Security researcher / Chile

00 / About

CamiloVera.

Open-source vulnerability research

Browse my research

RESEARCH FILE / 01 PUBLIC ADVISORY

CVE-2026-0540

DOMPurify

Mutation XSS through re-contextualization

Sanitized HTML can become dangerous again when it is parsed in a different context.

Sanitized HTML
Different context
Unsafe again

Research creditCamilo Vera, Cristian Vargas and Scott Moore

Read the write-up Original advisory ↗

Illustrative scene based on published research

I build tools for mobile security and maintain packages in Nixpkgs.

More about me

01 / Research

Explore the findings by product, with links to the write-ups and advisories.

Selected findings

Research and advisories

7 records

01CVE-2026-0540

DOMPurify

Mutation XSS through re-contextualization

Sanitized HTML can become dangerous again when it is parsed in a different context.

Research creditCamilo Vera, Cristian Vargas and Scott Moore

Read the write-up ↗Original advisory ↗

02CVE-2025-15265

Svelte

Hydratable key script-breakout XSS

A hydration key crosses from data into executable markup during server-side rendering.

Research creditCamilo Vera

Read the write-up ↗Original advisory ↗

03CVE-2025-9624

OpenSearch

Nested boolean/disjunction asymmetric DoS

A small query can force a disproportionately expensive amount of work.

Research creditCamilo Vera

Read the write-up ↗Original advisory ↗

04CVE-2025-9375

xmltodict

XML injection through crafted dictionary keys

Dictionary keys can be interpreted as XML structure rather than ordinary data.

Research creditCamilo Vera

Read the write-up ↗Original advisory ↗

05CVE-2025-8101

Linkify.js

Prototype pollution and HTML attribute injection

Inherited properties can influence generated links, leading to script execution.

Research creditCamilo Vera

Read the write-up ↗Original advisory ↗

06CVE-2025-7969

markdown-it

Fenced-code rendering XSS

Custom highlighting output can cross the trust boundary of rendered Markdown.

Research creditCamilo Vera

Read the write-up ↗Original advisory ↗

07CVE-2025-67635

Jenkins

Unauthenticated CLI full-duplex endpoint DoS

A race condition and missing timeouts can tie up server threads.

Research creditCamilo Vera · personal technical write-up

Read the write-up ↗

1 / 7 · DOMPurify

These are selected published findings. The short explanations are paraphrases; each record includes its research credits.

02 / Projects

Open-source projects

Tools I work on

2 projects

Tools I've built or contributed to for security research and reporting.

flutterdec

A Flutter AOT decompiler that produces pseudo-Dart for mobile reverse engineering, including obfuscated builds.

Camilo Vera · author

View repository

Grapefruit

Flutter instrumentation for iOS and Android, including a channel inspector for traffic between Dart and native code.

Camilo Vera · contributor

View repository

03 / Open source

Nixpkgs and open source

Reviews and pull requests

I maintain packages in Nixpkgs and review other contributors' changes. Here are selected reviews and pull requests.

NixOS/nixpkgs #581914 ↗Review comment

bitwarden-cli: manually prune deps

PR by pyrox0

Inline reviewpkgs/by-name/bi/bitwarden-cli/package.nix

caveravleft a review commentNov 1, 2026

Yes, this changes between npm 10.9.9 and 11.19.0. I tested both on the same Node 24 runtime with identical workspace dependencies and prune flags: npm 10 removes the root dev dependencies, while npm 11 leaves them behind. I also rebuilt Bitwarden with Node 22/npm 10 and the standard prune, and it removed the dev dependencies without the workaround. If we want to be more specific, the comment can mention npm 11 specifically.

Comment excerpt on another contributor’s PROpen discussion ↗

NixOS/nixpkgs #581914 ↗Review comment

bitwarden-cli: manually prune deps

PR by pyrox0

Inline reviewpkgs/by-name/bi/bitwarden-cli/package.nix

caveravleft a review commentNov 1, 2026

Just a nit about the comment wording: we can replace "for some reason" with the actual cause. npmInstallHook passes --workspace=apps/cli, which leaves root-level dev dependencies behind even though the root node_modules is copied into the output. The manual prune is fine, this would just make the reason clearer. Something like: # The default prune is scoped to npmWorkspace and leaves root devDependencies behind. # Prune from the repository root before installing node_modules.

