brycewang-stanford USENIX Security Review Process
USENIX Security reviews in two annual cycles, each a self-contained pipeline from registration to notification. The '26 cycle mechanics below were read from usenix.org pages on 2026-07-08 (via search renderings; direct fetch 403'd), with the '25 reviewing-model page as documented history. The '27 mechanics under chairs Adam Doupé (Arizona State) and Andrei Sabelfeld (Chalmers) are only partially posted — treat everything here as one venue-generation's snapshot.
The cycle pipeline, with '26 dates as the worked example
| Stage | Cycle 1 ('26) | Cycle 2 ('26) | What is happening |
|---|---|---|---|
| Registration | Aug 19, 2025 | Jan 29, 2026 | HotCRP record frozen; bidding data set |
| Submission | Aug 26, 2025 | Feb 5, 2026 | Format/policy screening begins |
| Early reject | Oct 7, 2025 | Mar 17, 2026 | Bottom of the pool exits after first-round reads |
| Final notification | Dec 4, 2025 | May 14, 2026 | Accept / shepherd-approval / reject |
| Finals due | Jan 15, 2026 | Jun 11, 2026 | Shepherding + Phase-1 AE complete |
The '25 model (last fully published) shows the shape inside those windows: two first-round reviews; online discussion; survivors gain two more reviews in round two; authors respond; the committee discusses to a decision. Expect roughly that structure until the current cycle's reviewing-model page says otherwise (待核实 for '27).
The decision vocabulary changed — don't plan on 2025 rules
Through '25, USENIX Security ran an Accept / Invited Major Revision / Reject scheme, where a major revision returned to the same reviewers in a later cycle. The '26 CFP retired major revisions. The current outcomes:
- Accept — proceed to camera-ready and mandatory artifact-availability check.
- Accepted on Shepherd Approval — the committee wants the paper conditional on changes a two-week shepherding pass can hold: clarifications, explicit limitations, calibrated claims. Not a lane for new experiments.
- Early reject / reject — out of the cycle. Whether and when the work may return is governed by per-year resubmission restrictions; the '26 CFP delegated restrictions on its Cycle-2 rejects to the '27 chairs. Always check the current restriction text before queuing a resubmission — this is the venue's most frequently misremembered rule.
Strategic consequence: the paper must be finished at submission. The old major-revision safety net — submit at 85% and repair through revision — no longer exists; its replacement (shepherding) fixes prose, not evidence.
What reviewers here weigh
Beyond ordinary top-venue novelty/rigor screens, this committee applies filters that surprise authors from adjacent fields:
- Ethics adequacy is a gate, not a score. Weak Ethical Considerations content can trigger required revisions mid-review or sink the paper regardless of technique (the ethics guidelines page is explicit that expectations bind).
- Threat-model realism. An attack needing implausible attacker position, or a defense evaluated only against non-adaptive adversaries, loses reviewers who ask "who does this actually affect?"
- Openness posture. Since the Open Science policy, a vague availability story is a review-visible weakness, not a post-acceptance detail.
- Measurement validity. Vantage-point, time-window, and ground-truth skepticism is a house specialty.
Reading each email
Early reject → 4 reviews were not spent on you; the pool's bottom tier.
Diagnose fit and framing, not just polish. Consider
usenixsec-topic-selection before re-targeting.
Reject (final) → full reviews attached; mine them for the resubmission and
check the current cross-cycle restriction before re-queuing.
Shepherd approval → acceptance-in-waiting; scope the change list within 48h
(see usenixsec-author-response).
Accept → the artifact clock is already running
(see usenixsec-camera-ready).
Confidentiality and conduct
Submissions are confidential to the committee; reviewer identities are hidden; contacting reviewers or attempting to identify them is out of bounds. Conflicts declared at registration drive assignment — fabricating conflicts to dodge a skeptical expert is chair-level misconduct at USENIX as elsewhere.
Reverify each cycle
- The current reviewing-model description, response phase, and round structure.
- Decision categories and shepherding window ('27 待核实).
- Resubmission restrictions between cycles and across years.
- Early-reject and notification dates for the live cycle.
Output format
[Cycle position] stage now + next date that matters
[Decision model] outcomes in force this cycle (verified on: <date>)
[Exposure] ethics gate / threat-model realism / openness / measurement validity
[Plan] action per possible outcome, with resubmission-restriction check