pip install red-specter-cipher
AI agents handle encrypted traffic, manage API keys, authenticate against services, and trust certificate chains they've never verified. CIPHER attacks every layer of that cryptographic stack — key generation, protocol negotiation, side-channel timing, quantum exposure, and trust chain integrity. Most organisations have no idea which of their AI systems are cryptographically vulnerable.
AI agent deployments generate ephemeral keys, API tokens, and session credentials constantly. Weak entropy sources, insufficient key lengths, and deprecated algorithms are pervasive — and invisible until they're exploited.
TLS 1.0, 3DES, RC4, and export-grade ciphers still exist in legacy stacks that AI agents connect to. Downgrade attacks silently negotiate weak cipher suites. Your agent thinks it's secure. It isn't.
Nation-state actors are harvesting encrypted AI agent traffic today. Post-quantum migration hasn't happened. KYBER and CRYSTALS-Dilithium aren't deployed. The countdown started without you.
Non-constant-time cryptographic implementations leak key material through measurable response time variations. Statistical analysis of timing deltas recovers secrets without breaking the algorithm itself.
Certificate pinning is absent. Certificate transparency log monitoring doesn't happen. Intermediate CA compromise goes undetected for months. Your trust chain has gaps you've never mapped.
AI agent plugins implement their own cryptographic routines — ECB mode, static IVs, reused nonces, unauthenticated encryption. Every plugin expands the vulnerable surface that CIPHER systematically enumerates.
Eight subsystems. Each one attacks a distinct layer of cryptographic infrastructure. From key generation weakness to quantum exposure to trust chain disruption — CIPHER covers the full cryptographic kill chain used against AI agent deployments.
| # | Subsystem | Command | What It Does | Clearance |
|---|---|---|---|---|
| 01 | KEYBREAK | cipher keybreak | Attacks key generation and management. Tests entropy quality, key length sufficiency, deprecated algorithm use, and key storage exposure in AI agent authentication flows. | Standard |
| 02 | DOWNGRADE | cipher downgrade | Executes protocol downgrade attacks against TLS/SSL stacks. Forces negotiation to TLS 1.0, 3DES, RC4, export-grade ciphers, and EXPORT-RSA. Maps the complete weak cipher surface. | Standard |
| 03 | KEYHARVEST | cipher keyharvest | Enumerates key material exposure vectors — environment variables, process memory leakage, log file contamination, and API response body leakage of secrets and tokens. | Standard |
| 04 | QUANTUM | cipher quantum | Identifies quantum-vulnerable cryptography in AI agent deployments. Enumerates RSA, ECC, DH key exchanges. Maps harvest-now-decrypt-later exposure. Flags missing CRYSTALS-Kyber/Dilithium migration. | Standard |
| 05 | TRUSTBREAK | cipher trustbreak | Attacks certificate trust chains. Tests missing pinning, CT log monitoring gaps, intermediate CA exposure, OCSP stapling failures, and cross-signed certificate substitution vectors. | Standard |
| 06 | TIMING | cipher timing | Statistical timing side-channel analysis against cryptographic operations. Measures response time distributions. Applies KS tests and Welch's t-tests to identify non-constant-time implementations leaking key material. | Elevated |
| 07 | HARVEST | cipher harvest | Passive and active cryptographic material collection. PCAP analysis for weak cipher negotiation, certificate fingerprinting, JWK endpoint enumeration, and JWKS rotation monitoring. | Standard |
| 08 | REPORT | cipher report | Aggregates all subsystem findings into a unified cryptographic risk report. NIST SP 800-131A mapping, FIPS 140-3 gap analysis, Ed25519 signed, RFC 3161 timestamped. AI Shield rule generation for crypto enforcement. | Standard |
Red Specter CIPHER is intended for authorised security testing and cryptographic assessment only. Timing side-channel analysis and protocol downgrade testing against systems you do not own or have explicit written permission to test may violate the Computer Misuse Act 1990 (UK), Computer Fraud and Abuse Act (US), and equivalent legislation in other jurisdictions. Always obtain written authorisation before conducting any cryptographic security assessment. Quantum vulnerability mapping of third-party infrastructure requires explicit scope approval. Apache License 2.0.