PaperlayerView exact revision on arXiv

Source-linked AI summary

Protocol Integration of Physical Layer Deception into EAP-TEAP Wi-Fi Authentication

Moustafa Ibrahim, Bin Han, Hans D. Schotten

arXiv:2610.01580v1cs.CRcs.NI

TL;DR

Credential-based EAP cannot distinguish a legitimate device from an adversary using compromised credentials, and prior PLD studies had not addressed integration into a running EAP stack. The paper adds batched PLD re-verification to EAP-TEAP and evaluates an end-to-end implementation, verifying construction behavior and analyzing recovery and security trade-offs.

  • Problem

    Credential-based EAP authenticates whoever presents valid credentials, leaving a claimant with compromised credentials indistinguishable at the credential layer; PLD integration into a running EAP stack had not been addressed, to the authors’ knowledge.

    Source: §I · p. 1

  • Method

    The paper adds a batched PLD re-verification step to EAP-TEAP, preserving primary-object opacity while sending recovery objects separately and requiring at least one active round.

    Source: §I · p. 1, C. Batched Construction and Protocol… · p. 3

  • Results

    The evaluation verifies the batched construction and derives informed-attacker and reliability results under stated independence assumptions.

    Source: §V · p. 6

  • Takeaways & Limitations

    The work realizes an authentication-oriented PLD step within EAP-TEAP/RADIUS/IEEE 802.11 in an end-to-end hostap 2.12 implementation.

    Source: §V · p. 6

    The repeated-observation analysis assumes channel-independent attempts, with retry decisions not conditioned on instantaneous channel state.

    Source: C. Repeated Observations and Retry… · p. 4

Abstract

from arXiv · show

Credential-based Extensible Authentication Protocol (EAP) authentication cannot distinguish a legitimate credential holder from an adversary using compromised credentials. Physical Layer Deception (PLD) complements credential-based authentication by exposing a deceptive primary object over a primary transport while a separate recovery object travels with differentiated reliability over a secondary channel. Existing PLD studies remain, to our knowledge, at the physical/link-model level; using PLD's activation/deactivation mechanism as an authentication gate creates an authentication-specific design requirement, since an all-inactive attempt would exercise no recovery path. We present a batched PLD-based re-verification step for Enterprise Wi-Fi's TEAP/RADIUS/IEEE 802.11 authentication chain, implemented end to end across the server, access point, and device in the open-source hostap 2.12 codebase. Each attempt carries three rounds, at least one active, with no dedicated activation flag. Across four campaigns totaling 1593 attempts, the prototype evaluates batched recovery behavior, rejects the implemented naive credential-bearing attacker in all 30 attempts, measures successful-path latency, and evaluates the security-reliability trade-off for one, two, and three active rounds under two modeled recovery regimes. The evaluation exercises the protocol and software-MAC behavior directly and analyzes informed and retry-seeking attackers under the software recovery model.

I. INTRODUCTION

Credential-based EAP accepts whoever presents valid credentials, so compromised credentials leave a legitimate device indistinguishable from an adversarial claimant. PLD offers a distinct recovery-path mechanism, but applying its activation mechanism as an authentication gate requires at least one active round.

I. INTRODUCTION

  • Compromised credentials let an adversarial claimant appear indistinguishable from the legitimate device at the credential layer alone.Enterprise WLANs commonly relay EAP through an access point to a RADIUS server, often using TEAP.Source: §I · p. 1
  • PLD pairs a deceptive primary object with a separate recovery object whose differentiated reliability favors a well-placed receiver.The paper distinguishes this active mechanism from channel-based authentication and challenge-response approaches.Source: §I · p. 1
  • Prior PLD studies had not addressed integrating activation and deactivation into a running EAP stack, to the authors’ knowledge.Using activation as an authentication gate requires at least one active round; an all-inactive attempt exposes every true challenge without exercising recovery.Source: §I · p. 1

A. PLD Mechanism

  • On active rounds, the server masks the true challenge and provides a key-bearing recovery object; inactive rounds carry the true challenge and non-key-bearing litter.A valid recovered key reconstructs the true challenge; otherwise the receiver falls back to the primary object, which is correct only for an inactive round or a successfully recovered active round.Source: A. PLD Mechanism · p. 2
  • The batch selects exactly A ≥1 active positions without replacement, sends no activation flag, and requires every round’s response to verify.An all-inactive batch would expose every true value directly and bypass the recovery path; randomized placement contributes uncertainty when active positions must be inferred.Source: A. PLD Mechanism · p. 2

B. Threat Model

The scheme targets a claimant who already has compromised client credentials and can complete ordinary TEAP authentication, distinguishing claimants through recovery-path reliability. Its protocol mapping keeps primary objects inside TEAP/TLS while sending recovery objects separately and gating success on the complete batch.

