Entra ID Conditional Access — the 14-day rollout pattern

Pawan Sharma Published 06 Jun 2026  ·  By Pawan Sharma  ·  Identity  ·  14 min read

Entra ID Conditional Access is the single most important security policy engine in a Microsoft-anchored estate — more impactful than any single Defender plan, and the first thing we tighten in every engagement. This post is the 14-day rollout pattern: emergency-access setup, baseline policy library, phishing-resistant MFA for admins, Sign-in Risk-Based enforcement, Token Protection. Plus the seven CA mistakes we audit-find most often.

Entra P1 / P2

Licensing

P1 = Conditional Access. P2 = Sign-in Risk + ID Protection + PIM.

14 days

Standard rollout

Break-glass → baseline → enforce → phishing-resistant admins.

99.9%

Microsoft published

Sign-in attack risk reduction from MFA. CA enforces it at the right places.

Phishing-resistant

FIDO2 / WHfB / passkeys

Required for admins + crown-jewel apps. CA enforces strength.

The baseline policy library — 8 policies every tenant needs

PolicyScopeAction
1. Require MFA for all usersAll users (except break-glass + service identities)Require MFA
2. Require phishing-resistant MFA for adminsAll directory + privileged rolesRequire authentication strength: phishing-resistant
3. Block legacy authenticationAll users (except break-glass)Block (POP/IMAP/SMTP auth + other legacy protocols)
4. Require compliant + hybrid-joined deviceAll users on M365 + Defender XDR appsRequire device compliance + hybrid-join
5. Block from non-corporate device / networkCrown-jewel apps (ERP, HRMS, finance)Block from non-compliant + non-network
6. High Sign-in Risk → blockAll users (P2 required)Block
7. Medium Sign-in Risk → require MFA + password changeAll users (P2 required)Require MFA, then password change
8. B2B guest MFA + session limitsAll guestsRequire MFA + session 8-hour sign-in frequency

The 14-day rollout pattern

1

Days 1-3 — Emergency access (break-glass) accounts

Create 2 cloud-only emergency accounts; permanent Global Administrator role; complex passwords stored in sealed safe; excluded from ALL Conditional Access policies; Sentinel alert on any sign-in. Quarterly fire-drill test.

2

Days 4-7 — Baseline policies in report-only

Policies 1, 3, 4, 8 in report-only mode. Watch sign-in logs for who would have been blocked. Identify legacy auth holdouts (typically: scanner / printer / on-prem mailbox client) + remediate or carve specific exemptions.

3

Days 8-10 — Flip to enforce

Baseline policies enforced. Monitor support tickets for first 48 hours; the legacy-auth ticket spike is the signal you missed an exemption. Reduce policy scope if needed; never reduce by disabling the whole policy.

4

Days 11-14 — Phishing-resistant + Sign-in Risk + Token Protection

Policy 2 (phishing-resistant MFA for admins) — issue FIDO2 keys / enable Windows Hello for Business. Policies 6 + 7 (P2 Risk-Based). Token Protection on M365 apps. PIM enrolment for all directory + privileged roles.

The 7 CA mistakes we audit-find most often

1. No break-glass accounts

Or break-glass excluded only from MFA policy but included in others — single CA policy change can lock out the tenant.

2. MFA covers users but not admins

Admin role MFA exclusion for "convenience" — the exact opposite of zero-trust.

3. Legacy authentication open

Even with modern auth required, POP/IMAP/SMTP basic auth still allowed on at least one mailbox — and that's the one attackers find.

4. Service principals + workload identities unmonitored

CA doesn't apply to service-principal sign-ins by default. Workload Identity Premium add-on closes this gap.

5. Guests inherit member privileges

Cross-Tenant Access Settings left default. Guests bypass MFA because their home tenant claimed MFA. Trust ours, not theirs.

6. No Risk-Based policies despite owning P2

Sign-in Risk + User Risk policies licensed but disabled. The single highest-signal trigger left dormant.