Comment excerpt on another contributor’s PROpen discussion ↗

NixOS/nixpkgs #510608 ↗Review comment

gemini-cli: 0.37.2 -> 0.38.1; gemini-cli-bin: 0.37.0 -> 0.38.1

PR by MiniHarinn

Inline reviewpkgs/by-name/ge/gemini-cli/package.nix

caveravleft a review commentApr 16, 2026

Looks like gitMinimal might not be needed here, preConfigure already pre-generates git-commit.ts with finalAttrs.src.rev, so git is never actually called during the build. Is there a reason to adding it?

Comment excerpt on another contributor’s PROpen discussion ↗

NixOS/nixpkgs #580821 ↗Comment

python3Packages.semgrep: fix `semgrep-core` patchelf and pin mcp

PR by pinarruiz

caveravcommentedOct 30, 2026

I tested the current PR head (9c1bee2) on x86_64-linux, and the workaround builds successfully. I also tested the upstream fix from NixOS/patchelf#672. Restoring the original semgrep-core recipe, with a single Patchelf invocation and the original longer RPATH, reproduces the assertion with unpatched Patchelf 0.19.2. Applying only #672's source changes fixes the build without adding the glibc symlinks. The resulting Semgrep package also passes its Nixpkgs-enabled tests and local Python/JavaScript scan checks. I retained the independent mcp_1 change for this comparison. I think using the upstream Patchelf fix is the cleaner solution to the ELF issue than adding this package-specific workaround.

Comment excerpt on another contributor’s PROpen discussion ↗

NixOS/nixpkgs #583658 ↗Pull request

neovide: fix source and cargo hashes

MergedPull request

caverav opened this pull request on Nov 5, 2026

Description of changes Fix the Neovide source and Cargo vendor hashes introduced in #583110. The Neovide and Skia sources had the same hash, so Nix could use the Skia source for both. Cargo vendoring then failed because the Skia source doesn't contain a Cargo.lock. Resolves #583616. Things done Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs. Follows the automation/AI policy. Built on platform: x86_64-linux aarch64-linux aarch64-darwin Tested, as applicable: Basic functionality of all binary files, usually in ./result/bin/. Ran nixpkgs-review on this PR. Package tests at passthru.tests. NixOS tests in nixos/tests. Tests in lib/tests or pkgs/test for functions and "core" functionality. The package build completed on…

+2−21 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

NixOS/nixpkgs #506451 ↗Package update

semgrep: 1.152.0 -> 1.161.0

MergedPackage update

caverav opened this pull request on Apr 4, 2026

Updates semgrep from 1.152.0 to 1.161.0. update.sh was adjusted in the same PR because upstream changed the Linux wheel naming on PyPI for current releases, and using nix-hash keeps the updater working without relying on the experimental nix hash interface. This release introduced structural changes to how the semgrep-core binary is distributed and linked in the upstream Python wheels, requiring adjustments to the Nix derivation to maintain a functional package. Starting with version 1.159.0, the semgrep-core binary is dynamically linked against several shared libraries (e.g., libtree-sitter, libzstd, libcurl, libgmp). Upstream now bundles these dependencies within a libs/ directory inside the Python wheel. Upstream has updated their build target…

+77−253 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

NixOS/nixpkgs #513253 ↗Package update

bitwarden-cli: 2026.3.0 -> 2026.4.1

MergedPackage update

caverav opened this pull request on Apr 25, 2026

bitwarden-cli: 2026.3.0 -> 2026.4.1 expose passthru.npmDeps so nix-update can refresh npmDepsHash during unattended updates Bitwarden describes cli-v2026.4.1 as a re-release of cli-v2026.3.0 after the malicious npm-only @bitwarden/cli@2026.4.0 publish was revoked: https://github.com/bitwarden/clients/releases/tag/cli-v2026.4.1 https://community.bitwarden.com/t/bitwarden-statement-on-checkmarx-supply-chain-incident/96127 The upstream source diff between cli-v2026.3.0 and cli-v2026.4.1 is only the CLI version metadata bump, but updating nixpkgs keeps the packaged version aligned with the clean re-release and avoids leaving the package marked as outdated.

