Jev for Vulnerability and Attack Surface Analysis

A new open-source tool called jev maps the attack surface of Python, JavaScript and TypeScript codebases and flags likely vulnerabilities down to the exact line for about one cent per repository, using a classifier that answers fixed questions with probabilities in under a second at roughly $42 per billion input tokens. Tested only against two deliberately vulnerable apps on 2026-10-02, jev scanned PyGoat (Python/Django) for $0.010 and NodeGoat (JavaScript/Express) for $0.007, surfacing issues including SQL injection, eval, pickle.loads, SSRF, open redirect and NoSQL $where injection. The tool requires a TypeSafe API key, caps each run at $0.05, and hands its ranked issue list to an AI agent with instructions to validate rather than fix each finding.

Maps the attack surface of a backend codebase and flags likely vulnerabilities, down to the suspicious line, for about one cent per repo. Try it without installing anything: https://franciscocarloserra.github.io/jev-attack-surface-analysis/ https://franciscocarloserra.github.io/jev-attack-surface-analysis/ read-only results on PyGoat and NodeGoat . You give it a repo Python, JavaScript or TypeScript and a budget in dollars. You get: - a map of the codebase, one block per file, colored by how suspicious it is; - a ranked list of potential vulnerabilities , each pointing to the exact line e.g. user input reaching SQL, eval , a shell or an outbound request ; - a copy button or print issues.py that hands that list to an AI agent, with the instruction to validate each issue, not to fix it. No LLM is involved. Plain Python reads the code; jev , a classifier that answers fixed questions with probabilities in under a second and at about $42 per billion input tokens, does all the judging. Like a magnifying glass: it looks at the whole repo coarsely, then zooms into the suspicious parts only. At each level jev rates every item, and only the hot ones are opened at the next level. repo │ ▼ 1. directories jev reads names only → drops tests, docs, migrations │ ▼ 2. files jev reads imports + signatures → exposure: none / low / medium / high │ ▼ 3. functions jev reads the code → does external input reach a dangerous operation? │ ▼ 4. lines jev picks one of the function's lines → where the vulnerability happens │ ▼ ranked issues ──► viewer / copy ──► your agent validates them - Heat 0 to 1 : how suspicious an item is, computed from jev's probabilities. - Budget : each level gets a share; what one level does not spend passes to the next. The hottest items go first, so if money runs out, what is left out is the least suspicious. Tested only against two apps that are vulnerable on purpose measured 2026-10-02 : | Repo | Language | Cost | Top issues found | |---|---|---|---| | PyGoat https://github.com/adeyosemanputra/pygoat | Python / Django | $0.010 | SQL injection, eval , pickle.loads , SSRF | | NodeGoat https://github.com/OWASP/NodeGoat | JavaScript / Express | $0.007 | eval on request body, open redirect, SSRF, NoSQL $where injection | You need a TypeSafe API key for jev. python3 -m venv .venv && .venv/bin/pip install -r requirements.txt export TYPESAFE API KEY=... git clone --depth 1 https://github.com/adeyosemanputra/pygoat repos/pygoat python3 viewer server.py open http://localhost:7801/heatmap viewer.html In the viewer pick the repo, set a budget max $0.05 per run and press Run analysis . From the shell instead: .venv/bin/python attack surface scan.py repos/pygoat --budget 0.03 python3 print issues.py examples/pygoat/scan result.json --top 10 Every run is saved in examples/<repo /runs/<run id .json with date, scanner version and settings hash and the latest one in examples/<repo /scan result.json . Agents: see AGENTS.md for the commands and the result format. | File | What it is | |---|---| | attack surface scan.py | the scanner | | classification levels.json | every setting: questions to jev, categories, thresholds, budget shares, languages | | heatmap viewer.html + viewer server.py | the viewer, and the small server that lets it start runs | | print issues.py | issue list as text, for agents | | examples/ | saved runs | - Ask different questions or change thresholds: edit classification levels.json . - Add a language : add an entry to languages in the same file file extensions and the parser's names for imports, functions and classes . - Add a level e.g. HTTP routes : write one extract <unit function in attack surface scan.py , register it in UNIT EXTRACTORS and add the level to the JSON. - It ranks candidates for review; it does not prove a vulnerability exists. - Each function is judged on its own, so a flaw spread across several functions can be missed.