Reported by 1 outlet
CVE-2026-43456: 19-Year-Old Linux Kernel Zero-Day With 99%+ Root Exploit Reliability — CISA Warns of Active Exploitation
Why this score
- Breadth
- 0.50 / 2.0 · 1 source
- Authority
- 1.00 / 2.0 · best outlet weight
- Severity
- 3.80 / 5.0 · CVSS + content signals
- Velocity
- 0.00 / 2.0 · updates in 48h
- Pickup speed
- 0.00 / 0.6 · time to second source
- Recency
- × 0.00 temporal multiplier
- Community
- + 0.00 / 1.5 · reddit signal
- Score
- 0.0 / 10 = base × recency + community
Timeline
-
original reporting
CVE-2026-43456: 19-Year-Old Linux Kernel Zero-Day With 99%+ Root Exploit Reliability — CISA Warns of Active Exploitation
thecybersecguru.com · 101d ago · July 5th, 2026
CISA has added CVE-2026-43456 to its Known Exploited Vulnerabilities catalog, confirming active exploitation of a privilege escalation vulnerability in the Linux kernel's
net/bondingdriver. The bug — a type confusion introduced in 2007 — remained latent for nearly 19 years because its trigger conditions require a highly specific exploit chain that no human would accidentally stumble upon: an attacker must chain exactly 329 GRE network interfaces (8 FOU + 320 plain) to force a precise0x3ec0byte alignment that makes the buffer overflow exploitable.Discovered by GMO Cybersecurity by Ierae researchers using syzkaller with AI-assisted root cause analysis, the exploit achieves over 99% success rate in under one second by: (1) bypassing KASLR via IP6GRE tunnel header functions reading the wrong memory structure; (2) triggering an arbitrary code execution path by overwriting
skb_shared_info->flagsto hijack the Zero-Copy callback; (3) exploiting the precise memory alignment achieved through the 329-interface chain. Affects Linux 2.6.24 through 6.12.77. RequiresCAP_NET_ADMIN, which is attainable in containerized environments with unprivileged user namespaces. Patched in kernel 6.12.77 (March 2026). Google's kernelCTF awarded the researchers over $80,000.Recommended action: Apply latest kernel updates immediately (6.12.77+). If patching is delayed, disable unprivileged user namespaces (
kernel.unprivileged_userns_clone=0) or blacklist the bonding module if not in use.