+5−41 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

NixOS/nixpkgs #504271 ↗Pull request

igf: init at 1.1.2

OpenPull request

caverav opened this pull request on Mar 28, 2026

Summary add frida-node-prebuilds for the pinned Frida Node.js N-API bindings used by igf add igf 1.1.2, the Grapefruit dynamic instrumentation server for Frida-based mobile application analysis package the published npm release because it ships the built CLI and web assets that are absent from the upstream v1.1.2 tag build r2hermes-wasm and better-sqlite3 locally, while sourcing the matching upstream Frida bindings from the reusable helper package Homepage: https://github.com/ChiChou/grapefruit npm package: https://www.npmjs.com/package/igf Upstream release: https://github.com/ChiChou/grapefruit/releases/tag/v1.1.2 Notes for reviewers igf intentionally ships both frida (17.x) and frida16 (16.x) because upstream supports runtime selection between…

+403−03 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

NixOS/nixpkgs #503500 ↗Pull request

ghidra-extensions.ghidra-mcp: init at 5.1.0

OpenPull request

caverav opened this pull request on Mar 25, 2026

Package the published GhidraMCP 5.1.0 release from bethington/ghidra-mcp as a Ghidra extension. This now builds the extension from the tagged source release using the shared ghidra.buildGhidraExtension helper instead of installing the upstream release ZIP asset. Upstream release notes: https://github.com/bethington/ghidra-mcp/releases/tag/v5.1.0

+36−02 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

NixOS/nixpkgs #523892 ↗Pull request

stdenv/check-meta: route warnings through handleEvalIssue

ClosedPull request

caverav opened this pull request on May 25, 2026

Fixes #523712. meta.problems (from #494416) currently can't be used on any package: the moment a maintainer adds e.g. meta.problems.removal.message = "...", the Eval CI job aborts on every system. The eval driver in ci/eval/default.nix treats any stderr output as a chunk failure, and meta.problems warnings are emitted via lib.warn, which writes to stderr. That's blocking #522364 (gemini-cli deprecation notice). The change giveWarning in pkgs/stdenv/generic/check-meta.nix now routes through the existing config.handleEvalIssue hook (the same one the error path already uses), falling back to lib.warn when no handler is set: if config ? handleEvalIssue then builtins.seq (config.handleEvalIssue warning.kind msg) acc else warn msg acc; To pass the…

+26−14 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

NixOS/nixpkgs #495238 ↗Pull request

objection: init at 1.12.3

ClosedPull request

caverav opened this pull request on Feb 28, 2026

Add new package objection (pkgs/by-name/ob/objection/package.nix). Add maintainer entry caverav (Camilo Vera Vidales) in maintainers/maintainer-list.nix. Package source: PyPI objection version 1.12.3. Upstream release notes: https://github.com/sensepost/objection/releases/tag/1.12.3

+61−02 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

caverav/flutterdec #74 ↗Review comment

Flutterdec dart serwalker

PR by kalixtez

Inline reviewcrates/flutterdec-serwalker/src/cluster/mod.rs

caveravleft a review commentNov 6, 2026

from_utf16_lossy changes valid Dart string contents here. Dart strings can contain unpaired surrogates, so [0xD800], [0xDC00], and [0xFFFD] are three distinct values. With both shipped profiles, all three decode to [0xFFFD], and the pool producer still marks them as Exact. Please preserve the original code units or a lossless escaped representation through model production. A lossy string can be useful for display, but it shouldn't replace the stored value. Add a regression that keeps these three constants distinct in the exported pool and also covers a normal surrogate pair.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Review comment

Flutterdec dart serwalker

PR by kalixtez

Inline reviewcrates/flutterdec-serwalker/src/info_producer/functions_info.rs

caveravleft a review commentNov 6, 2026

The full payload range is useful, but it doesn't give consumers the callable entrypoints. In the monomorphic fixture, the Function starts at 0x1010 while its Code-pool target is 0x1028. The existing exact-address join in build_pool_semantic_hints then loses the selector, owner, and library. Selecting that function with EntryVa(0x1028) also fails. Keep the full range, but carry the decoded callable aliases into metadata lookup and entry-address selection. Moving the range start forward would discard part of the body, and treating every interior address as an entry would be too broad. A shared-code test should also make sure we don't assign a unique owner when several declarations share a body. I reproduced this with the decoder and consumer…

