Block in Bundle            ID    
+===================================+====+
|         Primary Block             | B1 |
+-----------------------------------+----+                     
|             BIB                   | B2 |
|  OP(integrity, target=B1)         |    |
+-----------------------------------+----+                  
|             BCB                   | B3 |
|  OP(confidentiality, target=B4)   |    |
+-----------------------------------+----+
|      Extension Block              | B4 |
+-----------------------------------+----+
|             BIB                   | B5 |
|  OP(integrity, target=B6)         |    |
+-----------------------------------+----+
|      Extension Block              | B6 |
+-----------------------------------+----+
|             BCB                   | B7 |
|  OP(confidentiality,target=B8,B9) |    |
+-----------------------------------+----+                
|   BIB  (encrypted by B7)          | B8 |
|  OP(integrity, target=B9)         |    |
+-----------------------------------+----|                                                        
|         Payload Block             | B9 |
+-----------------------------------+----+
               

Figure 4: Sample Use of BSP Blocks

In this example a bundle has four non-security-related blocks: the primary block (B1), three extension blocks (B4,B6), and a payload block (B9). The following security applications are applied to this bundle.

• An integrity signature applied to the canonicalized primary block. This is accomplished by a single BIB (B2).
• Confidentiality for the first extension block (B4). This is accomplished by a BCB block (B3).
• Integrity for the second extension block (B6). This is accomplished by a BIB block (B5). NOTE: If the extension block B6 contains a representation of the serialized bundle (such as a hash over all blocks in the bundle at the time of its last transmission) then the BIB block is also providing an authentication service from the prior BPSEC-BPA to this BPSEC-BPA.
• An integrity signature on the payload (B10). This is accomplished by a BIB block (B8).
• Confidentiality for the payload block and it's integrity signature. This is accomplished by a BCB block, B7, encrypting B8 and B9.

4. Canonical Forms

By definition, an integrity service determines whether any aspect of a block was changed from the moment the security service was applied at the security source until the point of current evaluation. To successfully verify the integrity of a block, the data passed to the verifying cipher suite MUST be the same bits, in the same order, as those passed to the signature-generating cipher suite at the security source.

However, [BPBIS] does not specify a single on-the-wire encoding of bundles. In cases where a security source generates a different encoding than that used at a receiving node, care MUST be taken to ensure that the inputs to cipher suites at the receiving node is a bitwise match to inputs provided at the security source.

This section provides guidance on how to create a canonical form for each type of block in a bundle. This form MUST be used when generating inputs to cipher suites for use by BPSec blocks.

This specification does not define any security operation over the entire bundle and, therefore, provides no canonical form for a serialized bundle.

4.1. Technical Notes

The following technical considerations hold for all canonicalizations in this section.

• Any numeric fields defined as variable-length MUST be expanded to their "unpacked" form. For example, a 32-bit integer value MUST be unpacked to a four-byte representation.
• Each block encoding MUST follow the CBOR encodings provided in [BPBISCBOR].
• Canonical forms are not transmitted, they are used to generate input to a cipher suite for secuity processing at a security-aware node.
• Reserved flags MUST NOT be included in any canonicalization as it is not known if those flags will chaneg in transit.
• These canonicalization algorithms assume that endpoint IDs themselves are immutable and they are unsuitable for use in environments where that assumption might be violated.
• Cipher suites MAY define their own canonicalization algorithms and require the use of those algorithms over the ones provided in this specification. In the event of conflicting canonicalization algorithms, cipher suite algorithms take precedence over this specification.

4.2. Primary Block Canonicalization

The primary block canonical form is the same as the CBOR encoding of the block, with certain modifications to account for allowed block changes as the bundle traverses the DTN. The fields that compromise the primary block, and any special considerations for their representation in a canonical form, are as follows.

