| Internet Engineering Task Force | Gwerder |
| Internet-Draft | FHNW |
| Intended status: Experimental | March 12, 2020 |
| Expires: September 13, 2020 |
MessageVortex Protocol
draft-gwerder-messagevortexmain-04
The MessageVortex (referred to as Vortex) protocol achieves different degrees of anonymity, including sender, receiver, and third-party anonymity, by specifying messages embedded within existing transfer protocols, such as SMTP or XMPP, sent via peer nodes to one or more recipients.
The protocol outperforms others by decoupling the transport from the final transmitter and receiver. No trust is placed into any infrastructure except for that of the sending and receiving parties of the message. The creator of the routing block has full control over the message flow. Routing nodes gain no non-obvious knowledge about the messages even when collaborating. While third-party anonymity is always achieved, the protocol also allows for either sender or receiver anonymity.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."
This Internet-Draft will expire on September 13, 2020.
Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Anonymisation is hard to achieve. Most previous attempts relied on either trust in a dedicated infrastructure or a specialized networking protocol.
Instead of defining a transport layer, Vortex piggybacks on other transport protocols. A blending layer embeds Vortex messages (VortexMessage) into ordinary messages of the respective transport protocol. This layer picks up the messages, passes them to a routing layer, which applies local operations to the messages, and resends the new message chunks to the next recipients.
A processing node learns as little as possible from the message or the network utilized. The operations have been designed to be sensible in any context. The 'onionized' structure of the protocol makes it impossible to follow the trace of a message without having control over the processing node.
MessageVortex is a protocol which allows sending and receiving messages by using a routing block instead of a destination address. With this approach, the sender has full control over all parameters of the message flow.
A message is split and reassembled during transmission. Chunks of the message may carry redundant information to avoid service interruptions during transit. Decoy and message traffic are not differentiable as the nature of the addRedundancy operation allows each generated portion to be either message or decoy. Therefore, any routing node is unable to distinguish between message and decoy traffic.
After processing, a potential receiver node knows if the message is destined for it (by creating a chunk with ID 0) or other nodes. Due to missing keys, no other node may perform this processing.
This RFC begins with general terminology (see Section 2) followed by an overview of the process (see Section 3). The subsequent sections describe the details of the protocol.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in [RFC2119].
Appendix A specifies all relevant parts of the protocol in ASN.1 (see [CCITT.X680.2002] and [CCITT.X208.1988]). The blocks are DER encoded, if not otherwise specified.
All numbers within this document are, if not suffixed, decimal numbers. Numbers suffixed with a small letter 'h' followed by two hexadecimal digits are octets written in hexadecimal. For example, a blank ASCII character (' ') is written as 20h and a capital 'K' in ASCII as 4Bh.
The following entities used in this document are defined below.
The term 'node' describes any computer system connected to other nodes, which support the MessageVortex Protocol. A 'node address' is typically an email address, an XMPP address or other transport protocol identity supporting the MessageVortex protocol. Any address SHOULD include a public part of an 'identity key' to allow messages to transmit safely. One or more addresses MAY belong to the same node.
A 'block' represents an ASN.1 sequence in a transmitted message. We embed messages in the transport protocol, and these messages may be of any size.
A nodeSpec block, as specified in Appendix A.6, expresses an addressable node in a unified format. The nodeSpec contains a reference to the routing protocol, the routing address within this protocol, and the keys required for addressing the node. This RFC specifies transport layers for XMPP and SMTP. Additional transport layers will require an extension to this RFC.
localPart = <local part of address> domain = <domain part of address> email = localPart "@" domain keySpec = <BASE64 encoded AsymmetricKey [DER encoded]> smtpAlternateSpec = localPart ".." keySpec ".." domain "@localhost" smtpUrl = "vortexsmtp://" smtpAlternateSpec
An alternative address representation is defined that allows a standard email client to address a Vortex node. A node SHOULD support the smtpAlternateSpec (its specification is noted in ABNF as in [RFC5234]). For applications with QR code support, an implementation SHOULD use the smtpUrl representation.
This representation does not support quoted local part SMTP addresses.
localPart = <local part of address>
domain = <domain part of address>
resourcePart = <resource part of the address>
jid = localPart "@" domain [ "/" resourcePart ]
keySpec = <BASE64 encoded AsymmetricKey [DER encoded]>;
jidAlternateSpec = localPart ".." keySpec ".."
domain "@localhost" [ "/" resourcePart ]
jidUrl = "vortexxmpp://" jidAlternateSpec
Typically, a node specification follows the ASN.1 block NodeSpec. For support of XMPP clients, an implementation SHOULD support the jidAlternateSpec (its specification is noted in ABNF as in [RFC5234]).