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Review comment

Flutterdec dart serwalker

PR by kalixtez

Inline reviewcrates/flutterdec-serwalker/src/info_producer/functions_info.rs

caveravleft a review commentNov 6, 2026

A Function owner can be a PatchClass. Dart resolves that through wrapped_class, but this direct Class downcast drops the relationship. I reproduced this with a reachable PatchClass whose wrapped Class is already present in the model: the exported Function still has owner=None, so its class/library attribution is lost. Resolve the wrapper by reference before converting to ClassId. Direct-Class and PatchClass owners should resolve to the same class/library; genuinely unresolved owners can remain absent.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Review comment

Flutterdec dart serwalker

PR by kalixtez

Inline reviewcrates/flutterdec-serwalker/src/lib.rs

caveravleft a review commentNov 6, 2026

This verifies the file against profile_sha256, but SnapshotInput can also supply a different digest in compatibility.profile_sha256. Passing the correct LocalSend profile/digest with a conflicting binding returns a validated model containing both digests. The final validation doesn't catch the disagreement because its context copies the same binding. When a compatibility binding is present, require its profile digest to match the verified digest too. Add tests for matching, conflicting, and absent bindings. The built-in CLI currently passes None, so its normal path isn't affected. This check is needed for callers of the new public API; it shouldn't require adding a registry binding to native runs.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Review comment

Flutterdec dart serwalker

PR by kalixtez

Inline reviewcrates/flutterdec-core/src/pipeline/model.rs

caveravleft a review commentNov 6, 2026

The identity-rejection branch above can return heuristic output before this call is reached. It refuses pinned external backends, but still allows an explicit internal request to fall back. That doesn't match the new meaning of --adapter-backend internal. With the synthetic FullJIT header, both info and a capped decompile exit successfully with builtin_parser=null and core_fallback_reason=identity_rejected, even though internal was explicitly selected. Reject that request when Serwalker can't run, and keep heuristic recovery available under auto. Add a test for identity rejection before profile lookup, alongside the existing parse-error cases.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Review comment

Flutterdec dart serwalker

PR by kalixtez

Inline reviewcrates/flutterdec-serwalker/src/program_roots/mod.rs

caveravleft a review commentNov 6, 2026

Nonblocking for the current model output, since these roots aren't projected into it yet, but this leaves out the last recent slot. A 64-entry ring uses -64 for slot63; the maximum repeat count of 63 is a separate limit. The writer-shaped fixture misreads that token as a fresh Code index instead of returning the saved CodeIndex(64). Derive the bound from the ring size: Suggested change const RECENT_MIN: i64 = -MAX_REPEAT; const RECENT_MIN: i64 = -(RECENT_COUNT as i64); Add adjacent slot62 and slot63 cases, with an intervening reference so the writer doesn't encode the latter as a repeat.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Review comment

Flutterdec dart serwalker

PR by kalixtez

Inline reviewcrates/flutterdec-serwalker/src/cluster/mod.rs

caveravleft a review commentNov 6, 2026

Nonblocking: the test helpers rely on FLUTTERDEC_PROFILES_DIR to find checked-in fixtures. With it unset, cargo test --workspace --locked has 23 missing-profile failures. Setting the directory makes the suite pass. Load the fixture files from a manifest-relative path or a shared test loader, rather than requiring ambient configuration. Avoid setting a process-wide environment variable from parallel tests. The installed CLI already finds its packaged profiles, so the production package layout can stay as it is.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Review

Flutterdec dart serwalker

PR by kalixtez

caveravcommentedNov 6, 2026

I tested the built-in path on LocalSend, including the staged release and the Nix package. It parses successfully, and the configured workspace suite passes. I'm requesting changes for the cases below that lose recovered data, accept conflicting profile information, or fall back without running the explicitly selected parser. The dispatch-table and test-fixture comments are nonblocking for the current model output.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #103 ↗Review

build(deps): bump DeterminateSystems/magic-nix-cache-action from 14 to 15

PR by dependabot[bot]

caveravcommentedOct 26, 2026

Reviewed the dependency update and verified all required CI checks pass.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #104 ↗Review

build(deps): bump clap from 4.6.6 to 4.6.7

PR by dependabot[bot]

caveravcommentedOct 26, 2026

Reviewed the dependency update and verified all required CI checks pass.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #106 ↗Review

build(deps): bump libc from 0.2.189 to 0.2.190

PR by dependabot[bot]

caveravcommentedOct 26, 2026

Reviewed the dependency update and verified all required CI checks pass.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #108 ↗Review

build(deps): bump serde_json from 1.0.151 to 1.0.154

PR by dependabot[bot]

caveravcommentedOct 26, 2026

Reviewed the dependency update and verified all required CI checks pass.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #74 ↗Comment