B. Threat Model

  • The target adversary holds compromised client credentials and can complete ordinary EAP-TEAP/EAP-TLS; the scheme adds post-credential re-verification before final EAP-Success.The claimant terminates its own TEAP/TLS session and observes the primary object, while the server and access point are trusted.Source: B. Threat Model · p. 2

C. Batched Construction and Protocol Mapping

  • The cross-layer mapping keeps primary objects opaque inside TEAP/TLS and forwards each recovery object separately over an IEEE 802.11 path, correlated per round without fixed arrival order.Authentication is gated on all L rounds verifying, using existing TEAP, RADIUS, and IEEE 802.11 extensibility.Source: C. Batched Construction and Protocol… · pp. 2–3
  • Each 32-byte recovery object encodes a candidate key and hash check; decoding yields VALID KEY on a match and NO VALID KEY otherwise.The server accepts only when all L rounds occur exactly once and every per-round response matches its independently recomputed value.Source: C. Batched Construction and Protocol… · pp. 2–3
  • Normal operation selects exactly 1 ≤ A ≤ L active rounds; A = 0 is reserved for an explicitly authorized diagnostic condition.Each round is selected as active or inactive by the server, and the scheme transmits no dedicated active/inactive flag.Source: C. Batched Construction and Protocol… · pp. 2–3
  • One TEAP Vendor-Specific TLV carries all primary values, while L RADIUS VSAs carry recovery objects that the access point forwards as L Category-127 Action frames.The proposed exchange is an auxiliary TEAP Phase-2 TLV exchange alongside Crypto-Binding and before the final Result TLV.Source: C. Batched Construction and Protocol… · p. 3

A. Receiver and Attacker Strategies

The paper contrasts a naive claimant that ignores recovery keys with informed attackers that use the receiver’s recovery rule or guess unrecovered active challenges. It also considers attackers seeking repeated observations of the same recovery object.

A. Receiver and Attacker Strategies

  • The naive non-deceptive attacker sets each reconstructed challenge to the primary object, regardless of its decoder outcome, and never uses an available key.The informed Perception attacker applies the receiver’s rule using its own recovery probability, without needing the active-set realization.Source: A. Receiver and Attacker Strategies · p. 3
  • The informed random-guess attacker knows L, A, and the challenge-space size but must guess both the position and value of every unrecovered active round.The Perception strategy follows the semantic-PLD strategy of.Source: A. Receiver and Attacker Strategies · p. 3
  • A claimant may withhold its response or exploit retries to collect additional independent observations of the same recovery object; channel-state-aware attempt selection is outside the analysis.The modeled retry decision does not condition on instantaneous channel state.Source: A. Receiver and Attacker Strategies · p. 3

B. False Acceptance and PLD-Induced False Rejection

The section distinguishes an informed Perception attacker from a random-guess baseline and relates false acceptance to legitimate PLD-induced rejection.

B. False Acceptance and PLD-Induced False Rejection

  • The random-guess baseline assumes 128-bit challenges and no side information distinguishing inactive litter from missed active objects.If side information identifies inactive positions, the combinatorial factor decreases toward one.Source: B. False Acceptance and PLD-Induced… · pp. 3–4
  • The random-guess false-accept probability combines recovery outcomes with conditional guessing and assumes i.i.d. recovery.The no-side-information assumption matters because identifying inactive positions reduces the combinatorial factor.Source: B. False Acceptance and PLD-Induced… · p. 4
  • Under the informed Perception strategy, false acceptance occurs exactly when all active-round keys are recovered.The passage describes this as the condition T = A.Source: B. False Acceptance and PLD-Induced… · pp. 3–4
  • PLD-induced false rejection is defined symmetrically using the legitimate receiver’s recovery probability.The passage identifies this probability as rB.Source: B. False Acceptance and PLD-Induced… · p. 4

P PLD

Under i.i.d. per-round recovery, the section presents a Perception false-acceptance and PLD-induced false-rejection trade-off, conditioned on successful primary delivery.

P PLD

  • The stated trade-off is conditioned on successful primary/TEAP delivery.Source: P PLD · p. 4
  • As A increases, Perception false acceptance falls while PLD-induced false rejection rises.The trade-off is stated for the closed forms under i.i.d. per-round recovery.Source: P PLD · p. 4
  • The closed forms assume i.i.d. per-round recovery; correlated or position-dependent recovery uses the general probability Pr(T = A).Source: P PLD · p. 4

C. Repeated Observations and Retry Policy

Repeated observations can improve recovery for both legitimate receivers and attackers, so the section limits same-object retries and proposes controls on fresh-session retries.

C. Repeated Observations and Retry Policy

IV. PROTOTYPE AND EVALUATION

The prototype adds server, access-point, and device changes to hostap 2.12 and exercises the batched protocol through the software MAC-level simulator.