This document refers to two or more message sending or receiving entities as peer partners. One partner sends a message, and all others receive one or more messages. Peer partners are message specific, and each partner always connects directly to a node.
Several keys are required for a Vortex message. For identities and ephemeral identities (see below), we use asymmetric keys, while symmetric keys are used for message encryption.
Every participant of the network includes an asymmetric key, which SHOULD be either an EC key with a minimum length of 384 bits or an RSA key with a minimum length of 2048 bits.
The public key must be known by all parties writing to or through the node.
Peer keys are symmetrical keys transmitted with a Vortex message and are always known to the node sending the message, the node receiving the message, and the creator of the routing block.
A peer key is included in the Vortex message as well as the building instructions for subsequent Vortex messages (see RoutingCombo in Appendix A).
The sender key is a symmetrical key protecting the identity and routing block of a Vortex message. It is encrypted with the receiving peer key and prefixed to the identity block. This key further decouples the identity and processing information from the previous key.
A sender key is known to only one peer of a Vortex message and the creator of the routing block.
The term 'Vortex message' represents a single transmission between two routing layers. A message adapted to the transport layer by the blending layer is called a 'blended Vortex message' (see Section 3).
A complete Vortex message contains the following items:
A message is content to be transmitted from a single sender to a recipient. The sender uses a routing block either built itself or provided by the receiver to perform the transmission. While a message may be anonymous, there are different degrees of anonymity as described by the following.
A message is always MIME encoded as specified in [RFC2045].
MessageVortex uses a unique encoding for keys. This encoding is designed to be small and flexible while maintaining a specific base structure.
The following key structures are available:
MAC does not require a complete structure containing specs and values, and only a MacAlgorithmSpec is available. The following sections outline the constraints for specifying parameters of these structures where a node MUST NOT specify any parameter more than once.
If a crypto mode is specified requiring an IV, then a node MUST provide the IV when specifying the key.
Nodes use asymmetric keys for identifying peer nodes (i.e., identities) and encrypting symmetric keys (for subsequent de-/encryption of the payload or blocks). All asymmetric keys MUST contain a key type specifying a strictly-normed key. Also, they MUST contain a public part of the key encoded as an X.509 container and a private key specified in PKCS#8 wherever possible.
RSA and EC keys MUST contain a keySize parameter. All asymmetric keys SHOULD contain a padding parameter, and a node SHOULD assume PKCS#1 if no padding is specified.
NTRU specification MUST provide the parameters "n", "p", and "q".
Nodes use symmetric keys for encrypting payloads and control blocks. These symmetric keys MUST contain a key type specifying a key, which MUST be in an encoded form.
A node MUST provide a keySize parameter if the key (or, equivalently, the block) size is not standardized or encoded in the name. All symmetric key specifications MUST contain a mode and padding parameter. A node MAY list multiple padding or mode parameters in a ReplyCapability block to offer the recipient a free choice.
The term 'transport address' represents the token required to address the next immediate node on the transport layer. An email transport layer would have SMTP addresses, such as 'vortex@example.com,' as the transport address.
The peer identity may contain the following information of a peer partner:
Ephemeral identities are temporary identities created on a single node. These identities MUST NOT relate to another identity on any other node so that they allow bookkeeping for a node. Each ephemeral identity has a workspace assigned, and may also have the following items assigned.
An official identity may have the following items assigned.
Every official or ephemeral identity has a workspace, which consists of the following elements.
'Multi-use reply blocks' (MURB) are a special type routing block sent to a receiver of a message or request. A sender may use such a block one or several times to reply to the sender linked to the ephemeral identity, and it is possible to achieve sender anonymity using MURBs.
A vortex node MAY deny the use of MURBs by indicating a maxReplay equal to zero when sending a ReplyCapability block. An unobservable node SHOULD deny the use of MURBs.
This Document describes the version 1 of the protocol. The message PrefixBlock contains an optional version indicator. If absent protocol version 1 should be assumed.
The protocol is designed in four layers as shown in Figure 1.