• The Version field is included, without modification.
• The Bundle Processing Flags field is used, with modification. Certain bundle processing flags MAY change as a bundle transits the DTN without indicating an integrity error. These flags, which are identified below, MUST NOT be represented in the canonicalized form of the bundle processing flags and, instead, be represented by the bit 0.
  • Reserved flags.
  • Bundle is a Fragment flag.

• The CRC Type, Destination EID, Source Node ID, Report-To EID, Creation Timestamp, and Lifetime fields are included, without modification.
• The fragment ID field MAY change if the bundle is fragmented in transit and, as such, this field MUST NOT be included in the canonicalization.
• The CRC field MAY change at each hop - for example, if a bundle becomes fragmented, each fragment will have a different CRC value from the original signed primary block. As such, this field MUST NOT be included in the canonicalization.

4.3. Non-Primary-Block Canonicalization

All non-primary blocks (NPBs) in [BPBIS] share the same block structure and should be canonicalized in the same way.

Canonicalization for NPBs is dependent on whether the security operation being performed is integrity or confidentiality. Integrity operations consider every field in the block, whereas confidentiality operations only consider the block-type-specific data. Since confidentiality is applied to hide information (replacing plaintext with ciphertext) it provides no benefit to include in the confidentiality calculation information that MUST remain readable, such as block fields other than the block-type-specific data.

The fields that comprise a NPB, and any special considerations for their representation in a canonical form, are as follows.

• The Block Type Code field is included, without modification, for integrity operations and omitted for confidentiality operations.
• The Block Number field is included, without modification, for integrity operations and omitted for confidentiality operations.
• The Block Processing Control Flags field is included, without modification, for integrity operations and omitted for confidentiality operations, with the exception of reserved flags which are treated as 0 in both cases.
• The CRC type and CRC fields are included, without modification, for integrity operations and omitted for confidentiality operations.
• The Block Type Specific Data field is included, without modification, for both integrity and confidentiality operations, with the exception that in some cases only a portion of the payload data is to be processed. In such a case, only those bytes are included in the canonical form and additional cipher suite parameters are required to specify which part of the field is included.

5. Security Processing

This section describes the security aspects of bundle processing.

5.1. Bundles Received from Other Nodes

Security blocks MUST be processed in a specific order when received by a security-aware node. The processing order is as follows.

• All BCB blocks in the bundle MUST be evaluated prior to evaluating any BIBs in the bundle. When BIBs and BCBs share a security target, BCBs MUST be evaluated first and BIBs second.

5.1.1. Receiving BCB Blocks

If a received bundle contains a BCB, the receiving node MUST determine whether it has the responsibility of decrypting the BCB security target and removing the BCB prior to delivering data to an application at the node or forwarding the bundle.

If the receiving node is the destination of the bundle, the node MUST decrypt any BCBs remaining in the bundle. If the receiving node is not the destination of the bundle, the node MAY decrypt the BCB if directed to do so as a matter of security policy.

If the relevant parts of an encrypted payload block cannot be decrypted (i.e., the decryption key cannot be deduced or decryption fails), then the bundle MUST be discarded and processed no further. If an encrypted security target other than the payload block cannot be decrypted then the associated security target and all security blocks associated with that target MUST be discarded and processed no further. In both cases, requested status reports (see [BPBIS]) MAY be generated to reflect bundle or block deletion.

When a BCB is decrypted, the recovered plain-text MUST replace the cipher-text in the security target body data

If a BCB contains multiple security targets, all security targets MUST be processed if the BCB is processed by the Node. The effect of this is to be the same as if each security target had been represented by an individual BCB with a single security target.

5.1.2. Receiving BIB Blocks

If a received bundle contains a BIB, the receiving node MUST determine whether it has the responsibility of verifying the BIB security target and whether to remove the BIB prior to delivering data to an application at the node or forwarding the bundle.