A. Implementation

  • The server executes each attempt as one three-round batch and withholds EAP-Success unless every round appears exactly once and every response matches.Malformed, missing, duplicate, or incorrect batch material is rejected.Source: A. Implementation · p. 4
  • The access point relays TEAP without terminating or decrypting it, forwarding each recovery object in its own Action frame.It forwards active codewords and litter identically; validity checking occurs at the device.Source: A. Implementation · p. 4
  • The device correlates round objects by transaction and round identifiers regardless of arrival order, waits up to 2000 ms, then returns all response tags together.An early-arriving recovery object is cached pending the batch request.Source: A. Implementation · p. 4
  • Fig. 3 covers the security–reliability trade-off versus active rounds; Table I covers a 273-attempt batched recovery-model campaign.Source: Figure 3 · p. 5, Table I · p. 5

B. Experimental Setup

The evaluation uses four mac80211_hwsim campaigns totaling 1,593 attempts, with L = 3 in every campaign.

B. Experimental Setup

  • The four campaigns cover recovery behavior (273 attempts), the naive attacker (30), latency (90), and security-reliability trade-offs (1,200).The naive-attacker campaign uses an isolated test-only device binary; the server and access point remain unchanged.Source: B. Experimental Setup · p. 5
  • Each campaign uses a pre-generated fixed schedule, and controlled per-round withholding realizes degraded and trade-off recovery probabilities.The trade-off campaign tests A ∈ {1, 2, 3} under two modeled recovery regimes, with 100 attempts per condition for both Bob and Perception-Eve.Source: B. Experimental Setup · p. 5
  • A shared 2,000-ms deadline resolves rounds still missing at expiry as NO VALID KEY; reported latency covers successful paths, while reject timing is qualitative.The latency campaign compares ordinary TEAP with A ∈ {1, 3} full-recovery cases, using 30 attempts per condition.Source: B. Experimental Setup · p. 5

C. Batched Recovery-Model Evaluation

The recovery-model, security-reliability, and attacker campaigns report complete decisions, the modeled trade-off across active-round counts, and rejection of the implemented naive attacker in its tested configuration.

C. Batched Recovery-Model Evaluation

  • 273/273 recovery-model attempts completed without infrastructure failures or unexplained timeouts; ordinary TEAP and both full-recovery cases accepted 30/30.Under degraded recovery, acceptance fell as recovery probability decreased.Source: C. Batched Recovery-Model Evaluation · p. 5
  • A = 0 accepted 3/3 attempts despite all recovery objects being withheld; fixed and randomized A = 1 counts matched at both tested recovery probabilities.The A = 0 result motivates requiring A ≥ 1 in normal operation; the matching A = 1 counts are expected under i.i.d. recovery.Source: C. Batched Recovery-Model Evaluation · p. 5

D. Security-Reliability Trade-off Evaluation

  • Across both modeled recovery regimes, increasing A lowers analytical PFA and raises analytical PFR; all 1,200 attempts matched decisions predicted from realized active sets and recovery delivery.Every trade-off attempt reached an accept/reject decision without infrastructure failures or unexplained timeouts.Source: D. Security-Reliability Trade-off… · p. 5

E. Naive Credential-Bearing Attacker Baseline

  • All 30 attempts against the naive credential-bearing attacker were rejected, with zero unexpected accepts or infrastructure failures.This result covers one strategy at L = 3, A = 1 with randomized active positions; the semantic-PLD Perception strategy is analyzed separately.Source: E. Naive Credential-Bearing Attacker… · pp. 5–6

F. Authentication Latency

In the software-MAC testbed, re-verification added approximately 25 ms on successful paths, while missing required active recovery objects followed a much longer reject path. The paper also reports per-hop overheads and identifies physical-layer and attacker-model boundaries.

F. Authentication Latency

  • Mean successful-path latency was 231.698 ms for ordinary TEAP, 257.096 ms for L = 3, A = 1, and 256.558 ms for L = 3, A = 3.The reported increase over ordinary TEAP is approximately 25 ms for both full-recovery configurations.Source: F. Authentication Latency · p. 6
  • Attempts missing a required active recovery object waited for the 2,000-ms recovery deadline and rejected at approximately 2.4–2.5 s; reject timing was reported qualitatively.Source: F. Authentication Latency · p. 6

G. Protocol Overhead

  • At L = 3, overheads were 73 B/121 B for TEAP Batch Request/Response TLVs, 150 B for RADIUS recovery VSAs, and 147 B for Category-127 Action-frame bodies.The three quantities occur on distinct transport hops and are reported separately.Source: G. Protocol Overhead · p. 6

V. DISCUSSION AND CONCLUSION

  • Campaigns verify the batched construction, A ≥ 1, randomized positions, litter, and all-round gate; attacker and repeated-observation results are derived under independence assumptions.The proposed re-verification is implemented end-to-end in hostap 2.12 and evaluated alongside analytical attacker and reliability models.Source: §V · p. 6
  • The analysis models per-round recovery independently of channel state; physical-layer recovery asymmetry, real-hardware performance, and broader adaptive attackers remain open evaluation areas.The paper also identifies correlated fading, missed-active/litter reception, relay or collusion resistance, and recovery-frame origin authentication as open issues.Source: §V · p. 6
Loading 2610.01580v1…