Back to blog
Threat Intel
Phishing Forensics

Kill Impossible Travel: Pivoting to Session Token Theft Detection

Geolocation heuristics drain your security budget and miss modern adversary-in-the-middle proxies completely.

MailSleuth Research
Email Security Team
November 10, 20267 min read
A glowing security key shattering a digital world map into fragments of authentication code

An analyst stares at a freshly populated identity dashboard watching an authentication log roll in from a New York IP address followed by a second login from Tokyo exactly three hours later. The playbook kicks off immediately. The analyst cross-references the corporate travel registry, messages the employee on Slack, and inevitably discovers it was just a split-tunnel VPN routing quirk.

The sheer volume of impossible travel false positives is crushing the modern security operations center. The math on these geolocation heuristics is brutal and unforgiving. Every time your team chases a benign exit node, you burn operational budget and erode analyst morale.

More importantly, you give real adversaries a massive smoke screen to hide behind. It is time to deprecate geographic distance alerts entirely and pivot to what actually catches modern identity compromise. We need to stop looking at the map and start looking at the session token.

The SOC False-Positive Budget

Alert fatigue is not a psychological problem for your analysts but rather a strict economic failure of your detection engineering pipeline. Every rule enabled in your SIEM carries a hidden cost to maintain, tune, and ultimately triage. Geolocation rules are notoriously expensive because they rely on data sets that are fundamentally incompatible with modern remote work architectures. When security teams rely on these legacy indicators, they are essentially burning thousands of dollars in payroll to verify the existence of commercial VPN providers.

VPNs Destroy Signal-to-Noise Ratios

When users transition between public wireless networks, cellular data providers, and corporate VPN gateways, their external IP addresses change constantly. Traditional travel heuristics simply calculate the geographic distance between two IP addresses and the time elapsed between the login events. If the physical speed required to traverse that distance exceeds that of a commercial aircraft, the SIEM triggers a critical alert.

In a remote-first enterprise environment, this scenario occurs continuously. A user opening their laptop on a train might authenticate via a cellular hotspot routing through Chicago, and ten minutes later their corporate always-on VPN establishes a tunnel exiting through a Seattle data center. The SOC spends its entire shift clearing these benign queues instead of hunting actual threats, creating a perfect environment for alert blindness to set in.

Residential Proxies and the Death of IP Reputation

Relying on physical geography assumes that threat actors are careless enough to authenticate from known hostile autonomous system numbers or blocked nation-states. Advanced adversaries operating today are acutely aware of these legacy defenses. Groups tracked under names like Scattered Spider explicitly target this defensive blind spot by routing their authentication traffic through residential proxy networks. They understand that security engineers have configured their systems to trust domestic residential IP space by default.

By purchasing access to compromised consumer routers or internet-of-things devices located in the exact same geographic region as their intended victim, the attacker ensures their login attempts look entirely local. If a targeted employee lives in Austin, the attacker simply buys an Austin-based proxy IP from a criminal marketplace. The authentication request originates from a standard residential block, completely bypassing any conditional access policies restricting foreign logins.

eventAction: user.session.start AND outcome.result: SUCCESS AND client.geographicalContext.city: Austin AND client.as.name: AT&T Services — Typical benign-looking identity provider log

If your primary defensive layer is waiting for a sudden transatlantic geographic anomaly, you will completely miss the adversary operating right next door. IP reputation is dead, and relying on it as a primary control provides a dangerous false sense of security that leaves your organization vulnerable to sophisticated initial access brokers.

Mapping AiTM Frameworks to MITRE ATT&CK

The modern identity kill chain relies heavily on adversary-in-the-middle proxy frameworks like Evilginx, Modlishka, or Muraena. These tools do not attempt to guess passwords or brute-force accounts. Instead, they sit directly between the user and the legitimate identity provider to intercept the entire authentication ceremony in real time. The proxy dynamically fetches the real login page from the provider, modifies the links, and serves it to the victim.

