Patent Application
20080307110
The present disclosure relates generally to computer networking.
Virtual Private Networks (VPNs) provide a partitioning mechanism for isolating data transmitted and received between customer network nodes even though a corresponding physical network supporting propagation of the data is shared by many users. The data transmitted between such network nodes may be encrypted to protect against eavesdropping and tampering by unauthorized parties. A typical VPN may include a group, composed of network nodes in several subnets.
In one conventional system, when network nodes within the same group, but in different subnets wish to communicate, the network nodes (or their respective subnet routers) establish a point-to-point secure connection by negotiating a pairwise key for the communication. However, because establishing a pairwise key can require substantial computational resources, a Dynamic Group VPN (DGVPN) may be used instead. With a DGVPN, when network nodes which are members of the same group wish to communicate, no pairwise key is required. Instead, the two network nodes use a shared group key in order to communicate encrypted data. In this approach, a shared key server provides group security policies to nodes within particular groups. When a first network node wishes to communicate with a second node, both nodes being in a shared group, each node should have the group security policy from the shared key server. The security policy includes a group ID, a set of subnet prefixes identifying members of the group, and a group key. When the first node communicates with the second node, the first node determines if the first and the second node are in a shared group, by determining if there is a group ID in its routing table corresponding to the subnet prefix of the second node. If the two nodes are in a common group, the first node encrypts communications to the second node using the group key associated with the group ID. The two nodes may then communicate using that group key. A technique similar to this latter approach is described in co-pending U.S. patent application Ser. No. 10/867,266 (Wainner, et al.), filed Jun. 14, 2004, entitled “System and method for Dynamic Secured Group Communication.”
Objects, features, and advantages of particular embodiments of the invention will be apparent from the following description, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of various embodiments of the invention.
Unfortunately, the conventional DGVPN approach has some deficiencies in that there may be an asymmetry in the encryption states of the communicating network nodes. For example, if a first node has obtained the encryption state from the key server, but a second node is unable to do so, then the first node may send encrypted packets to the second node, which the second node is not able to decrypt, thereby wasting network resources. Also, in the situation that the second node is not able to obtain the encryption state, the second node may send unencrypted packets to the first node, thereby violating the security policy. Finally, even in a situation in which both nodes have obtained the encryption state, it is possible that the first node may not have received a routing update from the second node in another subnet. In that situation, no communication would be established.
An improved technique solves these deficiencies by using conditional advertisement of routing prefixes and conditional acceptance of routing prefixes in order to synchronize the encryption and routing states of two network hosts wishing to communicate. In one embodiment, routing information is only advertised if a security policy is in place within a network node. In addition, in another embodiment, received routing updates are only written into the node's routing table if the received routing update includes a reference to the proper security policy. In another embodiment, communication between the nodes only occurs once the routing information and security policy information has been properly distributed. In another embodiment, the routing advertisement includes a hash. Embodiments are also provided for apparatus and computer program products for practicing the invention.
An improved technique involves conditional route advertising. The route advertising is conditional on establishing a symmetric security policy within group members of a DGVPN. The group members wishing to communicate synchronize their routing states with their crypto-states. If the crypto-states do not synchronize, then the routing states do not synchronize either, and no communication is possible between the group members over a secure DGVPN. This may be referred to as a fail-close state. As a result, no communication is attempted unless both group members have a security policy in place.
In a DGVPN, groups are set up to allow for encrypted communications between subnets of the VPN. Typically, communications within each subnet is trusted, so the DGVPN is implemented at the routers controlling each subnet. These edge routers are referred to as network nodes. When a first computer in a first subnet wishes to communicate with a second computer in a second subnet, the communication will travel through the respective routers of each subnet. When the router of the first subnet (hereinafter, the first network node) receives the communication, it will determine the subnet prefix of the router of the second subnet (hereinafter, the second network node). If the second network node is part of a DGVPN group in common with the first network node (as determined by reference to a routing table), then secured communication between the first and second network nodes is possible by using a shared group key, provided by a key server. If the two network nodes are not within a common DGVPN group, then secure communication is not possible between the two nodes by means of a shared group key, in which case either unsecured communication must occur, or some other method of secured communication must be utilized.
In order to set up a DGVPN group, a network node, known as a group initiator, identifies a plurality of additional network nodes as members of the group. This involves sending a request to a key server with a list of the subnet prefixes of each of those network nodes. The key server responds by assigning a group ID and a group key to the group identified by the provided subnet prefixes. The key server also provides an authentication mechanism. The group initiator updates its routing table to indicate that the subnet prefixes indicate the group having the group ID. The group initiator also propagates the group ID and the list of subnet prefixes to all the network nodes within the group. Each network node within the group may then contact the key server with the group ID, be authenticated, and receive the group key. This group key is used for secured communications between the network nodes of the DGVPN group.
