Skip to content

High-Value Targets (Prioritized Programs)

This list highlights programs with the highest recent bounty payouts, prioritized for focused research. Rankings are based on aggregated HackerOne hacktivity for the 6-month window 2025-03 to 2025-08.

Last updated: 2025-09-01

Top Programs (last 6 months)

Rank Program Total Payout Reports Program Page
1 Uber $20,340.00 24 https://hackerone.com/uber
2 Eternal $9,300.00 12 https://hackerone.com/eternal
3 OKG $7,500.00 6 https://hackerone.com/okg
4 TikTok $6,000.00 12 https://hackerone.com/tiktok
5 Sheer $900.00 6 https://hackerone.com/sheer_bbp
6 GitLab $600.00 12 https://hackerone.com/gitlab
7 PayPal $600.00 6 https://hackerone.com/paypal
8 Ferrero $0.00 30 https://hackerone.com/ferrero
9 MediaTek $0.00 24 https://hackerone.com/mediatek
10 Zooplus $0.00 18 https://hackerone.com/zooplus

Strategic High-Value Additions

Newly Interesting Public HackerOne Targets (2026-03-31 scan)

These showed up again in the public HackerOne program page as especially worth a closer look: - 1Password — creative/manual research is explicitly favored; scanners are unlikely to help. - Akamai — CDN/origin/proxy trust boundaries; soft-launch / invite-only posture. - Airbnb — huge real-world workflow surface; strong fit for authZ, recovery, and trust-boundary chains. - Airlock Secure Access Hub — WAF + IAM stack protecting 30k+ apps; edge and identity bugs can be high impact. - Amazon Vulnerability Research Program — broad surface with mature triage expectations. - Anduril Industries — reproducible reports and structured disclosure language suggest higher signal. - Atlassian — large enterprise SaaS surface with frequent workflow/auth complexity.

Operator-quality cue from public advisories (2026-06-30)

Operator-quality cue from public advisories (2026-06-29)

Operator-quality cue from public advisories (2026-06-28)

Operator-quality cue from public advisories (2026-06-27)

Operator-quality cue from public advisories (2026-06-26)

Operator-quality cue from public advisories (2026-06-25)

Operator-quality cue from public advisories (2026-06-24)

Operator-quality cue from public advisories (2026-06-23)

Operator-quality cue from public advisories (2026-06-21)

Operator-quality cue from public advisories (2026-06-11)

Operator-quality cue from public advisories (2026-05-31)

Operator-quality cue from public releases (2026-05-29)

Operator-quality cue from public releases (2026-05-27)

  • Continuous agentic-pentesting products deserve an explicit watchlist bucket when their scope includes source-code ingestion, exploit execution, auth preflights, network egress checks, or local/container runner setup. KeygraphHQ/Shannon describes itself as an autonomous white-box AI pentester for web apps and APIs; its v1.3.0 release added auth-validation/email-login preflight support, blocked cloud metadata ranges in target URL checks, and hardened global npm installs with --ignore-scripts. Those are useful public signals for scoring AI security-testing targets: prioritize programs that document runner sandboxing, credential handling, target allow/block rules, and dependency-install boundaries. Sources: https://github.com/KeygraphHQ/shannon, https://github.com/KeygraphHQ/shannon/releases/tag/v1.3.0

Operator-quality cue from public advisories/research (2026-05-20)

  • AI coding-agent and workspace-edit products deserve a sharper EV bucket after Hacktron's VS Code Copilot applyPatchTool TOCTOU writeup: prioritize programs where issue/PR-to-agent automation, exact-effect approval, protected-path writes, Codespaces/dev-token exposure, or repository-control-file policies are in scope. Source: https://www.hacktron.ai/blog/rce-in-vscode-copilot
  • Identity federation disable/revoke paths are high-value when in scope. The May 20 Keycloak SAML broker advisory reinforces that “disabled IdP” must be enforced at every assertion/session-minting entrypoint, not just in admin UI discovery. Source: https://github.com/advisories/GHSA-x4p7-7chp-64hq
  • Developer control planes and automation glue are still underpriced target surface: Rclone RC, MLflow model serving, setup-php workflows, Diffusers trust_remote_code, and project-local LLM filters all show how tooling defaults can cross command, tenant, or trust boundaries.
  • JWT algorithm confusion and tenant-null authorization collapse remain practical identity seams. Prefer programs that explicitly include SSO, VPN/CAS, tenant/workspace administration, or support tooling in scope and allow safe negative testing.