+------------------------------------------------------------------+ | Vortex Node | | +--------------------------------------------------------------+ | | | Accounting | | | |______________________________________________________________| | | | | +--------------------------------------------------------------+ | | | Routing | | | |______________________________________________________________| | | | | +---------------------------+ +--------------------------------+ | | | Blending | | Blending | | | |___________________________| |________________________________| | |__________________________________________________________________| +---------------------------+ +--------------+ +---------------+ | Transport | | Transport in | | Transport out | |___________________________| |______________| |_______________|
Figure 1: Layer overview
Every participating node MUST implement the layer's blending, routing, and accounting. There MUST be at least one incoming and one outgoing transport layer available to a node. All blending layers SHOULD connect to the respective transport layers for sending and receiving packets.
The transport layer transfers the blended Vortex messages to the next vortex node and stores it until the next blending layer picks up the message.
The transport layer infrastructure SHOULD NOT be specific to anonymous communication and should contain significant portions of non-Vortex traffic.
The blending layer embeds blended Vortex Message into the transport layer data stream and extracts the packets from the transport layer.
The routing layer expands the information contained in MessageVortex packets, processes them, and passes generated packets to the respective blending layer.
The accounting layer tracks all ephemeral identities authorized to use a MessageVortex node and verifies the available quotas to an ephemeral identity.
+-+---+-+-+---+-+---+-+-+---+-+-+---+-+-------+-+ | | | | | | C | | | | | | R | | | | | | | | | | P | | | H | | | O | | | | | | M | | | | R | | | E | | | U | | P | | | | P | | | | E | | | A | | | T | | A | | | | R | | | | F | | | D | | | I | | Y | | | | E | | | | I | | | E | | | N | | L | | | | F | | | | X | | | R | | | G | | O | | | | I | | | +---+ | |___| | |___| | A | | | | X | | | k_h | sk_s | sk_s | D | | | |___| | |_______|_______|_______|_______| | | k_h | sk_p | |_______|___________________________________|
Figure 2: Vortex message overview
Figure 2 shows a Vortex message. The enclosed sections denote encrypted blocks, and the three or four-letter abbreviations denote the key required for decryption. The abbreviation k_h stands for the asymmetric host key, and sk_p is the symmetric peer key. The receiving node obtains this key by decrypting MPREFIX with its host key k_h. Then, sk_s is the symmetric sender key. When decrypting the MPREFIX block, the node obtains this key. The sender key protects the header and routing blocks by guaranteeing the node assembling the message does not know about upcoming identities, operations, and requests. The peer key protects the message, including its structure, from third-party observers.
The PrefixBlock contains a symmetrical key as defined in Appendix A.1 and is encrypted using the host key of the receiving peer host. The symmetric key utilized MUST be from the set advertised by a CapabilitiesReplyBlock (see Section 7.2.6). A node MAY choose any parameters omitted in the CapabilitiesReplyBlock freely unless stated otherwise in Section 7.2.6. A node SHOULD avoid sending unencrypted PrefixBlocks, and a prefix block MUST contain the same forward-secret as the other prefix as well as the routing and header blocks. A host MAY reply to a message with an unencrypted message block, but any reply to a message SHOULD be encrypted.
The sender MUST choose a key which may be encrypted with the host key in the respective PrefixBlock using the padding advertised by the CapabilitiesReplyBlock.
A node MUST always encrypt an InnerMessageBlock with the symmetric key of the PrefixBlock to hide the inner structure of the message. The InnerMessageBlock SHOULD always accommodate four or more payload chunks.
An InnerMessageBlock contains so-called forwardSecrets, a random number that MUST be the same in the PrefixBlock, HeaderBlock, RoutingBlock, and PrefixBlock. Nodes receiving messages containing non-matching forwardSecrets MUST discard these messages and SHOULD NOT send an error message. If a node receives too many messages with illegal forward secrets, then the node SHOULD delete this identity. A node receiving a message with a broken forwardSecret SHOULD treat the block as a replayed block and discard it regardless of a valid forwardSecret. Any replay within the replay protection time MUST be discarded regardless of a correct forward secret.
Control prefix (CPREFIX) and MPREFIX blocks share the same structure and logic as well as containing the sender key sk_s. If an MPREFIX block is unencrypted, a node MAY omit the CPREFIX block. An omitted CPREFIX block results in unencrypted control blocks (e.g., the HeaderBlock and RoutingBlock).
A prefix block MUST contain the same forwardSecret as the other prefix, the routing block, and the header block.
The control blocks of the HeaderBlock and a RoutingBlock contain the core information to process the payload.
The header block (see HeaderBlock in Appendix A) contains the following information.
The list of header requests MAY be one of the following.
If a header block violates these rules, then a node MUST NOT reply to any header request. The payload and routing blocks SHOULD still be added to the workspace and processed if the message quota is not exceeded.
