heySec

@heysec.com

9 Followers 2 Following

Security Made Simple: Improving cybersecurity for all by making it accessible and understandable. đź”— https://heysec.com

Screenshot of the heySec “Kempu — Dashboard” web app. Left menu: Dashboard, API Keys, Statistics, Request Logs, Documentation, Users, Settings. Summary cards show “Total Requests 1,351,” “Blocked % 1%,” “Avg Safety Score 0.56,” and “Quota Used 29.6% (2,956 / 10,000 requests).” A line chart titled “Requests Over Time (30 Days)” shows Total (blue) rising sharply, Passed (green) rising moderately, and Blocked (red) near zero. A donut chart titled “Verdict Breakdown” is mostly green “Passed” with a small red “Blocked” slice. Top right user: “Klemens Arro.”
heySec web app screenshot (heyseclab.ddev.site) on the Activity tab showing a “Researcher Run” marked “Completed,” duration “190s,” “2,779 in,” “8,955 out,” and “11 threats found.” Left pane labeled “AGENT OUTPUT” contains a long text/JSON-style report. Right pane titled “Discovered Threats” lists items including: “Constrained Decoding Attacks (CDA): Grammar-Based Containment Bypass” (tags: delimiter manipulation, critical, new); “Unicode Tag Character Invisible Prompt Injection” (unicode boundary break, high, new); “Special Token Manipulation (MetaBreak): Control Token Exploitation” (token escape, critical, new); “Data Smuggling via Unicode Variant Selectors (Sneaky Bits)” (encoding boundary bypass, high, new); “Control Character Injection: Backspace and Carriage Return Containment Escape” (context boundary escape, high, new); “Repeated Token Attack: Safety Divergence via Token Repetition” (payload reassembly, high, new); and “Adversarial Tokenization: Non-Canonical Token Segmentation