Program-quality cue from community chatter (2026-04-04)

  • Triage trust is still the strongest quality signal. Fresh Reddit chatter about boycotts, repeated N/A closures, and reward leakage reinforces that process quality can outweigh raw payout tables.
  • AI governance / browser-control products deserve their own EV bucket. If a program’s value depends on observing prompts, shadow-AI use, or policy enforcement across browser-side and alternate data paths, score it higher than a generic SaaS with the same payout range.
  • Supply-chain and release-provenance boundaries remain high-value. Public chatter around Cisco/Trivy and Adobe support compromise is another reminder that build pipelines, third-party support paths, and dependency trust are real attack surfaces when in scope.
  • Metrics-minded teams care about evidence that maps to actual risk reduction. Programs that define acceptable proof, remediation metrics, and re-validation criteria tend to be easier to work with and to profit from.
  • April 2’s heuristics still stand: prioritize programs that explain duplicate or “informative” closures, preserve context, and do not silently patch while refusing mediation.
  • Community distrust itself is a selection signal: when hunters start talking about boycotts, duplicate gaming, or reward leakage, treat that as a durable warning that the program’s process quality may be degrading even if the technical surface is still interesting.
  • Fairness for hard-to-demo classes is a selection signal in its own right: programs that clearly accept blind SSRF/OAST, authZ edge cases, cache poisoning, or desync-style evidence tend to have better ROI than programs that demand unsafe max-exploit proof.
  • Browser-side AI governance and shadow-AI monitoring are now a distinct premium target class: products that can inspect prompts, extension traffic, or in-browser copilots are exposing a trust boundary, not just a DLP use case.
  • VMs/process hygiene is a selection clue: if a team can’t connect scan volume to risk reduction, expect weaker prioritization and slower remediation, which usually means more triage friction for researchers.
  • Hidden build/bootstrap paths are still gold: CodeConnections-style connector tokens and CodeBuild pre-user-code requests are exactly the kind of seams that turn ordinary platform work into high-value trust-boundary research.
  • Outsourced support and account-recovery paths should be scored as first-class attack surfaces whenever the program can influence identity, tickets, or escalation state.

Program-quality cue from community chatter (2026-04-02)

  • April 2 Reddit chatter reinforced that triage transparency and contradiction handling matter more than raw payout numbers. Favor programs that explain duplicate or “informative” closures, preserve context, and don’t silently patch while refusing mediation.
  • Community distrust itself is a selection signal: when hunters start talking about boycotts, duplicate gaming, or reward leakage, treat that as a durable warning that the program’s process quality may be degrading even if the technical surface is still interesting.
  • AI/agent tooling programs should be weighted higher when they expose behavior-validation surfaces like sub-agents, retrieval, browser extensions, or tool-calling, because the high-EV bugs now live in state drift and wrong-tool execution as much as in prompt text.
  • Governance and browser-side mediation products are now a separate EV bucket: if a target’s value depends on observing or constraining user prompts, shadow-AI usage, or policy-bypass paths outside inline controls, score it higher than an ordinary SaaS with similar payout tables.
  • Build/release and dependency-provenance boundaries keep showing up in public incidents; if a program owns its own delivery pipeline or agent framework updates, treat patch cadence and version guidance as part of program quality.
  • Fairness for hard-to-demo classes is a selection signal in its own right: programs that clearly accept blind SSRF/OAST, authZ edge cases, cache poisoning, or desync-style evidence tend to have better ROI than programs that demand unsafe max-exploit proof.