Flutterdec dart serwalker

PR by kalixtez

caveravcommentedNov 7, 2026

I also found a bug in the shared validator that we should fix before merging. It already exists on main, so this PR didn't introduce it. A function can really be named unknown, but the validator treats that spelling as a placeholder and rejects the entire model. I reproduced it with a name recovered as Exact. The fix belongs in the validator. Serwalker should keep the name it recovered, not drop or rename it to get past validation. We should allow genuine names like unknown and Unknown, while keeping missing names absent.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #107 ↗Comment

build(deps): bump zip from 8.6.0 to 9.0.3

PR by dependabot[bot]

caveravcommentedOct 26, 2026

Blocking issue: CI fails after upgrading zip to 9.0.3 (lockfile resolved 9.0.5). ZipFile::name() now returns a Result, so entry.name().to_string() at crates/flutterdec-loader/src/lib.rs:88 fails to compile. Please handle the result or retain a compatible zip version, then rerun CI.

Comment excerpt on another contributor’s PROpen discussion ↗

caverav/flutterdec #91 ↗Pull request

research(decompiler): make register loss honest - value recovery, structure, stubs, and an audited annotation path

MergedPull request

caverav opened this pull request on Aug 8, 2026

flutterdec emits pseudo-Dart from stripped Flutter AOT ARM64 binaries. This branch is one research campaign on why that output does not read like Dart, measured against two real release APKs (LocalSend, Immich) at full scope. The previous title named the runtime-stub work, which is roughly a third of what is here. The through-line: AOT register allocation destroys most local value identity, and the decompiler was inventing values to cover the loss. Almost everything below is either stopping a fabrication, or replacing it with an annotation that says what is actually known. What shipped Verifiable from git log main..HEAD (~55 behaviour commits). Grouped by subsystem, not by research section: Value recovery and the lifter stop claiming values…

+21638−54175 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

caverav/flutterdec #97 ↗Pull request

feat(adapter)!: typed snapshot identity, adapter protocol v1, and ProgramModel v4

MergedPull request

caverav opened this pull request on Aug 31, 2026

Summary First of two stacked PRs for #96. This one lands the semantic contract layer: a typed, header-derived snapshot identity; a one-shot adapter protocol v1; and a clean cutover of every producer and consumer from the old ProgramModel (v2/v3) to ProgramModel v4. The registry, runtime-loaded content-addressed profiles, and sandboxed native subprocess execution are PR 2. Typed snapshot identity (flutterdec-loader::identity) Compatibility decisions used to rest on a bare snapshot_hash: String that could equally be a header read, a byte scan, or the literal "unknown", and on a snapshot kind that was never read at all. A 32-hex run in a data section could therefore authorize the same parser as a real header. SnapshotIdentity makes the difference a…

+12005−547943 files changed

Condensed PR excerpt · status at captureOpen pull request ↗

This selected GitHub activity is checked during a refresh build, not while you browse. The dates below show when the excerpts and PR statuses were last fetched.

Activity collected: 7 Nov 2026, 02:13 UTC
PR statuses checked: 7 Nov 2026, 02:13 UTC

Reviews on other contributors' pull requests come first, then my merged, open, and closed unmerged pull requests. Generated reports and comments on my own pull requests are excluded.

04 / CTFs

Selected results

CTF results

A selection of individual and team results, including international appearances.

1st place · Fluid Attacks team

OAS Cyber Americas Cup 2026

Camilo Vera Vidales, Adrian Castañeda, Oscar Uribe Londoño and Cristian Vargas

1st place · Fluid Attacks team

Ekoparty 2026 Main CTF

In-person at Ekoparty in Buenos Aires

1st place

Campo de Marte Attack & Defense CTF

Camilo Vera and Julio Cerda Villablanca

1st place

Campo de Marte CTF 2025

Camilo Vera, Julio Cerda Villablanca, Cristobal Cordova, Diego Arias Cauas and Pablo Aravena · 16 May

1st place · Team Chile

Latin America Cybersecurity Challenge 2025

Pwnvengers · Camilo Vera Vidales, Daniel Espinoza, Julio Cerda Villablanca, Robert Antonio P. and Pablo Aravena