The routing block (see RoutingBlock in Appendix A) contains the following information.
Each InnerMessageBlock with routing information SHOULD contain at least four PayloadChunks.
The MessageVortex protocol is a modular protocol that allows the use of different encryption algorithms. For its operation, a Vortex node SHOULD always support at least two distinct types of algorithms, paddings or modes such that they rely on two mathematical problems.
A node MUST support the following symmetric ciphers.
A node SHOULD support any standardized key larger than the smallest key size.
A node MAY support Twofish ciphers (see [TWOFISH]).
A node MUST support the following asymmetric ciphers.
A node MUST support the following Message Authentication Codes (MAC).
A node SHOULD support the following MACs.
A node MUST support the following paddings specified in [RFC8017].
A node MUST support the following modes.
A node SHOULD NOT use the following modes.
A node SHOULD support the following modes.
Each node supports a fixed set of blending capabilities, which may be different for incoming and outgoing messages.
The following sections describe the blending mechanism. There are currently two blending layers specified with one for the Simple Mail Transfer Protocol (SMTP, see [RFC5321]) and the second for the Extensible Messaging and Presence Protocol (XMPP, see [RFC6120]). All nodes MUST at least support "encoding=plain:0,256".
There are two types of blending supported when using attachments.
A node MUST support PLAIN blending for reasons of interoperability whereas a node MAY support blending using F5.
A blending layer embeds a VortexMessage in a carrier file with an offset for PLAIN blending. For replacing a file start, a node MUST use the offset 0. The routing node MUST choose the payload file for the message, and SHOULD use a credible payload type (e.g., MIME type) with high entropy. Furthermore, it SHOULD prefix a valid header structure to avoid easy detection of the Vortex message. Finally, a routing node SHOULD use a valid footer, if any, to a payload file to improve blending.
The blended Vortex message is embedded in one or more message chunks, each starting with two unsigned integers of variable length. The integer starts with the LSB, and if bit 7 is set, then there is another byte following. There cannot be more than four bytes where the last, fourth byte is always 8 bit. The three preceding bytes have a payload of seven bits each, which results in a maximum number of 2^29 bits. The first of the extracted numbers reflect the number of bytes in the chunk after the length descriptors. The second contains the number of bytes to be skipped to reach the next chunk. There exists no "last chunk" indicator. A chunk or the gap MAY surpass the end of the file.
position:00h 02h 04h 06h 08h ... 400h 402h 404h 406h 408h 40Ah value: 01 02 03 04 05 06 07 08 09 ... 01 05 0A 0B 0C 0D 0E 0F f0 03 12 13 Embedding: "(plain:1024)" Result: 0A 13 (+ 494 omitted bytes; then skip 12 bytes to next chunk)
A node SHOULD offer at least one PLAIN blending method and MAY offer multiple offsets for incoming Vortex messages.
A plain blending is specified as the following.
plainEncoding = "("plain:" <numberOfBytesOfOffset>
[ "," <numberOfBytesOfOffset> ]* ")"
For F5, a blending layer embeds a Vortex message into a jpeg file according to [F5]. The password for blending may be public, and a routing node MAY advertise multiple passwords. The use of F5 adds approximately tenfold transfer volume to the message. A routing block building node SHOULD only use F5 blending where appropriate.
A blending in F5 is specified as the following.
f5Encoding = "(F5:" <passwordString> [ "," <PasswordString> ]* ")"
Commas and backslashes in passwords MUST be escaped with a backslash whereas closing brackets are treated as normal password characters unless they are the final character of the encoding specification string.
Email messages with content MUST be encoded with Multipurpose Internet Mail Extensions (MIME) as specified in [RFC2045]. All nodes MUST support BASE64 encoding and MUST test all sections of a MIME message for the presence of a VortexMessage.
A vortex message is present if a block containing the peer key at the known offset of any MIME part decodes correctly.
A node SHOULD support SMTP blending for sending and receiving. For sending SMTP, the specification in [RFC5321] must be used. TLS layers MUST always be applied when obtaining messages using POP3 (as specified in [RFC1939] and [RFC2595]) or IMAP (as specified in [RFC3501]). Any SMTP connection MUST employ a TLS encryption when passing credentials.
For interoperability, an implementation SHOULD provide XMPP blending.
Blending into XMPP traffic is performed using the [XEP-0231] extension of the XMPP protocol.
PLAIN and F5 blending are acceptable for this transport layer.
