard31 commented Dec 18, 2020

sbernard31 commented Dec 18, 2020

rikard-sics commented Dec 18, 2020

rikard-sics commented Dec 18, 2020

rikard-sics commented Dec 18, 2020

rikard-sics commented Jan 20, 2021

rikard-sics commented Jan 20, 2021

rikard-sics commented Jan 21, 2021

sbernard31 commented Jan 21, 2021

rikard-sics commented Jan 21, 2021 • edited Loading

rikard-sics commented Jan 25, 2021

rikard-sics commented Jan 25, 2021

sbernard31 commented Jan 25, 2021

rikard-sics commented Jan 25, 2021

sbernard31 commented Jan 25, 2021

rikard-sics commented Jan 25, 2021

sbernard31 commented Jan 26, 2021

rikard-sics commented Jan 26, 2021

sbernard31 commented Jan 26, 2021

rikard-sics commented Jan 26, 2021

rikard-sics commented Jan 27, 2021

sbernard31 commented Jan 27, 2021

rikard-sics commented Jan 27, 2021

rikard-sics Jan 20, 2021 •

edited

Loading

rikard-sics Jan 4, 2021 •

edited

Loading

rikard-sics commented Jan 21, 2021 •

edited

Loading