registry / iflow-engine-base / 4.1.14
iflow-engine-base@4.1.14
BIM Engine SDK for Vue2, Vue3, React and HTML
AI Security Review
scanned 5d ago · by lpm-firewall-aiThe external request appears to support the SDK’s AI-assisted model encoding and is reached through a public method, so the source does not establish malicious exfiltration. However, a reusable DeepSeek key is embedded in the client bundle and its prefix is logged, creating a concrete key-abuse risk.
Static reason
High-risk behavior combination matched malicious policy.
Trigger
OneClickEncoding construction logs the key prefix; invoking its public startChat method sends the supplied text.
Impact
The key may be extracted and abused, and data supplied to startChat is sent to DeepSeek.
Mechanism
The SDK includes a default API key and uses it as a Bearer credential for a chat completion request.
AI rationale
The external request appears to support the SDK’s AI-assisted model encoding and is reached through a public method, so the source does not establish malicious exfiltration. However, a reusable DeepSeek key is embedded in the client bundle and its prefix is logged, creating a concrete key-abuse risk.
Evidence
dist/chunks/EngineKernel-CGMWSboO.jsdist/index.d.tslocalStorage key bim_encoding_<url>
Network endpoints1
api.deepseek.com/v1/chat/completions
Decision evidence
public snapshotAI called this Suspicious at 86.0% confidence as Dangerous Capability with medium false-positive risk.
Evidence for warning
- The SDK bundle embeds a DeepSeek API key and logs its prefix during OneClickEncoding construction.
- The public startChat(url, str) method sends its input to DeepSeek’s chat completions endpoint using the embedded key.
- A zero-width character is embedded in the repeated extended property identifier, making the source harder to inspect.
Evidence against
- The package describes itself as a BIM engine SDK and has no lifecycle scripts; the flagged declaration file is API typing.
- The DeepSeek request is reached through the SDK’s explicit startChat method, consistent with its AI-assisted model-encoding feature.
Behavioral surface
CryptoEvalFilesystemNetwork
HighEntropyStringsMinifiedPossibleObfuscationTelemetryUrlStrings
Source & flagged code
7 flaggeddistGaussian/gaussian.umd.jsView file
1(function(rt,St){typeof exports=="object"&&typeof module<"u"?St(exports,require("three"),require("@mkkellogg/gaussian-splats-3d")):typeof define=="function"&&define.amd?define(["ex...
L2: \0`,ie+=g(Q,2),ie+=C.magic,ie+=g(m,2),ie+=g(_,2),ie+=g(oe.crc32,4),ie+=g(oe.compressedSize,4),ie+=g(oe.uncompressedSize,4),ie+=g(i.length,2),ie+=g(H.length,2),{fileRecord:T.LOCAL_F...
L3: `)}return u},a.decode=function(d,x){if(typeof d!="string")throw new TypeError('"input" must be a string.');if(typeof x!="string")throw new TypeError('"alphabet" must be a string.')...
Low
Eval
Package source references a known benign dynamic code generation pattern.
distGaussian/gaussian.umd.jsView on unpkg · L1dist/chunks/EngineKernel-CGMWSboO.jsView file
7378contains invisible/control Unicode U+200C (zero width non-joiner)
extended<U+200C> = /* @__PURE__ */ new Map();
Critical
Trojan Source Unicode
Source contains bidi control or invisible Unicode characters associated with Trojan Source attacks.
dist/chunks/EngineKernel-CGMWSboO.jsView on unpkg · L7378Findings
2 Critical2 High4 Medium6 Low
CriticalTrojan Source Unicodedist/chunks/EngineKernel-CGMWSboO.js
CriticalTrigger Reachable Dangerous Capabilitydist/index.d.ts
HighShips High Entropy Blobdist/assets/hdr/17.hdr
HighSemantic Analysis Limiteddist/bim-engine-sdk.umd.js
MediumNetwork
MediumShips Wasm Moduledist/draco/draco_decoder.wasm
MediumOversized Source Filedist/bim-engine-sdk.umd.js
MediumStructural Risk Force Deep Review
LowScripts Present
LowEvaldistGaussian/gaussian.umd.js
LowFilesystem
LowHigh Entropy Strings
LowTelemetry
LowUrl Strings
Affected versions and remediation
This report applies to iflow-engine-base@4.1.14.
See all versions of iflow-engine-base for other recorded verdicts.
- Review the evidence and your use of iflow-engine-base@4.1.14 before allowing it. Restrict the permissions described in this report.
- Choose an independently verified alternative or release. This report does not establish that other versions are safe.
Evidence last updated: .
Think this verdict is wrong? Dispute it on GitHub or read how reports are produced.