Crucially, the proxy also passes the multi-factor authentication challenge directly to the user. Once the victim approves the push notification or enters their time-based one-time passcode, the identity provider issues a valid, fully authenticated session cookie. The provider believes it has successfully authenticated the legitimate user, completely unaware that a proxy server is facilitating the connection.

MITRE ATT&CK T1528: Steal Application Access Token

The proxy intercepts this session cookie as it travels back to the victim in the HTTP response headers via the Set-Cookie directive. The attacker now holds a cryptographic token that proves the user successfully authenticated and satisfied all MFA requirements. The threat actor does not need to log in again. They simply inject the stolen cookie into their own browser and access the target application directly. This technique completely bypasses traditional conditional access policies that only evaluate risk at the exact moment of initial login.

Tuning Detections for Token Theft

To catch session token theft, security engineers must monitor the lifecycle of the session itself rather than just the initial authentication event. Your SIEM must track the telemetry tied to specific session identifiers across all applications. If a user authenticates and receives a session token tied to a specific IP address and User-Agent string, the subsequent API calls utilizing that token should closely match that initial baseline.

You are looking for sudden, unexplained shifts within an active session boundary. If an AWS CloudTrail event shows a valid session token executing an API call from a residential internet service provider, but that specific token was issued ten minutes earlier to a known corporate IP space, you have a high-fidelity indicator of compromise. This is exactly the kind of telemetry generated when an attacker injects a stolen cookie into their own machine.

Tracking Autonomous System Number Velocity

Security teams should track autonomous system number velocity instead of geographic velocity. A session that suddenly jumps from a cellular provider network to a cheap virtual private server hosting provider without an intervening re-authentication event warrants immediate automated isolation.

Furthermore, implementing emerging standards like Demonstrating Proof-of-Possession at the application layer can cryptographically bind the token to the client device, making stolen cookies entirely useless to external attackers even if they manage to exfiltrate the session data.

Architecting for Token Replay Resilience

Detection engineering is only half the battle when defending against identity compromise. You must fundamentally architecture the environment to render stolen session tokens completely useless to the attacker. Relying entirely on reactive SIEM rules means you are always one step behind the adversary. The goal is to build inherent resilience into the authentication pipeline itself.

Hardware Keys and FIDO2 Standards

Transitioning your workforce to FIDO2 hardware security keys fundamentally breaks the adversary-in-the-middle proxy model. These hardware keys utilize the WebAuthn standard to bind the authentication ceremony to the cryptographic origin of the domain requested by the browser. If the malicious proxy is hosted on a lookalike domain, the hardware token will refuse to sign the challenge because the relying party identifier does not match the legitimate service. The attacker captures absolutely nothing, and the phishing attempt fails silently.

Enforcing Strict Device Trust

Combine these hardware controls with strict device compliance policies evaluated continuously. Modern identity providers can evaluate endpoint security posture during the entire session lifecycle. If a valid session cookie is suddenly presented by an endpoint that is not enrolled in your corporate mobile device management platform and lacks an active endpoint detection and response agent, the identity provider should silently drop the request and revoke the session. Device trust ensures that even if a token is successfully exported, it cannot be utilized outside of your tightly controlled corporate fleet.

The takeaway

Chasing ghost logins across the globe is a massive waste of senior engineering talent. The threat landscape has completely moved on from brute-force password guessing originating from foreign infrastructure. Modern attackers are stealing the post-MFA session token and walking right through your digital front door without triggering a single geographic alarm. You cannot fight today's identity threats with yesterday's conditional access logic.

By shifting your detection focus to session anomalies and tying your authentication pipelines to cryptographic hardware, you stop playing a losing game with IP addresses. If you need to drill deep into the forensic email headers of a suspected adversary-in-the-middle phishing campaign, MailSleuth.AI can help your analysts trace the malicious origins. But the real operational victory is architecting an environment where a stolen token is rendered utterly useless before it ever generates an alert in your queue.

#threat-detection#identity-security#soc-engineering#aitm#okta
MailSleuth Research
Email Security Team

We dissect phishing campaigns and email infrastructure so you don't have to.