7. Token Protection not enabled where supported

Token-theft replay attacks are the 2025-2026 frontier; Token Protection is the defence.

PIM + Privileged Access Workstation — the next layer

Once CA baseline is in

Privileged Identity Management + PAW for admin roles

PIM converts standing-privileged roles to just-in-time activation with MFA challenge + justification + approval workflow + auditable trail. Combine with Privileged Access Workstations (PAW) — dedicated, locked-down devices for admin work, kept separate from email + web browsing endpoints. The combination is what mature Microsoft-anchored estates run at the privileged tier. Ogma rolls these out as the layer-2 after CA baseline.

FAQ

Is Entra ID P2 mandatory for Conditional Access?
Conditional Access itself is P1. The Risk-Based policies — sign-in risk, user risk, ID Protection — are P2. For mature posture you want P2 + Defender XDR integration; P1 alone misses the highest-signal triggers.
Where does Conditional Access fit relative to MFA?
MFA is the action; Conditional Access is the policy engine deciding when MFA is required, when access is blocked, when session is restricted. Modern posture is 'CA-driven MFA only when warranted' rather than 'MFA on everything always' — better security + better user experience.
Phishing-resistant authentication — FIDO2 / passkeys / WHfB — how does it fit in?
Conditional Access can require 'phishing-resistant MFA' as the strength. Map admin roles + crown-jewel apps to phishing-resistant only (FIDO2 / Windows Hello for Business / passkeys / cert-based auth). User population to standard MFA. The strength taxonomy is the right primitive.
What about contractors / vendors / B2B guests?
Cross-Tenant Access Settings (CTAS) + Conditional Access policies scoped to guests. Common pattern: require MFA on guest sign-in even if their home tenant has MFA — 'don't trust their MFA, trust ours'. Plus session-time-bound access via Entitlement Management.
Break-glass accounts — how do we exclude safely?
Two break-glass accounts excluded from all CA policies, stored in a sealed safe, monitored continuously via Sentinel for sign-in events. Quarterly fire-drill access verification. CA policy compromise without break-glass = locked-out tenant for hours.
Token theft + replay — how do we defend?
Token Protection policy (preview-to-GA in 2025) binds tokens to device. Continuous Access Evaluation (CAE) revokes tokens within minutes on risk signal. Pair both: device-bound tokens that revoke on risk. Substantially harder for token-replay attacks.
How does CA tie back to DPDP / RBI / CERT-In compliance?
RBI Cyber Security Framework + SEBI CSCRF + IRDAI ICS all expect strong authentication on privileged access. CA policy library + Sign-in Risk + Privileged Identity Management (PIM) are how you operationalise + evidence that. Each CA policy has an audit trail Sentinel can ingest.
What's a good 14-day Conditional Access rollout?
Days 1-3: emergency-access (break-glass) accounts set up + tested. Days 4-7: baseline policies in report-only mode (require MFA all users, block legacy auth, block from non-compliant device). Days 8-10: flip to enforce after report-only review. Days 11-14: phishing-resistant MFA for admins, Sign-in Risk-Based policies, Token Protection.

Free Entra Conditional Access audit + remediation plan

Identify your CA policy gaps, missing risk-based controls, and break-glass posture in 5 working days

Ogma audits your Entra CA policy library against the 8-policy baseline + 7-mistake checklist, returns a prioritised remediation plan, and runs the 14-day rollout if you choose to proceed. INR + GST quote tied to your seat count.

Request the CA audit or explore the Entra ID landing

Related: Defender XDR vs CrowdStrike · E5 Security bundle math · 5 mistakes Indian buyers make

Stay ahead of cyber threats

One short email a week — curated Indian cybersecurity news, Fortinet releases, DPDPA updates. No fluff.


Cato Firewall as a Service
Cato ZTNA — Zero Trust Network Access
Cato SASE Solution