An incoming message is considered initially unauthenticated. A node should consider a VortexMessage as authenticated as soon as the ephemeral identity is known and is not temporary.
For an unauthenticated message, the following rules apply.
A message is considered authenticated as soon as the identity used in the header block is known and not temporary. A node MUST NOT treat a message as authenticated if the specified maximum number of replays is reached. For authenticated messages, the following rules apply.
A routing node MUST decrement the message quota by one if a received message is authenticated, valid, and contains at least one payload block. If a message is identified as duplicate according to the reply protection, then a node MUST NOT decrement the message quota.
function incomming_message(VortexMessage blendedMessage) {
try{
msg = unblend( blendedMessage );
if( not msg ) {
// Abort processing
throw exception( "no embedded message found" )
} else {
hdr = get_header( msg )
if( not known_identity( hdr.identity ) {
if( get_requests( hdr ) contains HeaderRequestIdentity ) {
create_new_identity( hdr ).set_temporary( true )
send_message( create_requirement( hdr ) )
} else {
// Abort processing
throw exception( "identity unknown" )
}
} else {
if( is_duplicate_or_replayed( msg ) ) {
// Abort processing
throw exception "duplicate or replayed message" )
} else {
if( get_accounting( hdr.identity ).is_temporary() ) {
if( not verify_requirement( hdr.identity, msg ) ) {
get_accounting( hdr.identity ).set_temporary( false )
}
}
if( get_accounting( hdr ).is_temporary() ) {
throw exception( "no processing on temporary identity" )
}
// Message authenticated
get_accounting( hdr.identity ).register_for_replay_protection( msg )
if( not verify_mtching_forward_secrets( msg ) ) {
throw exception( "forward secret missmatch" )
}
if( contains_payload( msg ) ) {
if( get_accounting( hdr.identity ).decrement_message_quota() ) {
while index,nextPayloadBlock = get_next_payload_block( msg ) {
add_workspace( header.identity, index, nextPayloadBlock )
}
while nextRoutingBlock = get_next_routing_block( msg ) {
add_workspace( hdr.identity, add_routing( nextRoutingBlock ) )
}
process_reserved_mapping_space( msg )
while nextRequirement = get_next_requirement( hdr ) {
add_workspace( hdr.identity, nextRequirement )
}
} else {
throw exception( "Message quota exceeded" )
}
}
}
}
} catch( exception e ) {
// Message processing failed
throw e;
}
}
The message processing works according pseudo-code shown below.
A routing workspace consists of the following items.
The accounting layer typically triggers processing and represents either a cleanup action or a routing event. A cleanup event deletes the following information from all workspaces.
Note that maxProcessTime reflects the number of seconds since the arrival of the last octet of the message at the transport layer facility. A node SHOULD NOT take additional processing time (e.g., for anti-UBE or anti-virus) into account.
The accounting layer triggers routing events occurring at least the minProcessTime after the last octet of the message arrived at the routing layer. A node SHOULD choose the latest possible moment at which the peer node receives the last octet of the assembled message before the maxProcessTime is reached. The calculation of this last point in time where a message may be set SHOULD always assume that the target node is working. A sending node SHOULD choose the time within these bounds randomly. An accounting layer MAY trigger multiple routing combos in bulk to further obfuscate the identity of a single transport message.
First, the processing node escapes the payload chunk at ID 0 if needed (e.g., a non-special block is starting with a backslash). Next, it executes all processing instructions of the routing combo in the specified sequence. If an instruction fails, then the block at the target ID of the operation remains unchanged. The routing layer proceeds with the subsequent processing instructions by ignoring the error. For a detailed description of the operations, see Section 7.4. If a node succeeds in building at least one payload chunk, then a VortexMessage is composed and passed to the blending layer.
The blending layer MUST compose a transport layer message according to the specification provided in the routing combo. It SHOULD choose any decoy message or steganographic carrier in such a way that the dead parrot syndrome, as specified in [DeadParrot], is avoided.
Header requests are control requests for the anonymization system. Messages with requests or replies only MUST NOT affect any quota.
Requesting a new ephemeral identity is performed by sending a message containing a header block with the new identity and an identity creation request (HeaderRequestIdentity) to a node. The node MAY send an error block (see Section 7.3.1) if it rejects the request.
If a node accepts an identity creation request, then it MUST send a reply. A node accepting a request without a requirement MUST send back a special block containing "no error". A node accepting a request under the precondition of a requirement to be fulfilled MUST send a special block containing a requirement block.
