DRIP Authentication Formats

UA Systems (UAS) are usually in a volatile environment when it comes to communication. UA are generally small with little computational (or flying) horsepower to carry standard communication equipment. This limits the mediums of communication to few viable options. Observer systems (e.g. smartphones and tablets) place further constraints on the communication options. The Remote ID Broadcast messages MUST be available to applications on these platforms without modifying the devices. The ASTM standard focuses on two ways of communicating to a UAS for RID: Broadcast and Network. This document will focus on adding trust to Broadcast RID in the current (and an expanded) Authentication Message format.

The following will be addressed: This will be addressed using the Certificate Message types ( and ). This requirement is addressed using the Wrapped ASTM Message , Manifest Message and Message Pack Signature types . This requirement is addressed using the Certificate Message types ( and ).

The key words “MUST”, “MUST NOT”, “REQUIRED”, “SHALL”, “SHALL NOT”, “SHOULD”, “SHOULD NOT”, “RECOMMENDED”, “NOT RECOMMENDED”, “MAY”, and “OPTIONAL” in this document are to be interpreted as described in BCP 14 when, and only when, they appear in all capitals, as shown here.

See for common DRIP terms. In this document whenever the word Aircraft is used it is referring to an Unmanned Aircraft (UA) not a Manned Aircraft.

The current standard for Remote ID (RID) does not, in any meaningful capacity, address the concerns of trust in the UA space with communication in the Broadcast RID environment. This is a requirement that will need to be addressed eventually for various different parties that have a stake in the UA industry. The following subsections will provide a high level reference to the ASTM standard for Authentication Messages and how their current limitations effect trust in the Broadcast RID environment.

The above diagram is the format defined by ASTM that is the frame which everything this document fits into. The specific details of the ASTM headers are abstracted away as they are not necessarily required for this document.

Currently the ASTM AuthType of 0xD should be used to denote DRIP based Authentication. The max page count of the Authentication Message is increased to 10, instead of being capped at 5.

The DRIP Header is used to signal what kind of Authentication under DRIP that the message is using and consists of two fields. The Most Significant Bit is used to signal if FEC is present in the final page of the Authentication Message. It MUST be set to 1 if FEC is being used. The lower 7 bits are used as the DRIP AuthType field denoting what Authentication type is being used. There are 5 major areas carved out of the DRIP AuthType defined by the following bitmaps:

This field has a maximum size of 223 bytes. If the data is less than 223 bytes and a page is only partially filled then the rest of the partially filled page must be null padded.

To help Bluetooth (specifically Bluetooth 4) achieve the goal of reliable receipt of paged messages a Forward Error Correction (FEC) scheme is introduced and SHOULD be used for Bluetooth 4 and SHOULD NOT be used for Bluetooth 5 under DRIP. Due to the nature of Bluetooth 4 and the existing ASTM paging structure an optimization can be used. If a Bluetooth frame fails its CRC check, then the frame is dropped without notification to the upper protocol layers. From the Remote ID perspective this means the loss of a complete frame/message/page. In Authentication Messages, each page is already numbered so the loss of a page allows the receiving application to build a “dummy” page filled with nulls (other than the ASTM Header and Auth Header which is known). A compliant implementation of this standard MUST use XOR for the FEC. When generating the parity the first two bytes of every Authentication Page MUST be exclude from the XOR operation. For pages 1 through N this leaves the data portion of the page while page 0 will include a number of headers along with 17 bytes of data. This parity is appended in one full page (always the last) allowing for recovery when any single page is lost in transmission. If more than one page is lost (>1/5 for 5 page messages, >1/10 for 10 page messages) than the error rate of the link is already beyond saving and the application has more issues to deal with.

This format MUST be encapsulated by the General Frame () and reside in its data field ().

To avoid needing the UAs HHIT via the ASTM Basic ID in a detached fashion the 16 byte HHIT of the UA is included in the wrapper frame. The HHIT for the UA (and other entities in the RID and greater UTM system under DRIP) is an enhancement of the Host Identity Tag (HIT) introducing hierarchy (and how they are used in UAS RID) as defined in .

