How Auth Sentry Works

Traditional security tools generate thousands of alerts. Your team investigates manually. Attackers move faster than humans can respond. Auth Sentry changes that.

The Problem with Traditional ITDR

What You Get:

"Unusual login detected from new location"

Severity: Medium
Recommended Action: Investigate

What You Have to Do:

  • • Manually query SIEM for related events
  • • Check SaaS platform logs
  • • Pull user's access history
  • • Contact user to validate
  • • Correlate with threat intel
  • • Determine if legitimate or malicious
  • ⏱ Time spent: 45 minutes

While you investigate, the attacker moves laterally.

The Auth Sentry Difference: Autonomous Investigation

1

Evidence Collection via Graph Database

Our AI Agents don't just flag anomalies—they continuously link evidence across your identity graph until patterns emerge. Only when sufficient evidence exists does an alert get created.

Example: Lateral Movement Detection

When user "james.hernandez14" accesses 4 systems in 5 minutes, the agent:

  1. Queries the graph for all identity events related to james.hernandez14
  2. Discovers account was dormant for 6 months (evidence #1)
  3. Correlates access pattern with lateral movement signatures (evidence #2)
  4. Links to previous suspicious OAuth token activity (evidence #3)
  5. Calculates confidence score based on evidence weight: 0.98 (98%)
  6. Creates alert only when threshold reached
Why Graph Databases Matter

Traditional tools analyze events in isolation. Graph databases connect the dots—linking users, permissions, apps, OAuth tokens, and actions across time. This reveals attack chains that single-point analysis misses.

2

Automatic Enrichment from Your Security Stack

Instead of handing you a generic "suspicious activity" alert, AI Agents autonomously gather context from your existing tools before creating the alert.

SIEM Query

Pulls related security events from same timeframe

SaaS Platform Logs

Retrieves authentication and access logs

MDM Systems

Checks device compliance status

Threat Intel

Correlates with known attack patterns

Result: Analysts get enriched context automatically, not raw alerts requiring manual investigation.

3

Human-in-the-Loop Validation

AI Agents can message affected users directly via Slack or Microsoft Teams to validate suspicious activity in real-time.

"Hey James, did you just access the production database from a new location in Singapore?"

Reply 'yes' to approve, or we'll revoke access in 2 minutes.

Possible Outcomes:

  • User confirms: "Yes, I'm traveling" → Alert closed as false positive
  • User denies: "No, my device was stolen!" → Escalate to critical
  • No response in 2 minutes → Auto-revoke tokens, block access

Result: Faster validation, fewer false positives, immediate containment when threats confirmed.

What You Receive: Not an Alert, a Complete Investigation

ALERT ID: 528
SEVERITY: HIGH

Suspicious Lateral Movement Detected - james.hernandez14

Identity [email protected] accessed 4 different systems within 5 minutes, indicating potential lateral movement.

EVIDENCE COLLECTED BY AI AGENT:
  • [Graph Database] Dormant account (no activity for 6 months) suddenly active
  • [Pattern Match] Access pattern matches known lateral movement signatures
  • [Auto-enriched from SIEM] Failed authentication attempts from same IP 30 min prior
  • [Auto-enriched from Okta] MFA push accepted after 3 denials (MFA fatigue attack)
  • [User validation via Slack] User reports device stolen 2 hours ago
THREAT VECTOR
Lateral Movement
CONFIDENCE SCORE
0.98 (98%)
INVESTIGATION TIME
<2 minutes
AI AGENT ACTIONS ALREADY TAKEN:
✓ Gathered evidence from 3 systems automatically
✓ Contacted user via Slack (confirmed compromise)
✓ Auto-revoked all OAuth tokens for this identity
✓ Blocked source IP address
⏳ Awaiting SOC approval to fully disable account
RECOMMENDED NEXT STEPS:
1. Review access logs for other compromised accounts
2. Force password reset for james.hernandez14
3. Audit all systems accessed in past 24 hours for data exfiltration
4. Consider triggering incident response playbook IR-001

Your SOC gets cases, not noise. 10x faster investigations.

Specialized AI Agents for Specialized Threats

One generic ML model misses nuanced attacks. Auth Sentry deploys dedicated agents, each expert in specific threat domains.

OAuth Token Agent

Monitors OAuth and refresh token lifecycles, detecting:

  • Token usage from impossible locations (impossible travel)
  • Cross-IP token sharing patterns
  • Tokens used outside normal application workflows
  • Dormant tokens suddenly activating
  • Token lifespans exceeding security policies

Service Account Agent

Learns "normal" behavior for non-human identities, flagging:

  • API key rotations outside change windows
  • Privilege escalations and permission changes
  • Dormant service accounts suddenly active
  • Access to resources outside typical patterns
  • Service accounts used from unexpected IP ranges

Lateral Movement Agent

Maps access chains across systems, identifying:

  • Unusual system-to-system pivots
  • Access to multiple high-value targets in short timeframes
  • Credential usage across disparate environments
  • Movement from dev → staging → production
  • Access path deviations from established patterns

Toxic Combo Agent

Detects dangerous application permission combinations:

  • GitHub access → AWS credential retrieval → production database
  • Email + file storage + admin console in sequence
  • CI/CD pipeline access combined with prod environment permissions
  • Third-party app integrations with over-privileged access
  • Chained access that individually appears legitimate but together forms attack path

Why specialized agents? Each agent operates independently, continuously learning YOUR organization's specific patterns. When threats intersect multiple domains (e.g., OAuth token abuse + lateral movement), agents collaborate to build the complete picture before alerting.

Identity Rx: Continuous Learning from YOUR Environment

Off-the-shelf AI doesn't understand YOUR business. Our AI Agents continuously learn your organization's unique identity DNA.

DevOps deploys on Friday afternoons?

Agent learns this is normal for YOUR team.

Sales accesses CRM + LinkedIn + ZoomInfo in sequence?

Agent baselines the pattern specific to YOUR workflow.

Finance runs monthly scripts touching prod databases?

Agent expects it based on YOUR calendar.

The longer it runs, the smarter it gets about YOUR specific environment.

70% reduction in false positives. Detections tailored to your organization, not generic rules.

See Auth Sentry in Action

Watch our AI Agents investigate a real lateral movement attack in under 2 minutes.

Request Free Trial