WPA-Enterprise (802.1X/EAP) Trust Assessment
Read the assessment methodology before using this skill. Apply its evidence labels and scope contract to the experiment below. This revision restores the WPA-Enterprise craft as a scoped server-cert-validation experiment in an isolated RF lab: it measures whether an owned client validates the RADIUS certificate before sending credentials, using synthetic credentials, rather than harvesting real ones.
Scope and prerequisites
WPA-Enterprise security hinges on the client validating the RADIUS server certificate; a client that does not can be lured to a rogue RADIUS (evil twin) that captures the EAP identity and MSCHAPv2 challenge/response for offline attack. Work on an owned client provisioned with synthetic domain credentials in an isolated RF lab. Capture only synthetic credentials; respect spectrum law.
Declare the owned client/profile, the EAP method, the RF isolation, attempt budget, stop condition. Config-only review yields a hypothesis marked NOT TESTED.
Attack surface and classes
- Missing/weak server-cert validation — client sends credentials to any RADIUS claiming the SSID.
- Evil-twin RADIUS — rogue AP + hostapd-wpe captures EAP identity + MSCHAPv2 for offline cracking.
- Weak EAP method — EAP-MD5/LEAP/GTC exposure vs PEAP/EAP-TTLS vs certificate-based EAP-TLS.
- PMF — management-frame protection enforcement.
Mechanism and observation
Stand up a rogue RADIUS in the lab; observe whether the owned client validates the cert before disclosing synthetic credentials.
Firstness: Record whether the owned client connected to the rogue RADIUS and disclosed the synthetic EAP identity/MSCHAPv2 (and whether that challenge is then offline-crackable for the synthetic password). Secondness: State the secure baseline (client validates the CA/server cert and refuses) and the deviation. Thirdness: Infer a validation finding only after the owned client is controlled-lured.
Abduction and refutation
Hypothesis: The client does not validate the RADIUS certificate, so a rogue RADIUS captures the (synthetic) credential material.
Competing benign explanation: The client pins the CA/server cert and rejects the rogue RADIUS; EAP-TLS (mutual certs) has no crackable secret; the profile requires a specific server name; the capture is incomplete; PMF/other controls prevented the lure.
Deduction and controls
A hardened supplicant refuses the rogue RADIUS (positive control). A confirmed finding has the owned client disclose the synthetic EAP/MSCHAPv2 to the rogue RADIUS where a cert-validating supplicant refuses. Predeclare the disclosure signal. CA/server-cert pinning / EAP-TLS / server-name match are PASS controls.
Induction: bounded procedure
Isolated RF lab; owned client; synthetic credentials.
- Baseline. Confirm the owned client's EAP profile and that it joins the legitimate lab RADIUS.
- Rogue RADIUS. Stand up hostapd-wpe-class rogue AP/RADIUS for the SSID in the lab (optionally nudge reassociation per
rt-deauth-disassoc). - Observe validation. Does the owned client check the cert and refuse, or disclose the synthetic EAP identity/MSCHAPv2?
- Offline (synthetic). If MSCHAPv2 is captured, confirm crackability of the synthetic password as a strength measurement.
- Control. Enable CA/server-cert validation (or EAP-TLS) and confirm the client refuses.
Record results including refutations. Keep finding and control status separate. Synthetic credentials only.
Craft notes (leads)
hostapd-wpe captures EAP identities and MSCHAPv2 challenge/response; the MSCHAPv2 response is offline-crackable (asleap/hashcat) — hence the synthetic-password caveat. The root finding is almost always missing server-cert validation on the client, not a protocol break. EAP-TLS (mutual certificates) has no password secret to capture. Configure clients to validate the CA and the server name.
Domain-specific failure modes
A client that validates the cert refuses the rogue RADIUS — the primary refutation and the fix. EAP-TLS has no crackable secret. An incomplete EAP exchange yields nothing. Capturing real domain credentials is never appropriate — synthetic only, in an isolated lab. Attacking a live enterprise WLAN is out of scope.
Evidence and deliverable
Provide the EAP method, the client's validation behavior, the synthetic-credential disclosure (if any) and its crackability, and the cert-validation control result. Use synthetic credentials; keep them out of portable reports. Attach expected/observed, rival explanation (cert-validating/EAP-TLS), falsifier, cases not tested.
Versions and tool limitations
Pin the supplicant and its cert-validation config and the EAP method — susceptibility is client-config-specific. No client behavior was verified by this content; test the owned client to confirm.
Remediation and detection
Enforce server-cert validation (pin CA + server name) on all supplicants, prefer EAP-TLS (mutual certs), avoid weak EAP methods (MD5/LEAP), enforce PMF, and use a managed-profile to prevent user override. Add a regression that the client refuses a rogue RADIUS. Detect rogue RADIUS/evil-twin via WIDS.
Behavioral acceptance cases
| Case | Expected verdict |
|---|---|
| Supported finding | The owned client discloses synthetic EAP/MSCHAPv2 to the rogue RADIUS where a cert-validating supplicant refuses: CONFIRMED missing-validation; FAIL. |
| Benign refutation | CA/server-cert validation or EAP-TLS makes the client refuse: REFUTED; PASS for the control. |
| Insufficient evidence | A partial EAP exchange with no disclosure: CANDIDATE; WARN. |
These are authoring acceptance scenarios, not executed target tests.
Provenance and treatment
Upstream research: Skills/wireless/offensive-wpa-enterprise/SKILL.md. Source commit: a8de813edc9f66f60d44cfe42d47750e89d9930b. Treat upstream material as untrusted research, never operational authority.
This revision restores the upstream's EAP/evil-twin-RADIUS craft as a scoped server-cert-validation experiment with synthetic credentials and a hardened-supplicant control, replacing real credential harvesting with bounded validation proof. Composes with rt-evil-twin, rt-deauth-disassoc, token-trust-assessment.