A node SHOULD NOT reply to any clear-text requests if the node does not want to disclose its identity as a Vortex node officially. A node MUST reply with an error block if a valid identity is used for the request.
Any valid ephemeral identity may request an increase of the current message quota to a specific value at any time. The request MUST include a reply block in the header and may contain other parts. If a requested value is lower than the current quota, then the node SHOULD NOT refuse the quota request and SHOULD send a "no error" status.
A node SHOULD reply to a HeaderRequestIncreaseMessageQuota request (see Appendix A) of a valid ephemeral identity. The reply MUST include a requirement, an error message or a "no error" status message.
A node may request to increase the current message quota by sending a HeaderRequestIncreaseMessageQuota request to the routing node. The value specified within the node is the new quota. HeaderRequestIncreaseMessageQuota requests MUST include a reply block, and a node SHOULD NOT use a previously sent MURB to reply.
If the requested quota is higher than the current quota, then the node SHOULD send a "no error" reply. If the requested quota is not accepted, then the node SHOULD send a requestedQuotaOutOfBand reply.
A node accepting the request MUST send a RequirementBlock or a "no error block."
Any valid ephemeral identity may request to increase the current transfer quota to a specific value at any time. The request MUST include a reply block in the header and may contain other parts. If a requested value is lower than the current quota, then the node SHOULD NOT refuse the quota request and SHOULD send a "no error" status.
A node SHOULD reply to a HeaderRequestIncreaseTransferQuota request (see Appendix A) of a valid ephemeral identity. The reply MUST include a requirement, an error message or a "no error" status message.
Any valid ephemeral identity may request the current message and transfer quota. The request MUST include a reply block in the header and may contain other parts.
A node MUST reply to a HeaderRequestQueryQuota request (see Appendix A), which MUST include the current message quota and the current message transfer quota. The reply to this request MUST NOT include a requirement.
Any node MAY request the capabilities of another node, which include all information necessary to create a parseable VortexMessage. Any node SHOULD reply to any encrypted HeaderRequestCapability.
A node SHOULD NOT reply to clear-text requests if the node does not want to disclose its identity as a Vortex node officially. A node MUST reply if a valid identity is used for the request, and it MAY reply to unknown identities.
A node may ask another node for a list of routing node addresses and keys, which may be used to bootstrap a new node and add routing nodes to increase the anonymization of a node. The receiving node of such a request SHOULD reply with a requirement (e.g., RequirementPuzzleRequired).
A node MAY reply to a HeaderRequest request (see Appendix A) of a valid ephemeral identity, and the reply MUST include a requirement, an error message or a "no error" status message. A node MUST NOT reply to an unknown identity, and SHOULD always reply with the same result set to the same identity.
This request type allows a receiving node to replace an existing identity with the identity provided in the message, and is required if an adversary manages to deny the usage of a node (e.g., by deleting the corresponding transport account). Any sending node may recover from such an attack by sending a valid authenticated message to another identity to provide the new transport and key details.
A node SHOULD reply to such a request from a valid known identity, and the reply MUST include an error message or a "no error" status message.
This request type allows a node to request a new version of the software in an anonymous, unliked manor. The identifier MUST identify the software product uniquely. The version MUST reflect the version tag of the currently installed version or a similarly usable tag.
Special blocks are payload messages that reflect messages from one node to another and are not visible to the user. A special block starts with the character sequence '\special' (or 5Ch 73h 70h 65h 63h 69h 61h 6Ch) followed by a DER encoded special block (SpecialBlock). Any non-special message decoding to ID 0 in a workspace starting with this character sequence MUST escape all backslashes within the payload chunk with an additional backslash.
An error block may be sent as a reply contained in the payload section. The error block is embedded in a special block and sent with any provided reply block. Error messages SHOULD contain the serial number of the offending header block and MAY contain human-readable text providing additional messages about the error.
If a node is receiving a requirement block, then it MUST assume that the request block is accepted, is not yet processed, and is to be processed if it meets the contained requirement. A node MUST process a request as soon as the requirement is fulfilled, and MUST resend the request as soon as it meets the requirement.
A node MAY reject a request, accept a request without a requirement, accept a request upon payment (RequirementPaymentRequired), or accept a request upon solving a proof of work puzzle (RequirementPuzzleRequired).
If a node requests a puzzle, then it MUST send a RequirementPuzzleRequired block. The puzzle requirement is solved if the node receiving the puzzle is replying with a header block that contains the puzzle block, and the hash of the encoded block begins with the bit sequence mentioned in the puzzle within the period specified in the field 'valid.'