Program-quality cue from community chatter (2026-03-31)

  • AskNetsec’s latest agent-validation discussion is another reminder that AI/tooling programs should be scored higher when they expose browser extensions, retrieval, sub-agents, or tool-calling flows. The EV is in behavior drift and policy bypass across alternate paths, not just prompt text visibility.
Program Priority Reason Program Page
Coinbase 🚨 CRITICAL Major crypto exchange, $50K+ critical bounties, 519 subdomains discovered https://hackerone.com/coinbase
1Password 🔥 HIGH Public page explicitly asks for creative researchers; scanners are unlikely to help, which usually means higher manual-EV surface https://hackerone.com/1password
Akamai 🔥 HIGH Public soft-launch / invite-only posture focused on CDN/origin/proxy-layer trust boundaries https://hackerone.com/akamai
Airlock Secure Access Hub 🔥 HIGH WAF + IAM platform protecting 30k+ apps; complex auth and edge paths tend to pay off https://hackerone.com/airlock
Airbnb 🔥 HIGH Massive real-world workflow surface plus rich public disclosure history; strong candidate for authZ, recovery, and trust-boundary chains https://hackerone.com/airbnb
Amazon VRP 🔥 HIGH Broad product surface with explicit reporting program and mature security posture; worth prioritizing for deep, chain-based testing https://hackerone.com/amazonvrp
Anduril Industries 🔥 HIGH Explicit reproducibility expectations and a mature disclosure policy suggest structured triage and higher-value operational/defense surfaces https://hackerone.com/anduril_industries
Atlassian 🔥 HIGH Large enterprise software surface with frequent releases and plenty of auth/workflow complexity https://hackerone.com/atlassian
Basecamp 🔥 HIGH Publicly values researcher insight and pays for quality; workflow-heavy SaaS tends to reward careful manual work https://hackerone.com/basecamp

Historical Notes (recent activity)

  • Uber: Consistent “BountyAwarded” activity during 2025-03..2025-08 with multiple awards per month.
  • Eternal: Frequent awards in late August; steady reporter engagement.
  • OKG: Notable award spikes mid-to-late August 2025.
  • TikTok: Regular bounty activity; multiple awards in August 2025.
  • Sheer: Lower absolute payouts but steady cadence of awards.
  • GitLab: Smaller individual awards; consistent monthly activity.
  • PayPal: Intermittent awards, including early September.
  • Ferrero/MediaTek/Zooplus: High volume of resolved activity; limited public award amount exposure.

Program Quality Signals (Community/Operational Heuristics)

When two programs pay similar amounts, “program quality” often dominates expected value. Signals that repeatedly show up in community discussions:

  • Clear impact standards (esp. for Blind SSRF / OAST findings):
  • Programs that document what evidence they accept (e.g., DNS/HTTP callbacks, timing-based internal reachability, status-code differentials, header leaks) reduce report churn.
  • Watch for programs that require data exfiltration only for SSRF; this can be a time sink unless you can safely demonstrate higher impact.
  • Bonus signal: they explicitly recognize internal network mapping / WAF-bypass-by-proxy / protocol pivot proofs as meaningful impact when exfiltration is unsafe or impractical.
  • Fair, consistent triage:
  • Low “Informative / N/A” rates for valid vuln classes and predictable severity mapping.
  • Willingness to engage on nuance (business logic, chains, internal reachability) rather than rubber-stamping.
  • Fast feedback loop:
  • Reasonable response SLAs, low ghosting, and credible mediation outcomes.
  • Duplicate pressure vs. depth:
  • Programs with heavy “spray-and-pray” tooling overlap can have high duplicate rates; higher EV often comes from programs where manual, product-understanding bugs (logic, authZ, SSRF, cache poisoning/desync edge cases) are rewarded.
  • Community meta-signal: if hunters consistently report success with a “lighter” stack (proxy + notes + targeted automation) it often correlates with programs where understanding workflows beats running scanners.
  • Scope quality:
  • Modern asset inventory (well-defined subdomain patterns, APIs, mobile, integrations) and explicit third-party boundaries.
  • Disclosure and learning culture:
  • Public write-ups / hacktivity that show interesting classes being rewarded (authZ, SSRF, desync, supply-chain) is a strong indicator the program pays for real risk.
  • Remediation posture (especially VDPs):
  • Some high-signal targets (e.g., government VDPs) can have very long remediation timelines. Great for reputation/impact, but low expected value if you’re optimizing for payout/velocity.
  • “Contribution surface” / CI-CD exposure:
  • Community write-ups about exploits starting from PR comments, CI pipelines, webhooks, integrations, or supply-chain touchpoints are a signal to prioritize programs where those surfaces are in-scope (often higher severity, lower duplicate pressure).

Next Steps

  • Review each program page and policy for current scope and exclusions
  • Start or update program documentation directories using the templates
  • Track changes to scope over time (dated entries)