A BIB MUST NOT be processed if the security target of the BIB is also the security target of a BCB in the bundle. Given the order of operations mandated by this specification, when both a BIB and a BCB share a security target, it means that the security target MUST have been encrypted after it was integrity signed and, therefore, the BIB cannot be verified until the security target has been decrypted by processing the BCB.

If the security policy of a security-aware node specifies that a bundle should have applied integrity to a specific security target and no such BIB is present in the bundle, then the node MUST process this security target in accordance with the security policy. This MAY involve removing the security target from the bundle. If the removed security target is the payload or primary block, the bundle MAY be discarded. This action may occur at any node that has the ability to verify an integrity signature, not just the bundle destination.

If the bundle has a BIB and the receiving node is the destination for the bundle, the node MUST verify the security target in accordance with the cipher suite specification. If a BIB check fails, the security target has failed to authenticate and the security target SHALL be processed according to the security policy. A bundle status report indicating the failure MAY be generated. Otherwise, if the BIB verifies, the security target is ready to be processed for delivery.

If the bundle has a BIB and the receiving node is not the bundle destination, the receiving node MAY attempt to verify the value in the security result field. If the check fails, the node SHALL process the security target in accordance to local security policy. It is RECOMMENDED that if a payload integrity check fails at a waypoint that it is processed in the same way as if the check fails at the destination.

If a BIB contains multiple security targets, all security targets MUST be processed if the BIB is processed by the Node. The effect of this is to be the same as if each security target had been represented by an individual BIB with a single security target.

5.2. Bundle Fragmentation and Reassembly

If it is necessary for a node to fragment a bundle and security services have been applied to that bundle, the fragmentation rules described in [BPBIS] MUST be followed. As defined there and repeated here for completeness, only the payload may be fragmented; security blocks, like all extension blocks, can never be fragmented.

Due to the complexity of bundle fragmentation, including the possibility of fragmenting bundle fragments, integrity and confidentiality operations are not to be applied to a bundle representing a fragment (i.e., a bundle whose "bundle is a Fragment" flag is set in the Bundle Processing Control Flags field). Specifically, a BCB or BIB MUST NOT be added to a bundle fragment, even if the security target of the security block is not the payload. When integrity and confidentiality must be applied to a fragment, we RECOMMEND that encapsulation be used instead.

6. Key Management

Key management in delay-tolerant networks is recognized as a difficult topic and is one that this specification does not attempt to solve.

7. Policy Considerations

When implementing BPSec, several policy decisions must be considered. This section describes key policies that affect the generation, forwarding, and receipt of bundles that are secured using this specification.

• If a bundle is received that contains more than one security operation, in violation of BPSec, then the BPA must determine how to handle this bundle. The bundle may be discarded, the block affected by the security operation may be discarded, or one security operation may be favored over another.
• BPAs in the network MUST understand what security operations they should apply to bundles. This decision may be based on the source of the bundle, the destination of the bundle, or some other information related to the bundle.
• If an intermediate receiver has been configured to add a security operation to a bundle, and the received bundle already has the security operation applied, then the receiver MUST understand what to do. The receiver may discard the bundle, discard the security target and associated BPSec blocks, replace the security operation, or some other action.
• It is recommended that security operations only be applied to the payload block, the primary block, and any block-types specifically identified in the security policy. If a BPA were to apply security operations such as integrity or confidentiality to every block in the bundle, regardless of the block type, there could be downstream errors processing blocks whose contents must be inspected at every hop in the network path.
• Adding a BIB to a security target that has already been encrypted by a BCB is not allowed. Therefore, we recommend three methods to add an integrity signature to an encrypted security target.
  1. At the time of encryption, an integrity signature may be generated and added to the BCB for the security target as additional information in the security result field.
  2. The encrypted block may be replicated as a new block and integrity signed.
  3. An encapsulation scheme may be applied to encapsulate the security target (or the entire bundle) such that the encapsulating structure is, itself, no longer the security target of a BCB and may therefore be the security target of a BIB.