The Trust Timestamp is of the format defined in . That is a UNIX timestamp offset by 01/01/2019 00:00:00. An offset is then added to push the timestamp a short time into the future to avoid replay attacks. When wrapping a Vector (Position) Message the payload WILL contain (by ASTM rules) constantly changing data, this includes its own timestamp. This timestamp is only 2 bytes, which is easily attacked and only expresses the 1/10th of seconds since the last hour. Other ASTM message types, such as Basic ID and Self-ID are static messages with no changing data. To protect a replay of these signed messages the Trust Timestamp is the field during signing to be guaranteed to change. The offset used against the UNIX timestamp is not defined in this document. Best practices to identify a acceptable offset should be used taking into consideration the UA environment, and propagation characteristics of the messages being sent.

This field has a maximum of 116 bytes in length.

The wrapper signature is generated using the private key half of the the UAs Host Identity (HI) and is done over all preceding data. ASTM/DRIP Headers are exclude from this operation only information within the Wrapper Fame () is signed.

This format MUST be encapsulated by the General Frame () and reside in its data field (). This format is typically used to form a complete certificate using attestation data from a Registry defined in .

Any data up to 155 bytes in length that the UA wishes to assert truth to.

Generated by the UA during the creation of the Authentication message. It is set a short time into the future to protect against replay attacks of this DRIP format. It shares the same format as the Trust Timestamp ().

Performed by the UA using the onboard keypair which matches the HHIT in the Basic ID Message (0x0).

With Bluetooth 4.X formatting the goal is to attempt to bring reliable receipt of paged messages. Unless otherwise specified the FEC Bit of the DRIP Header MUST be set to 1 when using Bluetooth 4 to take advantage of FEC for lost frames in transmission.

This DRIP Authentication type uses the Wrapper Frame format (), filling the data field () with up to 4 ASTM Messages (all types except Message Pack (0xF) and Authentication (0x2)). When multiple ASTM Messages are being wrapped they MUST be in Message Type order.

This DRIP Authentication type uses the Wrapper Frame () format, filling the data () with ASTM Messages (all types except Message Pack (0xF) and Authentication (0x2)). ASTM Messages being wrapped MUST be in Message Type order.

When wrapping a single ASTM Message the 25 byte payload actually causes an inefficiency in the framing format, create a whole page unused except for a single byte. This can be optimized by removing a single byte out of the wrapped message but creates an issue on the receiver of where the byte was removed. When sending a Location Message (Message Type 0x1) a single byte can be removed at the end of the message as it is currently unused. Many other messages in the ASTM Message set however do not have this ability. The first byte can not be removed as it is the key to know how to decode the message.

This DRIP Authentication type uses the Wrapper Frame (), filling the data field () with hashes of previously sent messages. By hashing previously sent messages and signing them we gain trust in UAs previous reports. An observer who has been listening for any length of time can hash received messages and cross check against listed hashes. The signature is signed across the list of hashes.

The hash algorithm used for the Manifest Message is the same hash algorithm used in creation of the HHIT that is signing the Manifest. A standard HHIT would be using cSHAKE128 from . With cSHAKE128, the hash is computed as follows:

The message MAC is prepended to the message, as the MAC is the only information that links UA messages from a specific UA.

Two special hashes are included; a previous manifest hash, which links to the previous manifest message, as well as a current manifest hash. This gives a pseudo-blockchain provenance to the manifest message that could be traced back if the observer was present for extended periods of time. During creation and signing of this message format this field MUST be set to 0. So the signature will be based on this field being 0, as well as its own hash. It is an open question of if we compute the hash, then sign or sign then compute. There a few different ways to cycle this message. We can “roll up” the hash of ‘current’ to ‘previous’ when needed or to completely recompute the hash. This mostly depends on the previous note.

A potential limitation to this format is dwell time of the UA. If the UA is not sticking to a general area then most likely the Observer will not obtain many (if not all) of the messages in the manifest. Without the original messages received no verification can be done. Examples of such scenarios include delivery or survey UA.

This DRIP Authentication type uses the Attestation Frame (), filling the data field ().

What this grants is the ability to authenticate UA Remote ID when the receiving device of the observer (e.g. a smartphone with a dedicated RID application) has no Internet service (e.g. LTE signal).

The Attestation: Registry on Aircraft is in practice a binding claim between the Registry and the Aircraft, asserting the relationship between the two entities. Importantly this certificate allows offline signature verification from the UA. This is as the UA HI is included in the structure. Also included is the HHIT of the Registry to check the local shortlist of Registries that the Observer device trusts (mapping HHITs to HIs). More details about this Attestation and other certificates and the provisioning process can be found in .