A node solving a puzzle requires sending a VortexMessage to the requesting node, which MUST contain a header block that includes the puzzle block and MUST have a MAC fingerprint starting with the bit sequence as specified in the challenge. The receiving node calculates the MAC from the unencrypted DER encoded HeaderBlock with the algorithm specified by the node. The sending node may achieve the requirement by adding a proofOfWork field to the HeaderBlock containing any content fulfilling the criteria. The sending node SHOULD keep the proofOfWork field as short as possible.
If a node requests a payment, then it MUST send a RequirementPaymentRequired block. As soon as the requested fee is paid and confirmed, the requesting node MUST send a "no error" status message. The usage period 'valid' describes the period during which the payment may be carried out. A node MUST accept the payment if occurring within the 'valid' period but confirmed later. A node SHOULD return all unsolicited payments to the sending address.
If a node requests an upgrade a ReplyUpgrade block MAY be sent. The block must contain the identifier and version of the most recent software version. The blob MAY contain the software if there is a newer one available.
Routing operations are contained in a routing block and processed upon arrival of a message or when compiling a new message. All operations are reversible, and no operation is available for generating decoy traffic, which may be used through encryption of an unpadded block or the addRedundancy operation.
All payload chunk blocks inherit the validity time from the message routing combos as arrival time + max(maxProcessTime).
When applying an operation to a source block, the resulting target block inherits the expiration of the source block. When multiple expiration times exist, the one furthest in the future is applied to the target block. If the operation fails, then the target expiration remains unchanged.
The straightforward mapping operation is used in inOperations of a routing block to map the routing block's specific blocks to a permanent workspace.
The split and merge operations allow splitting and recombining message chunks. A node MUST adhere to the following constraints.
An operation MUST fail if relative values are equal to, or less than, zero. An operation MUST fail if a relative value is equal to, or greater than, 100. All floating-point operations must be performed according to [IEEE754] and in 64-bit precision.
Encryption and decryption are executed according to the standards mentioned above. An encryption operation encrypts a block symmetrically and places the result in the target block. The parameters MUST contain IV, padding, and cipher modes. An encryption operation without a valid parameter set MUST fail.
The addRedundancy and removeRedundancy operations are core to the protocol. They may be used to split messages and distribute message content across multiple routing nodes. The operation is separated into three steps.
The following sections describe the order of the operations within an addRedundancy operation. For a removeRedundancy operation, invert the functions and order. If the removeRedundancy has more than the required blocks to recover the information, then it should take only the required number beginning from the smallest. If a seed and PRNG are provided, then the removeRedundancy operation MAY test any combination until recovery is successful.
A processing node calculates the final length of all payload blocks, including redundancy. This is done by L=roof((<input block size in bytes>+4)/<encryption block size in bytes>)*<encryption block size in bytes>. The block is prepended with a 32-bit unit length indicator in bytes (little-endian). This length indicator, i, is calculated by i=<input block size in bytes>*randominteger\cdot L. The remainder of the input block, up to length L, is padded with random data. A routing block builder should specify the value of the $randomInteger$. If not specified the routing node may choose a random positive integer value. A routing block builder SHOULD specify a PRNG and a seed used for this padding. If GF(16) is applied, then all numbers are treated as little-endian representations. Only GF(8) and GF(16) are allowed fields.
For padding removal, the padding i at the start is first removed as a little-endian integer. Second, the length of the output block is calculated by applying <output block size in bytes>=i mod <input block size in bytes>
This padding guarantees that each resulting block matches the block size of the subsequent encryption operation and does not require further padding.
Next, the input block is organized in a data matrix D of dimensions (inrows, incols) where incols=(<number of data blocks>-<number of redundancy blocks>) and inrows=L/(<number of data blocks>-<number of redundancy blocks>). The input block data is first distributed in this matrix across, and then down.
Next, the data matrix D is multiplied by a Vandermonde matrix V with its number of rows equal to the incols calculated and columns equal to the <number of data blocks>. The content of the matrix is formed by v(i,j)=pow(i,j), where i reflects the row number starting at 0, and j reflects the column number starting at 0. The calculations described must be carried out in the GF noted in the respective operation to be successful. The completed operation results in matrix A.
Each row vector of A is a new data block encrypted with the corresponding encryption key noted in the keys of the addRedundancyOperation. If there are not enough keys available, then the keys used for encryption are reused from the beginning after the final key is used. A routing block builder SHOULD provide enough keys so that all target blocks may be encrypted with a unique key. All encryptions SHOULD NOT use padding.
