A 2002-era hijack technique against firewalls that have been EOL longer than most of your analysts have been alive
Tenable plugin 11057 flags hosts whose TCP Initial Sequence Numbers are predictable enough to enable blind connection spoofing — the classic Mitnick-style hijack. The primary CVE reference is CVE-2002-1463, which covers Symantec Raptor Firewall / Enterprise Firewall and similar appliances whose ISN generator keyed only off <srcport, dstport> tuples. A nearby reference is the broader BID 5387 class of weak-ISN findings against Novell NetWare and miscellaneous embedded TCP stacks. Affected versions are firmware lines from the early 2000s; modern Linux (RFC 6528 hashed ISN), Windows (post-XP SP2), and BSD stacks are not in scope.
Tenable still labels this HIGH (CVSS 7.3) because the CVSS base assumes a network-reachable spoof-and-hijack of an authenticated TCP session. That mapping is correct in a vacuum and wrong in practice: Tenable's own VPR score on this plugin is 2.2 (Low), which is the honest assessment. The base CVSS ignores that the affected appliances are 20+ years out of support, that TCP session hijacking has been mooted by ubiquitous TLS, and that the chain requires an off-path attacker who can also guess or sniff sequence state. Treat the HIGH label as legacy CVSS inflation.
4 steps from start to impact.
Identify a target with weak ISN generation
hping3 --seqnum, scapy, or Nmap's --script ipidseq / older tcp-sequence heuristic profile the predictability class. If the ISN is a function only of port tuple or a low-entropy counter, the target qualifies.- Network reachability to a TCP-listening port on the target
- Target runs an ISN-weak stack (Raptor, NetWare 5/6, embedded TCP)
- Virtually no production fleet still runs Raptor Firewall or NetWare in 2026
- Modern OS stacks all implement RFC 6528 randomized ISNs
--osscan sequence-predictability output, Qualys QID 82054Locate a target TCP session to hijack or spoof
- Knowledge of a trusted client IP
- A cleartext or weakly-authenticated service on the target
- Management planes are almost universally over TLS/SSH today
- Source-IP trust is rare outside small embedded environments
Spoof source IP and inject crafted TCP segments
scapy or custom raw-socket code. If the predicted ISN lands within the receiver window, the segment is accepted and the spoofed session proceeds.- Egress path that does not enforce BCP38 / uRPF
- Ability to suppress the real client (or it is offline)
- BCP38 ingress filtering at most ISPs and enterprise edges drops spoofed source IPs
- Stateful firewalls drop out-of-window segments
Execute attacker objective over hijacked channel
- The hijacked protocol is unauthenticated or weakly authenticated at the application layer
- Any TLS / SSH / IPsec termination kills the spoof — attacker cannot complete the handshake without the key material
- Modern protocols use per-session nonces, not bare TCP trust
The supporting signals.
| In-the-wild exploitation | None observed. No active campaigns. Last credible discussion is the 2002–2004 mailing-list era. |
|---|---|
| Public PoC | Conceptual PoCs in *Phrack* 48 (Mitnick-Shimomura technique, 1995) and Michal Zalewski's *Strange Attractors and TCP/IP Sequence Number Analysis* (2001). No modern weaponized tooling. |
| EPSS | Effectively 0 — CVE-2002-1463 is not scored meaningfully by FIRST EPSS; no recent telemetry. |
| CISA KEV | Not listed. |
| CVSS vector (legacy) | CVSS:2.0/AV:N/AC:L/Au:N/C:P/I:P/A:P (7.5). Tenable maps to CVSS:3.0/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H (7.3). Both ignore ecosystem reality. |
| Tenable VPR | 2.2 (Low) — Tenable's own predictive scoring contradicts the static CVSS HIGH label. |
| Affected versions | Symantec Raptor / Enterprise Firewall, Novell NetWare 5.x/6.x TCP stack, miscellaneous early-2000s embedded stacks. All EOL. |
| Fixed versions | Symantec hotfix released 2002; Novell TCPIP.NLM updates; for everything else the fix is decommissioning the device. |
| Exposure data | Shodan / Censys show negligible Raptor/NetWare footprint in 2026 — low thousands globally, mostly honeypots and forgotten DMZ appliances. |
| Disclosure / reporter | Disclosed 2002-08-13; Michal Zalewski's work on TCP/IP ISN entropy is the canonical academic reference. |
noisgate verdict.
The single decisive factor is installed-base collapse: the affected stacks (Raptor Firewall, NetWare) have been end-of-life for ~20 years and represent a vanishingly small share of any modern enterprise fleet. Even where the finding does fire, ubiquitous TLS/SSH/IPsec termination has eliminated the trust-by-source-IP application patterns the attack depends on.
Why this verdict
- Friction — chain is broken at step 3: BCP38 / uRPF / stateful firewalling drops the spoofed segments in virtually every enterprise egress and ingress path built in the last decade.
- Friction — chain is broken at step 4: application-layer cryptography (TLS, SSH, IPsec) has displaced bare-TCP source-IP trust, so even a successful ISN guess buys the attacker nothing on modern services.