8. Security Considerations

Given the nature of delay-tolerant networking applications, it is expected that bundles may traverse a variety of environments and devices which each pose unique security risks and requirements on the implementation of security within BPSEC. For these reasons, it is important to introduce key threat models and describe the roles and responsibilities of the BPSEC protocol in protecting the confidentiality and integrity of the data against those threats throughout the DTN. This section provides additional discussion on security threats that BPSEC will face and describe in additional detail how BPSEC security mechanisms operate to mitigate these threats.

It should be noted that BPSEC addresses only the security of data traveling over the DTN, not the underlying DTN itself. Additionally, BPSEC addresses neither the fitness of externally-defined cryptographic methods nor the security of their implementation. It is the responsibility of the BPSEC implementer that appropriate algorithms and methods are chosen. Furthermore, the BPSEC protocol does not address threats which share computing resources with the DTN and/or BPSEC software implementations. These threats may be malicious software or compromised libraries which intend to intercept data or recover cryptographic material. Here, it is the responsibility of the BPSEC implementer to ensure that any cryptographic material, including shared secret or private keys, is protected against access within both memory and storage devices.

The threat model described here is assumed to have a set of capabilities identical to those described by the Internet Threat Model in [RFC3552], but the BPSEC threat model is scoped to illustrate threats specific to BPSEC operating within DTN environments and therefore focuses on man-in-the-middle (MITM) attackers.

8.1. Attacker Capabilities and Objectives

BPSEC was designed to protect against MITM threats which may have access to a bundle during transit from its source, Alice, to its destination, Bob. A MITM node, Mallory, is a non-cooperative node operating on the DTN between Alice and Bob that has the ability to receive bundles, examine bundles, modify bundles, forward bundles, and generate bundles at will in order to compromise the confidentiality or integrity of data within the DTN. For the purposes of this section, any MITM node is assumed to effectively be security-aware even if it does not implement the BPSec protocol. There are three classes of MITM nodes which are differentiated based on their access to cryptographic material:

• Unprivileged Node: Mallory has not been provisioned within the secure environment and only has access to cryptographic material which has been publicly-shared.
• Legitimate Node: Mallory is within the secure environment and therefore has access to cryptographic material which has been provisioned to Mallory (i.e., K_M) as well as material which has been publicly-shared.
• Privileged Node: Mallory is a privileged node within the secure environment and therefore has access to cryptographic material which has been provisioned to Mallory, Alice and/or Bob (i.e. K_M, K_A, and/or K_B) as well as material which has been publicly-shared.

If Mallory is operating as a privileged node, this is tantamount to compromise; BPSec does not provide mechanisms to detect or remove Mallory from the DTN or BPSec secure environment. It is up to the BPSec implementer or the underlying cryptographic mechanisms to provide appropriate capabilities if they are needed. It should also be noted that if the implementation of BPSec uses a single set of shared cryptographic material for all nodes, a legitimate node is equivalent to a privileged node because K_M == K_A == K_B.

A special case of the legitimate node is when Mallory is either Alice or Bob (i.e., K_M == K_A or K_M == K_B). In this case, Mallory is able to impersonate traffic as either Alice or Bob, which means that traffic to and from that node can be decrypted and encrypted, respectively. Additionally, messages may be signed as originating from one of the endpoints.

8.2. Attacker Behaviors and BPSec Mitigations

8.2.1. Eavesdropping Attacks

Once Mallory has received a bundle, she is able to examine the contents of that bundle and attempt to recover any protected data or cryptographic keying material from the blocks contained within. The protection mechanism that BPSec provides against this action is the BCB, which encrypts the contents of its security target, providing confidentiality of the data. Of course, it should be assumed that Mallory is able to attempt offline recovery of encrypted data, so the cryptographic mechanisms selected to protect the data should provide a suitable level of protection.

