RFC 5056 Quiz (EN)

Q1: What security question can a channel binding answer?

Multiple Choice

An agent authenticates with a SASL mechanism inside a TLS connection. The profile wants to detect an attacker that terminates TLS and relays the application authentication over a different TLS connection.

A. Whether the authenticated agent is authorized to invoke the requested A2A operation B. Whether the application-layer authentication and each peer refer to the same secure-channel instance C. Whether the TLS certificate remains within its validity interval and certification path constraints D. Whether two authenticated application messages express the same business transaction

**Explanation:** A: Authorization is a separate application or deployment decision. B: RFC 5056 Sections 2 and 3 define channel binding as tying application-layer authentication to a secure channel so both peers can establish that they observed the same binding. This can expose a channel-relay attack. C: Certificate path validation is not the channel-binding function described by RFC 5056. D: Business-message equivalence requires application semantics that a channel binding does not supply.

Q2: How should a profile bind authentication in nested channels?

Multiple Choice

An A2A connection has two nested secure channels of the same channel-binding type. Independent implementations also serialize the selected binding data before comparing it.

A. Use the outermost channel and allow each implementation to choose its native object encoding B. Concatenate every nested binding and compare the resulting implementation-specific structures C. Name the binding type, use the innermost channel when the same type is nested, and define one canonical octet encoding D. Use whichever channel completed most recently and treat locally successful encodings as equivalent

**Explanation:** A: RFC 5056 Section 2.1 selects the innermost instance for nested channels of the same type and requires a canonical octet encoding. B: RFC 5056 does not define concatenation as the nesting rule, and implementation-specific structures are not an interoperable comparison value. C: This applies all three Section 2.1 requirements relevant to the scenario: identify the type, select the innermost same-type channel, and use canonical bytes. D: Completion order and local success do not establish that both peers compared the same binding.

Q3: What can the backend conclude from a forwarded binding value?

Multiple Choice

A gateway terminates the client TLS connection, derives a channel binding, and forwards that value in an ordinary HTTP header over a separate gateway-to-backend TLS connection. The backend wants to claim that its own application authentication is bound directly to the client TLS channel.

A. It may make the direct-binding claim because both network hops use TLS and the binding value is unchanged B. It may make the claim if the gateway authenticated the client, even when the forwarded header is not integrity-protected as a gateway assertion C. It must reject every gateway architecture because channel bindings permit only end-to-end transports D. No. The header proves neither direct client-backend binding nor its own integrity; use backend-verifiable evidence or a protected gateway assertion

**Explanation:** A: Two TLS hops are two channel instances; copying a value does not merge their endpoints. B: Gateway authentication can support a delegated trust model, but an unprotected header does not establish an authentic gateway assertion to the backend. C: RFC 5056 does not prohibit trusted intermediaries; it requires the security claim to match the channel and application authentication actually bound. D: RFC 5056 Sections 3 and 8 require the application authentication and channel-binding construction to be analyzed together. A gateway assertion can be a valid separate design, but its trust boundary and integrity are local profile responsibilities.

Q4: Which kind of binding fits a per-session replay requirement?

Multiple Choice

A profile requires an authentication proof captured on one secure-channel session to fail after either peer establishes a new session, even when the same server certificate is used.

A. An endpoint binding, because certificate continuity is the profile's decisive property B. A unique channel binding whose value changes across the sessions under review C. Either kind, because application authentication makes binding uniqueness irrelevant D. An endpoint binding plus a longer validity interval, so the proof remains stable during reconnection

**Explanation:** A: RFC 5056 Section 2.1 distinguishes endpoint bindings, which can identify a channel endpoint across multiple channel instances, from unique bindings. B: A binding must be unique across the sessions relevant to the threat model to detect replay on a re-established channel; Section 8.1 warns about non-unique bindings in such cases. C: Application authentication is necessary, but it does not turn a reused channel-binding value into a per-session value. D: Increasing the lifetime strengthens continuity, the opposite of the stated replay criterion.

Q5: What is missing when binding data is compared on a separate exchange?

Multiple Choice

Two agents derive the same channel-binding type from a secure channel. They later send their binding values to an A2A coordinator over a separate exchange whose messages can be modified in transit. The coordinator treats equality as proof of the original binding.

A. Nothing, because equality of correctly derived values authenticates every later carrier of those values B. Only confidentiality, because modified binding messages cannot affect the equality result C. Integrity-protect the separate exchange and authenticate each reporter so the coordinator can attribute the compared values D. A second endpoint binding, because two binding types automatically authenticate the coordinator exchange

**Explanation:** A: Correct derivation does not protect a different exchange that carries the value. B: RFC 5056 Section 3 requires at least integrity when a separate protected exchange is used; the coordinator also needs application authentication to know whose observation it is comparing. C: This preserves the integrity of the comparison and supplies the identity association required by the channel-binding model. D: Adding another value does not authenticate an attacker-modifiable carrier.

Q6: Which gateway design preserves an accurate security claim?

Multiple Choice

A gateway validates an agent grant and verifies that the agent's authentication is bound to the client-to-gateway channel. A backend owns the requested resource and makes the final authorization decision over a different TLS connection. The backend receives only a gateway-generated assertion.

A. Make the gateway a trusted verifier and protect an assertion binding its result to client identity and request context; the backend still applies resource authorization B. Forward the original binding bytes and let the backend describe them as its own channel binding because both hops are within one deployment C. Treat a valid grant as proof of channel binding and omit the gateway's binding result from the assertion D. Treat a successful gateway binding as final resource authorization because it proves both agent identity and operation permission

**Explanation:** A: RFC 5056 Sections 3 and 8 support the gateway's client-channel claim, while the protected assertion and backend authorization are explicit composition and local trust decisions. The backend must not relabel the gateway-observed channel as a direct client-to-backend channel. B: Administrative proximity does not change the endpoints or instance from which binding data was derived. C: Grant validation and channel binding answer independent questions; one does not imply the other. D: Binding authentication to a channel does not grant permission to a backend-owned resource.

On the Use of Channel Bindings to Secure Channels

References (URLs)

Q1: What security question can a channel binding answer?

Q2: How should a profile bind authentication in nested channels?

Q3: What can the backend conclude from a forwarded binding value?

Q4: Which kind of binding fits a per-session replay requirement?

Q5: What is missing when binding data is compared on a separate exchange?

Q6: Which gateway design preserves an accurate security claim?