The swarm contains a known number n of low-end devices and an honest verifier. Its communication graph must remain connected and quasi-static during an attestation session; mobility is allowed only when it does not disrupt the protocol's message propagation.

1. threat model

   Remote and local adversary

   defined

   The adversary may alter unprotected software and state and fully control communication by eavesdropping, injection, deletion, delay, or modification. Physical non-intrusive and intrusive attacks are expressly outside the security model.

   Swarm, connectivity, adversary, and verifier model · Attack vectors and informal security analysis
2. assumption

   SMART+ root-of-trust assumptions

   assumed

   Each prover supplies immutable attestation code, protected keys and variables, controlled entry/exit, interrupt handling, and memory-protection rules sufficient to keep malware from invoking the key outside AttCode or modifying protected protocol state.

   SMART+ hybrid attestation basis and memory-access rules
3. assumption

   Timing and state assumptions

   assumed

   The verifier selects a swarm-size-dependent overall timeout, knows every device identity and expected measurement, and assumes only static or predictably changing attested memory; dynamic runtime integrity is not established.

   Swarm, connectivity, adversary, and verifier model · LISA-alpha protocol, state machines, QoSA, and complexity

Both protocols flood an authenticated fresh request, derive a parent relation, invoke protected single-device attestation, and route authenticated results back toward the verifier; they differ in whether devices merely forward reports or verify and aggregate them.

1. protocol LISA-alpha asynchronous forwarding specified

   A device authenticates a fresh request, records the sender as its parent, rebroadcasts the request, measures its memory, and sends an authenticated individual report upward. Intermediate devices forward current-session reports without authenticating them.

   LISA-alpha protocol, state machines, QoSA, and complexity

   1. output

      List-level outcome and optional topology

      specified

      The verifier partitions known identifiers into attested, failed, and no-report sets, giving L-QoSA. Parent fields can expose topology only after extra authentication; the basic parent value is not trustworthy enough for reliable F-QoSA.

      LISA-alpha protocol, state machines, QoSA, and complexity
2. protocol LISA-s synchronous aggregation specified

   Devices establish children during a bounded request phase, attest locally, authenticate child reports, aggregate descendant identifiers or results, and send one authenticated report to the parent. Verification before aggregation suppresses forged-report propagation.

   LISA-s aggregation protocol, state machines, QoSA, and complexity

   1. output

      Configurable aggregate QoSA

      specified

      Changing what each aggregate carries supports outcomes from B-QoSA through lists, counts, and a recursively encoded descendant tree for F-QoSA, trading information granularity against payload size and protected state.

      Binary, list, intermediate, and full Quality of Swarm Attestation · LISA-s aggregation protocol, state machines, QoSA, and complexity

The source argues protocol-level authenticity and freshness under its root-of-trust and MAC assumptions and reports distinct resource tradeoffs; these are not claims of physical security, lower-layer availability, or fully formal verification.

1. claim

   Resistance to report and request forgery

   informally analyzed

   Accepted forged reports or requests would require MAC forgery, disclosure of protected key K, or modification of protected state; the analysis relies on SMART+ isolation, freshness counters, and a secure MAC rather than a game-based proof.

   Attack vectors and informal security analysis
2. claim

   Best effort under connectivity change

   conditional claim

   If quasi-static connectivity fails, healthy devices may be omitted and produce a false-negative or no-report outcome, but the paper states that affected devices are not positively attested merely because links change.

   Swarm, connectivity, adversary, and verifier model
3. claim

   Communication-computation tradeoff

   analyzed and measured

   LISA-alpha imposes little extra attestation logic but can forward n reports per device in the worst case; LISA-s adds report verification, timing, child state, and aggregation while reducing transmitted reports and payload growth.

   LISA-alpha protocol, state machines, QoSA, and complexity · LISA-s aggregation protocol, state machines, QoSA, and complexity · CORE implementation, experimental setup, metrics, and results

Evidence consists of executable pseudocode and state machines, an informal attack-by-attack security analysis, asymptotic complexity discussion, and experiments with Python implementations in the CORE network emulator.

1. evidence

   Security-analysis coverage

   informal argument

   The analysis considers report forgery, request forgery, attestation-layer denial of service, and lower-layer denial of service separately for both protocols, tracing accepted forgeries to cryptographic or protected-state failures.

   Attack vectors and informal security analysis
2. evidence

   CORE emulation study

   reported experiment

   Random connected topologies in a 1,500 by 800 area use a 200-unit link threshold, 802.11, OLSR, and cryptographic delays derived from a laptop and Raspberry Pi 2. Each plotted point averages 30 generated scenarios while varying swarm and attested-memory size.

   CORE implementation, experimental setup, metrics, and results
3. evidence

   Reported performance pattern

   reported measurement

   Average device CPU time is similar because memory hashing dominates, whereas LISA-alpha bandwidth is higher; at 40 nodes the reported payload difference reaches about threefold, and bandwidth grows roughly linearly with swarm size.

   CORE implementation, experimental setup, metrics, and results

The results are conditional on protected SMART+ execution, an honest verifier, known membership, suitable timeouts, quasi-static connected topology, static-memory measurement, and absence of physical compromise; the experiments are emulations rather than a deployed hardware swarm.

1. limitation

   Shared-key physical-compromise boundary

   explicitly out of scope

   With one swarm-wide master key, physical compromise of one device exposes the key and permits impersonation of devices and verifier. Per-device symmetric keys or public-key designs reduce that blast radius but change computation, bandwidth, and neighbor-authentication costs.

   Cryptographic choices and physical-compromise tradeoffs
2. limitation

   Availability boundary

   explicitly out of scope

   Radio jamming, packet dropping, and other network-, link-, or physical-layer denial of service are not prevented; protocol timeouts bound waiting but cannot distinguish every adversarial omission from natural loss or mobility.

   Attack vectors and informal security analysis