- Installed base is sub-1%: Censys/Shodan 2026 sweeps show effectively no surviving Raptor or NetWare TCP stacks in production enterprise networks; the population is dominated by honeypots and orphaned DMZ kit.
- Role multiplier — low-value role (workstation / embedded device): chain is theoretical; blast radius is per-session spoofing of an unauthenticated legacy protocol. No fleet impact.
- Role multiplier — typical role (perimeter firewall, NetWare file server): if a Raptor or NetWare host still fronts a real network, chain could in principle yield management-plane spoof, but installed base is so small (<0.1% of enterprise edges) that it does not lift the floor. Documented evidence: Shodan 2026 product banner counts.
- Role multiplier — high-value role: none applicable. The affected components are not hypervisors, IdPs, PAM, backup, kernel-mode agents, or modern network edge. The high-value-role floor does not engage.
- Tenable's own VPR (2.2) agrees — the static CVSS HIGH is legacy inflation, not current risk.
Why not higher?
MEDIUM would imply a non-trivial fraction of real fleets still expose the affected stacks behind unauthenticated trust relationships. They don't. Pure CVSS-on-paper inflation is exactly the kind of noise this re-assessment exists to filter out.
Why not lower?
IGNORE would be wrong because the plugin can legitimately fire on forgotten embedded gear (industrial gateways, ancient firewall appliances still in service in OT enclaves), and those niches do warrant inventory action. Backlog-hygiene LOW captures that cleanly.
What to do — in priority order.
- Inventory and retire the affected appliance — If 11057 fires, the underlying device is almost certainly EOL Raptor, NetWare, or similar legacy gear. Replace it; there is no LTS branch to pull from. Treat this as backlog hygiene — no mitigation SLA, no remediation SLA beyond the 365-day window.
- Enforce BCP38 ingress filtering and uRPF at edges — Already standard practice, but explicitly verify the path between any internet egress and the device flagged by 11057. This breaks step 3 of the chain regardless of ISN entropy.
- Front legacy services with a modern TLS terminator or IPsec tunnel — If decommissioning is blocked by OT / contractual constraints, isolate the device behind a jump host or VPN concentrator that adds cryptographic session integrity. Source-IP trust dies; ISN entropy stops mattering.
- Disable management protocols that rely on source-IP trust — rlogin, rsh, unauthenticated SNMP-set, and bare TCP RPC channels are the only application-layer payoffs for an ISN spoof. Turn them off on the flagged device and use SSH or out-of-band only.
- OS-level patches on modern Linux/Windows hosts — they were never the vulnerable population for CVE-2002-1463.
- Network IDS signatures for 'predictable ISN' — they flag the property, not the exploitation, and produce noise without action.
- Increasing TCP window size or tuning sysctl on the *scanning* host — does not affect the target's ISN generator.
Crowdsourced verification payload.
Run from your auditor workstation (or a scan jump-host) against the flagged target. Requires nmap and root (raw sockets). Example: sudo ./check-isn.sh 10.0.12.34 443. Outputs VULNERABLE, PATCHED, or UNKNOWN based on Nmap's TCP sequence-predictability class.
#!/usr/bin/env bash
# noisgate verification for tenable:11057 / CVE-2002-1463
# Uses Nmap's TCP Sequence Prediction heuristic to classify ISN entropy.
set -u
TARGET="${1:-}"
PORT="${2:-80}"
if [[ -z "$TARGET" ]]; then
echo "usage: $0 <target> [port]" >&2
exit 2
fi
if ! command -v nmap >/dev/null 2>&1; then
echo "UNKNOWN: nmap not installed" >&2
exit 3
fi
if [[ $EUID -ne 0 ]]; then
echo "UNKNOWN: must run as root for raw-socket OS detection" >&2
exit 3
fi
# -O forces OS detection which emits the TCP Sequence Prediction line.
OUT=$(nmap -O -p "$PORT" -Pn --osscan-guess "$TARGET" 2>/dev/null)
SEQLINE=$(echo "$OUT" | grep -i "TCP Sequence Prediction")
if [[ -z "$SEQLINE" ]]; then
echo "UNKNOWN: nmap returned no sequence prediction data"
exit 3
fi
# Nmap difficulty classes: "trivial joke", "easy", "medium", "formidable",
# "worthy challenge", "good luck!" Anything below 'formidable' is weak.
if echo "$SEQLINE" | grep -Eiq "trivial|easy|medium|random positive increments"; then
echo "VULNERABLE: $SEQLINE"
exit 1
fi
if echo "$SEQLINE" | grep -Eiq "formidable|worthy challenge|good luck"; then
echo "PATCHED: $SEQLINE"
exit 0
fi
echo "UNKNOWN: unparsed nmap output -> $SEQLINE"
exit 3
If you remember one thing.
Sources
What defenders are saying.
Crowdsourced verification outputs.
Results submitted by users who ran the verification payload against their environment.