Under ASTM Bluetooth 4.X rules, transmission of dynamic messages are at least every 1 second while static messages (which is what Authentication is classified under) are sent at least every 3 seconds. Under DRIP the Certificate Message MUST be transmitted to properly meet the GEN 1 and GEN 3 requirement. The ASTM Message Wrapper and Manifest both satisfy the GEN 2 requirement. At least one MUST be implemented to comply with the GEN 2 requirement. A single Manifest can carry at most (using the full 10 page limit and 8 byte hashes) 12 unique hashes of previously sent messages (of any type). This results in a total of 22 (12 + 10) frames of Bluetooth data being transmitted over Bluetooth. In comparison the Message Wrapper sends 6 pages (each a single frame) for each wrapped message. For backwards compatibility the implementation should also send the standard ASTM message that was wrapped for non-DRIP compliant receivers to obtain. This method results in 84 total Bluetooth frames (12 + (12 * 6)) sent. The question of which is better suited is up to the implementation.

Under ASTM specification, Bluetooth 5 transport of Remote ID is to use the Message Pack (Type 0xF) format for all transmissions. Under Message Pack all messages are sent together (in Message Type order) in a single Bluetooth frame (up to 250 bytes). Message Packs are required by ASTM to be sent at a rate of 1 per second (like dynamic messages). This gives the benefit of no longer is there any message or page fragmentation in transmission. For this reason the recommended use of FEC such as XOR used in Bluetooth 4.X is not needed here and is impractical. Any of the Bluetooth 4.X formats can theoretically be used during Bluetooth 5 operation under ASTM, however the following subsections define a number of formats optimized for Message Pack and Bluetooth 5.

This format uses the Wrapper Frame () and virtually signs all other messages in the Message Pack.

This DRIP Authentication type uses the Attestation Frame (), filling the data field (). It is identical to the DRIP Certificate under Bluetooth 5 () except that FEC is turned off due to being under Bluetooth 5.

(Ed. Note: do we mandate the transmission of a Location Message as well?)

Increase Authentication Max Page Count from 5 to 10. Add Authentication Type for DRIP.

This document does not require any actions by IANA.

TODO (Ed. Note: Hash lengths (length vs strength/collision rate); replay attacks with timestamps; static Cra (issue but nulled if UA signing other stuff dynamically meaning signatures will fail as HI won’t match - this is probably a deeper discussion topic for provisioning security considerations when we get to there))

Ryan Quigley and James Mussi of AX Enterprize, LLC for early prototyping to find holes in the draft specifications.

The format standardized by the ASTM is designed with a few major considerations in mind, which the authors of this document feel put significant limitations on the expansion of the standard. The primary consideration (in this context) is the use of the Bluetooth 5.X Extended Frame format. This method allows for a 255 byte payload to be sent in what the ASTM refers to as a “Message Pack”. The idea is to include up to five standard ASTM Broadcast RID messages (each of which are 25 bytes) plus a single authentication message (5 pages of 25 bytes each) in the Message Pack. The reasoning is then the Authentication Message is for the entire Message Pack. The authors have no issues with this proposed approach; this is a valid format to use for the Authentication Message provided by the ASTM. However, by limiting the Authentication Message to ONLY five pages in the standard it ignores the possibility of other formatting options to be created and used. Another issue with this format, not fully addressed in this document is fragmentation. Under Bluetooth 4.X, each page is sent separately which can result in lose of pages on the receiver. This is disastrous as the loss of even a single page means any signature is incomplete. With the current limitation of 5 pages, Forward Error Correction (FEC) is nearly impossible without sacrificing the amount of data sent. More pages would allow FEC to be performed on the Authentication Message pages so loss of pages can be mitigated. All these problems are further amplified by the speed at which UA fly and the Observer’s position to receive transmissions. There is no guarantee that the Observer will receive all the pages of even a 5 page Authentication Message in the time it takes a UA to traverse across their line of sight. Worse still is that is not including other UA in the area, which congests the spectrum and could cause further confusion attempting to collate messages from various UA. This specific problem is out of scope for this document and our solutions in general, but should be noted as a design consideration.