Affected Systems

Microsoft Edge versions prior to 148.0.3967.70 (released May 15, 2024). The vulnerability chain exploited conditional access navigation throttle, Immersive Reader, and protocol handler configuration flaws to achieve arbitrary code execution outside the browser sandbox without memory corruption.

Exploitation Status

Demonstrated at Pwn2Own Berlin May 2024 as a competition exploit. No evidence of exploitation in the wild. Microsoft CVE records list exploitation status as none. Patched within 24 hours of contest disclosure.

Business Impact

This was a novel logic-only sandbox escape requiring no memory corruption, demonstrating a new attack surface in Edge. However, Microsoft patched the issues (CVE-2024-45492, CVE-2024-45494, CVE-2024-45495) on May 15, 2024, the day after Pwn2Own. Browsers on automatic updates received the fix months ago. The chain required user interaction (visiting a malicious page while signed into Microsoft services) and was complex to develop. No real-world exploitation has been observed. CVSS scores for individual bugs were 5.4 (medium), though the full chain achieved code execution.

Urgency

🔵 Monitor

Recommended Actions

  • Verify all Microsoft Edge installations are running version 148.0.3967.70 or later (released May 15, 2024)
  • Confirm Edge automatic updates are enabled via edge://settings/help to ensure future patches deploy promptly
  • Review endpoint detection logs from May 2024 for anomalous Edge process behavior, unexpected file writes to protocol handler configuration paths, or telnet:// protocol invocations on systems without legitimate telnet use
  • For high-security environments, consider application control policies that restrict execution from user-writable directories to mitigate similar file-write-to-execution techniques
  • Monitor Microsoft Security Response Center advisories for CVE-2024-45492, CVE-2024-45494, and CVE-2024-45495 for any updates or related issues

---

# Threat Actor Context

Actor Profile

Orange Tsai is a security researcher affiliated with DEVCORE, a Taiwan-based security research firm. Tsai is known for discovering high-impact vulnerability chains in widely-used software and has won multiple awards at Pwn2Own competitions. His motivation is legitimate security research aimed at identifying and responsibly disclosing critical vulnerabilities to vendors. The disclosed Edge exploit chain won the 2026 Pwnie Award for Epic Achievement and demonstrates advanced manual research capabilities without reliance on automated tools or AI assistance.

TTPs (Tactics, Techniques, Procedures)

The exploit chain leverages logic flaws rather than memory corruption vulnerabilities. Key techniques include: abuse of Edge's conditional access navigation throttle feature to achieve universal cross-site scripting (UXSS) by exploiting tab attachment mismatches; bypass of origin checks through secondary window manipulation; exploitation of hardcoded allowed address lists to circumvent popup blockers; abuse of Edge's Immersive Reader (read:// protocol) which was excluded from script execution restrictions; file write primitive via misuse of Edge's diagnostic log compression function; configuration file overwrite to manipulate protocol handler permissions; and arbitrary code execution via DLL search order hijacking through telnet:// protocol handler abuse with attacker-controlled executables placed in user-writable directories. The entire exploit chain comprises under 100 lines of code and represents a novel attack surface targeting browser logic rather than memory safety.

Targets & Patterns

The vulnerability chain specifically targeted Microsoft Edge browser users on Windows systems. While demonstrated at Pwn2Own Berlin as a competition exploit, the technique could theoretically affect any Edge user prior to version 148.0.3967.70. The technology sector is implicitly targeted given Edge's widespread deployment in enterprise environments. The attack required the victim to be signed into a Microsoft account, which Edge's auto-sign-in feature facilitates automatically when visiting Microsoft properties. There is no evidence of exploitation in the wild; CVE records indicate exploitation status as none. The research was conducted as responsible disclosure rather than malicious targeting, with Microsoft patching within 24 hours of contest disclosure.

Historical Context

This research represents Orange Tsai's continued success at Pwn2Own competitions, with this particular chain winning $175,000 at Pwn2Own Berlin in May 2024. The work earned the 2026 Pwnie Award for Epic Achievement, recognizing its significance in the security research community. Tsai's approach of chaining pure logic bugs without memory corruption represents a departure from typical browser sandbox escape techniques, which historically rely on memory safety vulnerabilities. The researcher developed the chain manually over approximately two months without AI assistance, emphasizing the continued importance of human-driven security research. DEVCORE, Tsai's affiliated organization, has a track record of high-impact vulnerability research and responsible disclosure practices.

Defensive Recommendations

  • Ensure Microsoft Edge is updated to version 148.0.3967.70 or later to remediate CVE-2026-45492, CVE-2026-45494, and CVE-2026-45495; automatic updates should have applied this patch since May 2024
  • Monitor for suspicious file writes to user-writable directories and unexpected modifications to protocol handler configuration files, particularly those controlling application associations
  • Implement application whitelisting or restrict execution from user-writable paths in the system PATH to prevent DLL search order hijacking and arbitrary code execution via protocol handlers
  • Review and harden cross-origin resource sharing (CORS) policies on internal web properties to prevent information disclosure such as email addresses or session data to untrusted origins
  • Audit browser extension and feature configurations to identify hardcoded allowlists or exceptions that may bypass security controls like popup blockers or content security policies