Thus, when the first computer wishes to communicate with the second computer, the network packet sent to the second computer is received at the first network node. The first network node looks up the subnet prefix of the second computer in its routing table (by examining the destination field of the packet) and determines if the subnet prefix is associated with a group ID in its routing table. If it is, the first network node looks up the group key corresponding to that group ID and encrypts the contents of the packet with that key. The packet is then forwarded on to the second network node according to standard routing protocols. Upon receiving the encrypted packet, the second network node looks up the subnet prefix of the source field of the packet in its routing table to determine if it is from a DGVPN group member, and if so, what the group ID is. Once the group ID is identified, the second network node may decrypt the contents of the packet by using the group key for the group.
The DGVPN as described above, may suffer from a deficiency if one of the network nodes has not yet received the group information (e.g., the group ID, the subnet prefixes, or the group key). Once a network node has received this information, updated its routing table to reflect the correspondence between subnet prefixes and the group ID, and established a correspondence between the group ID and the group key, the network node is said to have a group security policy in place.
The illustrated system is desirable to use in order to establish communications between network nodes 32, 34 both residing within the same DGVPN group 38 of a VPN 40. The arrangement, as hereinbelow described, allows for secure communication to occur between members of the group 38, while preventing insecure communication. During operation, each network node 32, 34 performs a procedure 200 in order to communicate.
The group security policy is a collection of information, including keys, ipsec proxies, lifetimes, algorithms, etc. The group security policy also contains a group reference, which identifies a specific policy. The group reference (or group ID) can be an abstract number or string. An example group security policy might look like the following:
Once a group security policy is in place, the method 200 proceeds to step 230. In step 230, the first network node 32 advertises routing information 50 across the network 36 according to the BGP protocol. The BGP protocol is described in “RFC 1771—A Border Gateway Protocol 4 (BGP 4)” by Y. Rekhter, et al. (1995). The routing information 50 includes a routing prefix 54, a group reference 56 (indicating that the first network node 32 has obtained the group security policy), and a hash 58, cryptographically combining the routing prefix 54 and the group reference 56. This is done, for example, by concatenating the routing prefix 54 with the group ID 56, and hashing the result using a pre-defined cryptographic hashing algorithm as is well-understood in the art. In step 240, the first network node 32 receives a corresponding routing update 52 from the second network node 34, with which communication is desired over VPN 40. If the routing update 52 from the second network node 34 contains a routing prefix 60, a group reference 62 (indicating that the second network node 34 has obtained the group security policy) corresponding to the group security policy stored in location 110 of memory 100 (which will only be in place if the first network node 32 has the group security policy in place), and a cryptographic hash 64 which properly combines the second routing prefix 60 and the group reference 62 (step 250), then the first network node 32 accepts the routing information 52 and inserts it into an entry 108 within the routing table 106, within memory 100 (step 260). Otherwise, either the second network node 34 has not provided a proper routing prefix 60 or the second network node has not properly confirmed that it possesses the proper group security policy for group communication over the VPN 40. In that case, communication between the network nodes 32, 34 is not established over the VPN 40 (step 270). The method may then loop back to step 240 until proper routing information is received. In step 280, communication commences between network nodes 32, 34 using the security policy over VPN 40.
The second network node 34 also performs the method 200. It should be understood, that either network node 32, 34 may initiate communications. If the first node 32 initiates communications, it will first make sure that the security policy is in place locally, and if not, it will query the key server 48. Only then will it send routing information across the network so that the second node 34 will know how to contact it. Upon receiving the routing information from the first node 32, the second node 34 checks the packet 50 to make sure that it includes a group reference 56 and a routing prefix 54. It also makes sure that these are properly hashed together 58, so as to avoid spoofing. If the second node 34 does not yet have a security policy in place, it should query the key server 48 at this time to obtain the security policy associated with the group reference 56. The second node 34 then inserts the routing information into its routing table 106 and sends back its own routing information in a similar packet 52.
Thus, the system 30 and method 200 are effective to prevent communications from occurring between the network nodes 32, 34 except while encrypted. They are also effective to ensure that communications only occurs once each node 32, 34 knows how to reach the other node across the network.
While various embodiments of the invention have been particularly shown and described, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
For example, embodiments have been described as including a hash 58, 64 in a routing update 50, 52. However, if the routing update is encrypted according to the group security policy 46, the hash 58, 64 may be omitted.
As an additional example, embodiments have been described as operating over a VPN on an IP network. However, the invention may also be applied over other forms of virtual networks operating over other forms of physical networks or collections thereof. Similarly, embodiments have been described as operating according to the BGP protocol, but the invention may also be applied over a different routing protocol.
In addition, it should be understood that a network node may also be described as a network host, both terms referring to network devices capable of communicating with each other over a VPN over an IP network.