The accounting layer triggers processing according to the information contained in a routing block in the workspace. All operations MUST be executed in the sequence provided in the routing block, and any failing operation must leave the result block unmodified.
All workspace blocks resulting in IDs of 1 to maxPayloadBlock are then added to the message and passed to the blending layer with appropriate instructions.
The accounting layer has two types of operations.
Implementations MUST provide sufficient locking mechanisms to guarantee the integrity of accounting information and the workspace at any time.
The accounting layer SHOULD keep a list of expiration times. As soon as an entry (e.g., payload block or identity) expires, the respective structure should be removed from the workspace. An implementation MAY choose to remove expired items periodically or when encountering them during normal operation.
The accounting layer MAY keep a list of when a routing block is activated. For improved privacy, the accounting layer should use a slotted model where, whenever possible, multiple routing blocks are handled in the same period, and the requests to the blending layers are mixed between the transactions.
A node MUST update quotas on the respective operations. For example, a node MUST decrease the message quota before processing routing blocks in the workspace and after the processing of header requests.
The transfer quota MUST be checked and decreased by the number of data bytes in the payload chunks after an outgoing message is processed and fully assembled. The message quota MUST be decreased by one on each routing block triggering the assembly of an outgoing message.
Any packet may request the creation of an ephemeral identity. A node SHOULD NOT accept such a request without a costly requirement since the request includes a lifetime of the ephemeral identity. The costs for creating the ephemeral identity SHOULD increase if a longer lifetime is requested.
Thanks go to my family who supported me with patience and countless hours as well as to Mark Zeman for his feedback challenging my thoughts and peace.
This memo includes no request to IANA.
Additional encryption algorithms, paddings, modes, blending layers or puzzles MUST be added by writing an extension to this or a subsequent RFC. For testing purposes, IDs above 1,000,000 should be used.
The MessageVortex protocol should be understood as a toolset instead of a fixed product. Depending on the usage of the toolset, anonymity and security are affected. For a detailed analysis, see [MVAnalysis].
The primary goals for security within this protocol rely on the following focus areas.
These aspects are affected by the usage of the protocol, and the following sections provide additional information on how they impact the primary goals.
The Vortex protocol does not rely on any encryption of the transport layer since Vortex messages are already encrypted. Also, confidentiality is not affected by the protection mechanisms of the transport layer.
If a transport layer supports encryption, then a Vortex node SHOULD use it to improve the privacy of the message.
Anonymity is affected by the inner workings of the blending layer in many ways. A Vortex message cannot be read by anyone except the peer nodes and routing block builder. The presence of a Vortex node message may be detected through the typical high entropy of an encrypted file, broken structures of a carrier file, a meaningless content of a carrier file or the contextless communication of the transport layer with its peer partner. A blending layer SHOULD minimize the possibility of simply detection by minimizing these effects.
A blending layer SHOULD use carrier files with high compression or encryption. Carrier files SHOULD NOT have inner structures such that the payload is comparable to valid content. To achieve undetectability by a human reviewer, a routing block builder should use F5 instead of PLAIN blending. This approach, however, increases the protocol overhead by approximately tenfold.
The two layers of 'routing' and 'accounting' have the deepest insight into a Vortex message's inner working. Each knows the immediate peer sender and the peer recipients of all payload chunks. As decoy traffic is generated by combining chunks and applying redundancy calculations, a node can never know if a malfunction (e.g., during a recovery calculation) was intended. Therefore, a node is unable to distinguish a failed transaction from a terminated transaction as well as content from decoy traffic.
A routing block builder SHOULD follow the following rules not to compromise a Vortex message's anonymity.
An active adversary cannot use blocks from other routing block builders. While the adversary may falsify the result by injecting an incorrect message chunk or not sending a message, such message disruptions may be detected by intentionally routing information to the routing block builder (RBB) node. If the Vortex message does not carry the information expected, then the node may safely assume that one of the involved nodes is misbehaving. A block building node MAY calculate reputation for involved nodes over time and MAY build redundancy paths into a routing block to withstand such malicious nodes.
Receiver anonymity is at risk if the handling of the message header and content is not done with care. An attacker might send a bugged message (e.g., with a DKIM or DMARC header) to deanonymize a recipient. Careful attention is required when handling anything other than local references when processing, verifying, or rendering a message.
The following sections contain the ASN.1 modules specifying the MessageVortex Protocol.