When evaluating the risk of eavesdropping attacks, it is important to consider the lifetime of bundles on a DTN. Depending on the network, bundles may persist for days or even years. If a bundle does persist on the network for years and the cipher suite used for a BCB provides inadequate protection, Mallory may be able to recover the protected data before that bundle reaches its intended destination.

8.2.2. Modification Attacks

As a node participating in the DTN between Alice and Bob, Mallory will also be able to modify the received bundle, including non-BPSec data such as the primary block, payload blocks, or block processing control flags as defined in [BPBIS]. Mallory will be able to undertake activities which include modification of data within the blocks, replacement of blocks, addition of blocks, or removal of blocks. Within BPSec, both the BIB and BCB provide integrity protection mechanisms to detect or prevent data manipulation attempts by Mallory.

The BIB provides that protection to another block which is its security target. The cryptographic mechansims used to generate the BIB should be strong against collision attacks and Mallory should not have access to the cryptographic material used by the originating node to generate the BIB (e.g., K_A). If both of these conditions are true, Mallory will be unable to modify the security target or the BIB and lead Bob to validate the security target as originating from Alice.

Since BPSec security operations are implemented by placing blocks in a bundle, there is no in-band mechanism for detecting or correcting certain cases where Mallory removes blocks from a bundle. If Mallory removes a BCB block, but keeps the security target, the security target remains encrypted and there is a possibility that there may no longer be sufficient information to decrypt the block at its destination. If Mallory removes both a BCB (or BIB) and its security target there is no evidence left in the bundle of the security operation. Similarly, if Mallory removes the BIB but not the security target there is no evidence left in the bundle of the security operation. In each of these cases, the implementation of BPSec MUST be combined with policy configuration at endpoints in the network which describe the expected and required security operations that must be applied on transmission and are expected to be present on receipt. This or other similar out-of-band information is required to correct for removal of security information in the bundle.

A limitation of the BIB may exist within the implementation of BIB validation at the destination node. If Mallory is a legitimate node within the DTN, the BIB generated by Alice with K_A can be replaced with a new BIB generated with K_M and forwarded to Bob. If Bob is only validating that the BIB was generated by a legitimate user, Bob will acknowledge the message as originating from Mallory instead of Alice. In order to provide verifiable integrity checks, both a BIB and BCB should be used. Alice creates a BIB with the protected data block as the security target and then creates a BCB with both the BIB and protected data block as its security targets. In this configuration, since Mallory is only a legitimate node and does not have access to Alice's key K_A, Mallory is unable to decrypt the BCB and replace the BIB.

8.2.3. Topology Attacks

If Mallory is in a MITM position within the DTN, she is able to influence how any bundles that come to her may pass through the network. Upon receiving and processing a bundle that must be routed elsewhere in the network, Mallory has three options as to how to proceed: not forward the bundle, forward the bundle as intended, or forward the bundle to one or more specific nodes within the network.

Attacks that involve re-routing the packets throughout the network are essentially a special case of the modification attacks described in this section where the attacker is modifying fields within the primary block of the bundle. Given that BPSec cannot encrypt the contents of the primary block, alternate methods must be used to prevent this situation. These methods MAY include requiring BIBs for primary blocks, using encapsulation, or otherwise strategically manipulating primary block data. The specifics of any such mitigation technique are specific to the implementation of the deploying network and outside of the scope of this document.

Furthermore, routing rules and policies may be useful in enforcing particular traffic flows to prevent topology attacks. While these rules and policies may utilize some features provided by BPSec, their definition is beyond the scope of this specification.

8.2.4. Message Injection

Mallory is also able to generate new bundles and transmit them into the DTN at will. These bundles may either be copies or slight modifications of previously-observed bundles (i.e., a replay attack) or entirely new bundles generated based on the Bundle Protocol, BPSec, or other bundle-related protocols. With these attacks Mallory's objectives may vary, but may be targeting either the bundle protocol or application-layer protocols conveyed by the bundle protocol.

