Multiple DoS Vulnerabilities in Quicly Affecting HTTP/3 Servers

🔓 VULNERABILITIES & CVEs

• Quicly Connection State Corruption — CVE-2026-44436
  Multiple denial-of-service vulnerabilities identified in Quicly (IETF QUIC protocol implementation used in H2O HTTP server). Prior to commit 8b178e6, attackers can cause connection state corruption leading to service disruption.

• Quicly Remote DoS via Assertion Failure — CVE-2026-44435
  Assertion failure triggered by malformed CRYPTO stream handshake frames in Quicly versions before commit 937d0e9 causes remote denial of service.

• Quicly Stateless Reset Injection — CVE-2026-44434
  Stateless reset injection vulnerability due to lack of packet validation in Quicly prior to commit dccf5d4 allows attackers to disrupt connections.

• Quicly Memory Exhaustion — CVE-2026-44433
  Memory exhaustion denial of service possible by sending minimal STREAM frames in Quicly versions before commit 8b178e6.

• Bouncy Castle ARM SHA3/SHAKE Out-of-Bounds Write — CVE-2026-15997
  Out-of-bounds write in Legion of the Bouncy Castle bcprov-lts8on on ARM platforms due to restoreFullState underflow detection failure.

• wger IDOR Allows Access to Private Workout Data — CVE-2026-43977
  Authenticated users in wger fitness manager (pre-2.6) can read other users’ private workout notes and history.

• wger Privilege Escalation via Session Chaining — CVE-2026-43978
  Gym trainers can escalate privileges to manager roles by chaining trainer-login sessions in wger versions before 2.6.

• Jupyter Enterprise Gateway Kubernetes Manifest Injection — CVE-2026-44182
  Pre-3.3.0 versions interpolate untrusted input in Jinja2 templates, enabling Kubernetes manifest injection and potential remote code execution.

• Jupyter Enterprise Gateway SSTI via Jinja2 Templates — CVE-2026-44181
  Server-side template injection vulnerability in Jupyter Enterprise Gateway (2.0.0rc2 to pre-3.3.0) allows remote code execution.

• Jupyter Enterprise Gateway ContainerProcessProxy Privilege Bypass — CVE-2026-44180
  Improper enforcement of prohibited user IDs in Jupyter Enterprise Gateway versions before 3.3.0 can lead to privilege bypass.

• Microsoft SharePoint Server Spoofing (XSS) — CVE-2026-62826
  Authorized attackers can exploit input neutralization flaws during web page generation to perform spoofing attacks over the network.

• Windows Backup Service Privilege Escalation (Race Condition) — CVE-2026-58598
  Local privilege escalation via race condition in Windows Backup Engine due to improper synchronization of shared resources.

• Windows Admin Center Cross-Site Scripting (XSS) — CVE-2026-58643
  Unauthenticated attackers can perform network spoofing via input neutralization flaws in Windows Admin Center web pages.

• Windows Terminal Remote Code Execution — CVE-2026-59117
  Integer overflow vulnerability in Windows Terminal allows remote code execution by unauthenticated attackers.

• Kirby CMS XSS via KirbyTags and Image Blocks — CVE-2026-45368
  Versions prior to 4.9.1 and 5.4.1 fail to sanitize URLs in KirbyTags and image blocks, enabling stored cross-site scripting attacks.

• Kirby CMS Content Lock Info Disclosure — CVE-2026-45334
  Content-locking feature leaks IDs and emails of inaccessible users due to missing access checks in Kirby CMS before 4.9.1 and 5.4.1.

• Kirby CMS XSS via List Field Content — CVE-2026-44175
  Unsanitized list field content in Kirby CMS allows stored cross-site scripting in versions before 4.9.1 and 5.4.1.

• YAML::Syck Perl Out-of-Bounds Read — CVE-2026-57077
  Versions before 1.47 allow out-of-bounds read via unbounded newline scan in bundled libsyck, risking memory disclosure.

• YAML::Syck Perl Heap Use-After-Free — CVE-2026-57076
  Heap use-after-free triggered by anchor name reuse in YAML::Syck versions prior to 1.47, leading to potential memory corruption.

• YAML::Syck Perl Out-of-Bounds Read via Base64 Decoder — CVE-2026-57075
  Signed-char lookup table index flaw in base64 decoder causes out-of-bounds read in YAML::Syck before 1.47.

🕵️ THREAT RESEARCH & DEEP DIVES

• Agentic AI Is Untamable: Ask the Right Security Questions
  Emerging agentic AI systems introduce novel security risks that go beyond traditional attacker models, requiring a fundamental rethinking of organizational security strategies to address autonomous AI-driven threats.