The anatomy of modern identity attacks: Defending public sector missions in an AI-driven threat landscape

Across federal, state, and local government agencies, mission delivery has fundamentally shifted. From delivering digital citizen services and modernizing legacy infrastructure to enabling hybrid public workforces, the era of static network perimeters and traditional castle-and-moat security models is over. Today, identity is the critical perimeter connecting agency personnel, mission partners, contractors, constituents, and automated machine workloads across multi-cloud and on-premises environments.

Consequently, adversaries targeting government infrastructure have shifted tactics: They no longer spend months attempting to exploit hardened software vulnerabilities; they simply log in with compromised credentials. According to the Verizon 2026 Data Breach Investigations Report, compromised credentials were the initial access vector in 22% of confirmed breaches. In the public sector, the stakes are magnified—every breach directly threatens mission continuity, critical infrastructure, and citizen trust.

Download: 5 Identity Attacks that Exploit Your Broken Authentication

Compounding this challenge is the rapid proliferation of generative AI tools and autonomous AI agents acting as non-human identities across government workflows. Here is an inside look at how modern identity attacks unfold across public-sector architectures and how agencies can implement a resilient Zero Trust defense.

How do attackers use AI to infiltrate government systems?

Historically, initial infiltration relied on generic phishing emails or manual password spraying. Today, adversaries deploy automated AI agents to execute hyper-targeted, context-aware campaigns against government employees and contractors.

Primary initial access vectors

• AI-generated spear phishing: By scraping public agency directories, government procurement portals, and public records, adversaries use AI to craft highly convincing spear-phishing campaigns that mimic agency leadership communications, inter-agency memos, or urgent compliance directives. 
• Adversary-in-the-middle proxies: Attackers deploy adversary-in-the-middle (AitM) reverse proxies that intercept credentials and bypass basic, legacy multi-factor authentication (MFA) methods—such as SMS codes and mobile push notifications—in real time.
• Credential stuffing using residential proxies: Simultaneously, threat actors weaponize massive credential repositories harvested using infostealer malware. By routing automated attacks through rotating residential proxy networks, adversaries execute credential-stuffing attacks across citizen-facing portals and agency single sign-on (SSO) gateways without tripping basic rate limits. 

While the 2025 Okta Secure Sign-in Trends Report shows workforce MFA adoption reaching 70%, nearly one-third of enterprise accounts remain unprotected by MFA, and basic factors remain vulnerable to MFA fatigue and AitM interception.

How are autonomous AI agents and machine identities exploited in cyberattacks?

As government agencies accelerate IT modernization and explore autonomous AI agents to triage constituent inquiries, analyze public data, and automate operational workflows, the identity attack surface has expanded beyond human operators.

According to Okta's Businesses at Work 2026 report, while 58% of executives identify AI governance as their foremost security concern, less than one-third secure AI agents with the same rigorous controls applied to human personnel.

In a modern public sector attack lifecycle, adversaries specifically target:

• Long-lived API keys, OAuth tokens, and machine certificates
• Elevated service accounts provisioned to automated workloads
• Under-governed AI agents granted broad read/write access across agency data repositories

When an AI agent or machine identity is manipulated through prompt injection or compromised service tokens, the malicious command executes with the full authorization of that service role at machine speed. What once took human adversaries days of manual lateral movement can now unfold across cloud environments in seconds.

What is session hijacking, and how does it bypass legacy MFA?

In response to federal Zero Trust mandates (such as OMB M-22-09 and CISA Zero Trust Maturity models), agencies have accelerated the adoption of stronger MFA. In turn, threat actors have adapted by avoiding the authentication gate entirely:

1. Token and cookie extraction: Instead of attacking the initial login, adversaries deploy infostealer malware or browser exploits to extract active session tokens and cookies directly from government-furnished equipment or contractor endpoints. 
2. Session replay: Replaying these stolen sessions enables attackers to bypass MFA controls and impersonate legitimate agency operators.
3. Lateral movement: Once inside, threat actors exploit fragmented identity architectures. A central identity provider (IdP) might enforce strict access controls at initial login. Still, if downstream SaaS applications, cross-agency shared services, or legacy infrastructure remain siloed, an attacker can move laterally across agency ecosystems without triggering security alerts.

The three types of insider threats in government

Identity threats in government do not always start with external threat actors. Insider risk remains a top vulnerability across public sector environments, categorized across three primary profiles:

• Negligent insiders: Well-meaning agency staff who inadvertently expose sensitive constituent records or Controlled Unclassified Information (CUI) by using ungoverned public AI tools or misconfiguring public cloud storage buckets.
• Compromised insiders: Legitimate public servants whose devices or credentials have been silently hijacked by infostealers, turning valid accounts into unwitting conduits for state-sponsored or cybercriminal intrusion.
• Malicious insiders: Disgruntled employees or contractors with standing administrative privileges who intentionally exfiltrate critical datasets or disrupt essential public services.

Without continuous risk scoring and centralized identity governance, traditional monitoring tools cannot distinguish between authorized mission activity and insider exploitation.

Modernizing identity-first security: The core pillars of a Zero Trust public sector defense framework

To protect critical missions, public safety, and citizen privacy against AI-velocity threats, government security architectures must move beyond basic perimeter and point-in-time checks. Modernizing defense requires four core architectural pillars:

1. Enforce phishing-resistant MFA by default

Transition agency personnel, contractors, and critical service accounts to phishing-resistant authentication mechanisms (such as FIDO2/WebAuthn, PIV card and CAC integration, and passwordless authentication). Phishing-resistant credentials cryptographically bind the authentication process to the verified agency domain, reducing AitM phishing and credential harvesting.

2. Implement comprehensive governance for non-human and AI identities

Extend identity governance and administration to service accounts, automated pipelines, and autonomous AI agents. Implement the principle of least privilege, reduce hardcoded and standing API keys, issue short-lived credentials, and establish mandatory human-in-the-loop approvals for sensitive, high-impact actions.

3. Establish continuous adaptive risk assessment and threat signal sharing

Implement continuous session evaluation before, during, and after authentication. Integrating shared threat intelligence across your security stack—including security information and event management (SIEM), extended detection and response (XDR), endpoint detection and response (EDR), and IdPs—enables automatic session revocation, step-up authentication, and privilege containment the instant suspicious endpoint behavior, impossible travel, or token theft is detected.

4. Unify IAM, IGA, and PAM

Break down identity silos by unifying identity and access management (IAM), identity governance and administration (IGA), and privileged access management (PAM). Replace standing administrative privileges with just-in-time access and automated lifecycle management to verify departing personnel and contractors lose access across all agency systems instantly.

Securing the mission with Okta for Government

Safeguarding public sector infrastructure requires an identity partner built for government-authorized security, scalability, and compliance—including FedRAMP High, GovRAMP, and DoD Impact Level 5 (IL5) authorizations. To learn how your agency can implement a unified, policy-based identity governance framework that strengthens security without slowing operations, download our e-book, How to govern access without slowing the mission.

Contact our public sector team to explore how your agency can implement a phishing-resistant Zero Trust identity architecture that helps protect public trust.

These materials are intended for general informational purposes only and are not intended to be legal, privacy, security, compliance, or business advice. © 2026 Okta, Inc. and/or its affiliates.