Releases·Tool Release
Show HN: SecMask – a 66M parameter model for finding secrets in source code
A developer released SecMask, a 66M parameter model for detecting secrets in source code, as a Show HN post linking to GitHub. An earlier version, distilbert-secret-masker, averages 228 downloads per month and was named one of the 'five leading open-source scanners' in a security company's advertisement benchmark despite scoring lowest. The author updated the model in response and claims it is now the top open-source secret detector by their own benchmarks.
3 engagement1 sourceWed, Sep 2, 2026, 12:18 PM
AI Overview
- The author's earlier distilbert-secret-masker model averages 228 downloads per month and was ranked lowest in a security company's benchmark despite being named one of five leading open-source scanners.
- The author updated the model in response and claims it is now the #1 open-source secret detector, according to their own benchmarks.
Entities
Related
Social·Community
Typesafe Jev: Secret Detection Test
A Reddit user with early access to Typesafe.ai's Jev model ran it against a secret-detection benchmark suite built from gitleaks, GitGuardian, and similar sources. Results were described as promising and consistent, though the tester questioned whether Jev's low costs alone justify its use for this task.
A Reddit user with early access to Typesafe.ai's Jev model ran it against a secret-detection benchmark suite built from gitleaks, GitGuardian, and similar sources. Results were described as promising and consistent, though the tester questioned whether Jev's low costs alone justify its use for this task.
7 engagement1 sourceredditThu, Sep 17, 2026, 01:29 PM
Developments·Paper
Researchers propose string-based secret detection to replace regex
A new arXiv paper argues that hardcoded secrets in source code are best detected by analyzing strings themselves rather than relying on regex patterns or context-aware models. The approach aims to overcome limitations in precision, recall, and cross-language generalization.
A new arXiv paper argues that hardcoded secrets in source code are best detected by analyzing strings themselves rather than relying on regex patterns or context-aware models. The approach aims to overcome limitations in precision, recall, and cross-language generalization.
0 engagement1 sourcearxivWed, Aug 5, 2026, 06:55 AM
Products·launched
Sentrint: Security scanner for LLM-built projects
Sentrint is a security scanner designed for projects built with LLMs. It helps developers identify vulnerabilities specific to AI-generated code, addressing the unique security challenges introduced by LLM-assisted development.
Sentrint is a security scanner designed for projects built with LLMs. It helps developers identify vulnerabilities specific to AI-generated code, addressing the unique security challenges introduced by LLM-assisted development.
7 engagement1 sourcehackernewsThu, Aug 20, 2026, 12:46 PM
Products·launched
Bonsai-ninja: local code intelligence and security analysis for LLMs
Bonsai-ninja is an open-source tool that provides local code intelligence and security analysis, designed for humans and local-first LLMs. It helps developers understand and secure their codebase without sending data to the cloud, leveraging LLMs for analysis.
Bonsai-ninja is an open-source tool that provides local code intelligence and security analysis, designed for humans and local-first LLMs. It helps developers understand and secure their codebase without sending data to the cloud, leveraging LLMs for analysis.
1 engagement1 sourceredditMon, Aug 17, 2026, 05:33 AM
Products·building
NeuralScan: scans AI-generated code for exposed secrets
NeuralScan is a tool that scans AI-generated code projects for exposed API keys and dangerous patterns, explaining findings in plain English. It helps non-developers and agents identify and fix security issues in code generated by tools like Lovable, Bolt, and Cursor.
NeuralScan is a tool that scans AI-generated code projects for exposed API keys and dangerous patterns, explaining findings in plain English. It helps non-developers and agents identify and fix security issues in code generated by tools like Lovable, Bolt, and Cursor.
6 engagement1 sourceredditSun, Aug 23, 2026, 08:04 AM