BPSec relies on cipher suite capabilities to prevent replay or forged message attacks. A BCB used with appropriate cryptographic mechanisms (e.g., a counter-based cipher mode) may provide replay protection under certain circumstances. Alternatively, application data itself may be augmented to include mechanisms to assert data uniqueness and then protected with a BIB, a BCB, or both along with other block data. In such a case, the receiving node would be able to validate the uniqueness of the data.

9. Ciphersuite Authorship Considerations

Cipher suite developers or implementers should consider the diverse performance and conditions of networks on which the Bundle Protocol (and therefore BPSec) will operate. Specifically, the delay and capacity of delay-tolerant networks can vary substantially. Cipher suite developers should consider these conditions to better describe the conditions when those suites will operate or exhibit vulnerability, and selection of these suites for implementation should be made with consideration to the reality. There are key differences that may limit the opportunity to leverage existing cipher suites and technologies that have been developed for use in traditional, more reliable networks:

• Data Lifetime: Depending on the application environment, bundles may persist on the network for extended periods of time, perhaps even years. Cryptographic algorithms should be selected to ensure protection of data against attacks for a length of time reasonable for the application.
• One-Way Traffic: Depending on the application environment, it is possible that only a one-way connection may exist between two endpoints, or if a two-way connection does exist, the round-trip time may be extremely large. This may limit the utility of session key generation mechanisms, such as Diffie-Hellman, as a two-way handshake may not be feasible or reliable.
• Opportunistic Access: Depending on the application environment, a given endpoint may not be guaranteed to be accessible within a certain amount of time. This may make asymmetric cryptographic architectures which rely on a key distribution center or other trust center impractical under certain conditions.

10. Defining Other Security Blocks

Other security blocks (OSBs) may be defined and used in addition to the security blocks identified in this specification. Both the usage of BIB, BCB, and any future OSBs MAY co-exist within a bundle and MAY be considered in conformance with BPSec if each of the following requirements are met by any future identified security blocks.

• Other security blocks (OSBs) MUST NOT reuse any enumerations identified in this specification, to include the block type codes for BIB and BCB.
• An OSB definition MUST state whether it can be the target of a BIB or a BCB. The definition MUST also state whether the OSB can target a BIB or a BCB.
• An OSB definition MUST provide a deterinistic processing order in the event that a bundle is received containing BIBs, BCBs, and OSBs. This processing order MUST NOT alter the BIB and BCB processing orders identified in this specification.
• An OSB definition MUST provide a canonicalization algorithm if the default non-primary-block canonicalization algorithm cannot be used to generate a deterministic input for a cipher suite. This requirement MAY be waived if the OSB is defined so as to never be the security target of a BIB or a BCB.
• An OSB definition MAY NOT require any behavior of a BPSEC-BPA that is in conflict with the behavior identified in this specification. In particular, the security processing requirements imposed by this specification MUST be consistent across all BPSEC-BPAs in a network.
• The behavior of an OSB when dealing with fragmentation MUST be specified and MUST NOT lead to ambiguous processing states. In particular, an OSB definition should address how to receive and process an OSB in a bundle fragment that may or may not also contain its security target. An OSB definition should also address whether an OSB may be added to a bundle marked as a fragment.

Additionally, policy considerations for the management, monitoring, and configuration associated with blocks SHOULD be included in any OSB definition.

NOTE: The burden of showing compliance with processing rules is placed upon the standards defining new security blocks and the identification of such blocks shall not, alone, require maintenance of this specification.

11. Conformance

All implementations are strongly RECOMMENDED to provide some method of hop-by-hop verification by generating a hash to some canonical form of the bundle and placing an integrity signature on that form using a BIB.

12. IANA Considerations

This protocol has fields that have been registered by IANA.

12.1. Bundle Block Types

This specification allocates three block types from the existing "Bundle Block Types" registry defined in [RFC6255] .