SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

ScamShield AI Week 2: Hardening Docker Firewalls and Fail-Closed Sandbox Design

0
·8 views

A developer building ScamShield AI, a threat-intelligence platform for analyzing Indian scam messages, documented security hardening of an isolated Docker sandbox used to safely open suspicious links. The sandbox pairs headless Chromium via Playwright with a Mitmproxy network guard. The author found that iptables rule ordering created a bypass, since a stateful ESTABLISHED,RELATED accept rule preceded the drops for private and cloud-metadata addresses, risking SSRF. The firewall was restructured into six phases so prohibited destinations are dropped before any connection-state acceptance. The startup script was also changed from fail-open, which merely warned when iptables was missing, to fail-closed, which aborts the container if rules cannot be applied or verified.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Hand-Built Neural Network Reproduces an If-Else Comparison Without Training

A developer demonstrates how a simple if-else comparison, returning 10 when a is greater than 4 and 12 otherwise, can be encoded as a neural network with fixed weights rather than learned parameters. The approach ports Cheng Lou's handwired JavaScript project to Rust as handwired-rs. The construction chains two ReLU operations to produce a flag that is 1 for values above 4 and 0 otherwise, then converts that flag into the output with a weighted sum. The author argues the value lies not in speed but in producing a fixed sequence of arithmetic steps that can be formally verified across every input. The article notes that a pure linear combination cannot represent such a decision, which is why the nonlinear ReLU is required.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Voice AI Development Gets Easier in 2026 With Mature STT, TTS, and Cloning Tools

Voice interfaces now power applications such as smart-home assistants, customer support agents, and audiobook generators. The core technologies are speech-to-text, text-to-speech, and voice cloning, all of which have improved significantly in recent years. A developer can reportedly build a production-ready voice system in hours rather than weeks. The article offers a practical starting guide centered on ElevenLabs, covering required tools, API key setup, and sample code in Python and Node.js. It also explains how to create a custom cloned voice from a short audio sample using the platform's API.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer finds Claude effort setting, not model choice, drove most of a week's quota burn

A developer exhausted a full week of Claude Ultra quota in three days and initially blamed the model. Benchmarking Opus across effort levels on a routing task showed that costs rose sharply past medium effort, with no gain in correctness. Testing the cheapest model, Haiku, on a regex-engine task matched Python's re on 3,000 patterns, and routine well-specified tasks cost far less on low effort. The author concludes that low effort is safe when outputs can be verified by automated checks, though findings are based on small samples and some runs were excluded. The author also released a free, open-source Claude Code skill that selects models and effort per step and runs only the tests affected by a change.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AU Small Finance Bank probes possible leak from vendor-hosted vehicle loan platform

AU Small Finance Bank says it is investigating a possible data leak involving a Salesforce-hosted SaaS platform that held limited vehicle loan customer data. The bank states its own core systems are unaffected and banking services are operating normally. A clarification filed with stock exchanges on 2 November says the platform was used to onboard about 2.85 million vehicle loan customers, though the number of records actually exposed has not been disclosed. The bank has informed the Reserve Bank of India and CERT-In, and a forensic investigation is underway. A group calling itself ScarlettGroup has claimed responsibility and threatened to publish data, but its claims remain unconfirmed.

0 comments Read more at DEV Community