1st place · Pwnvengers

Campo de Marte CTF · 29 November

Camilo Vera, Daniel Espinoza, Matia Cornejo, Julio Cerda and Pablo Aravena · 10 teams

1st place · Chile

OEA Cyber Challenge 2023

Pablo Aravena, Diego Arias, Camilo Vera Vidales and Sebastián Zapata Ascencio

1st place · Pwnvengers · 1,250 points

Campo de Marte CTF · 20 Oct 2023

Camilo Vera, Pablo Aravena, Caetano Borges, Julio Cerda and Daniel Espinoza · held with Chile's PDI

1st place · Red Pwndas

Hack.ING CTF · Security UC Club

Pablo Aravena, Matia Cornejo and Camilo Vera Vidales · 24 March · 160+ participants

1st place · Pwnvengers

Hack.ING 2024

Daniel Espinoza, Matia Cornejo, Camilo Vera, Julio Cerda and Pablo Aravena · 300+ participants

05 / Writing

Write-ups and field notes

Writing

10 posts

I write about vulnerability research, reverse engineering, and mobile security, with the occasional note on CTFs.

Writing archive10 articles

Find a post

ArticleOct 12, 2026

How flutterdec works

A guided look at Flutter AOT decompilation, Dart snapshots, Serwalker deserialization, and how flutterdec reconstructs readable code.

Read article

ArticleApr 11, 2026

CTFs Are Not Dead, They’re Just Growing Up

Are LLMs killing CTFs? Or are they just forcing them to evolve? Exploring the new landscape of cybersecurity competitions.

Read article

ArticleMar 29, 2026

CVE-2026-0540: How a CTF Detour Led Us to a DOMPurify mXSS - Daft

A detailed write-up of the DOMPurify mXSS we found during a CTF detour, affecting 3.1.3 through 3.3.1 and fixed through a small patch series in 3.3.2.

Read article

ArticleMar 17, 2026

CVE-2025-15265: Svelte Hydratable Key SSR XSS - Lydian

A technical write-up of the Svelte async SSR XSS I discovered in hydratable() key serialization, affecting versions 5.46.0 through 5.46.3.

Read article

ArticleDec 12, 2025

CVE-2025-67635: Unauthenticated DoS in the Jenkins CLI full-duplex endpoint

Race conditions and missing timeouts in Jenkins' plain CLI endpoint let anyone exhaust servlet threads without Overall/Read.

Read article

ArticleNov 25, 2025

CVE-2025-9624: Nested Boolean/Disjunction Asymmetric DoS in Amazon's OpenSearch query_string - Chick

How I found CVE-2025-9624, an asymmetric Denial of Service in Amazon's OpenSearch's query_string handling, and how it was fixed with search.query.max_query_string_length.

Read article

ArticleSep 1, 2025

CamiloVera.

DOMPurify

Research and advisories

DOMPurify

Svelte

OpenSearch

xmltodict

Linkify.js

markdown-it

Jenkins

Tools I work on

flutterdec

Grapefruit

Reviews and pull requests

CTF results

OAS Cyber Americas Cup 2026

Ekoparty 2026 Main CTF

Campo de Marte Attack & Defense CTF

Campo de Marte CTF 2025

Latin America Cybersecurity Challenge 2025

Campo de Marte CTF · 29 November

OEA Cyber Challenge 2023

Campo de Marte CTF · 20 Oct 2023

Hack.ING CTF · Security UC Club

Hack.ING 2024

Writing

How flutterdec works

CTFs Are Not Dead, They’re Just Growing Up

CVE-2026-0540: How a CTF Detour Led Us to a DOMPurify mXSS - Daft

CVE-2025-15265: Svelte Hydratable Key SSR XSS - Lydian

CVE-2025-67635: Unauthenticated DoS in the Jenkins CLI full-duplex endpoint

CVE-2025-9624: Nested Boolean/Disjunction Asymmetric DoS in Amazon's OpenSearch query_string - Chick

The Obsolescence of SSL Pinning in Mobile App Security

CVE-2025-9375: XML Injection Vulnerability in xmltodict 0.14.2 - Mono

CVE-2025-7969: Markdown-it Fence Rendering XSS - Fito

CVE-2025-8101: Linkify.js Prototype Pollution & XSS - Charly