eBPF technology, long praised for legitimate Linux kernel observability and networking, is now being weaponized by underground actors to create rootkits that evade traditional detection.
Flare research uncovered Russian-language forum posts on Duty Free sharing full source code and tutorials for implants that hide processes via getdents64 hooks, activate via crafted TCP SYN packets, and encrypt C2 traffic with AES-256-GCM. The guide also shows how to use XDP to drop packets from 43 IP addresses linked to antivirus vendors.
Historical examples include BPFDoor used by Red Menshen, Symbiote’s dynamic linker preload technique, and the more recent LinkPro discovered by Synacktiv after a Jenkins compromise. Public research tools such as ebpfkit, TripleCross, and Boopkit demonstrated the same capabilities years earlier at Black Hat and DEF CON.
Additional exploits circulating include a reposted PoC for CVE-2017-16995 and an unverified local privilege-escalation targeting kTLS and eBPF sockmaps that attempts to corrupt /etc/passwd.
Defenders are advised to restrict unprivileged BPF access with kernel.unprivileged_bpf_disabled=1, regularly audit programs with bpftool prog list, watch for magic packets, and compare multiple host telemetry sources because some implants attempt to hide from bpftool itself.


