Red Specter SPECTER SOCIAL

Autonomous Social Engineering Adversary — 6 channels. 10 attack types. 1,242 tests.

Overview

SPECTER SOCIAL is an autonomous social engineering adversary. Point it at a human target. It builds a psychological profile, designs a multi-channel campaign, deploys adaptive personas, and pursues its objective until it succeeds or exhausts all vectors.

Every existing social engineering tool is a phishing simulator — static email templates, no reasoning, no adaptation, no psychological profiling, no multi-channel coordination. SPECTER SOCIAL thinks. It uses an LLM reasoning engine to analyse the target, select attack vectors, craft personalised content, interpret responses, calibrate trust, and adapt its strategy in real time.

Six channels. Ten attack types. One reasoning engine that decides which channel to use, what persona to deploy, and how to decompose a malicious objective into a sequence of innocent-looking interactions. The human is the attack surface. SPECTER SOCIAL is the adversary.

Installation

Optional extras:

Quick Start

The Seven-Phase Campaign Loop

Every SPECTER SOCIAL campaign cycles through seven phases. The reasoning engine decides when to advance, pivot, escalate, or switch channels based on target responses and trust calibration.

Attack Channels

SPECTER SOCIAL coordinates attacks across six communication channels. The reasoning engine selects the optimal channel for each interaction based on target profile and campaign state.

Attack Types

Ten social engineering attack types, each with specialised content generation and persona management.

Human Target Model

SPECTER SOCIAL extends the NEMESIS reasoning engine with a Human Target Model — a psychological profile that evolves throughout the engagement.

CLI Reference

LLM Backends

SPECTER SOCIAL requires an LLM to reason about human targets. Three backends are supported:

UNLEASHED Mode

Standard mode probes and reports. UNLEASHED mode does not stop until the objective is achieved.

Ed25519 cryptographic override. One private key. Dual-gate: --override (dry-run) then --override --confirm-destroy (live).

UNLEASHED Modules

• SS-01 Credential Harvest — real credential capture via fake portals
• SS-02 Wire Transfer — live wire transfer social engineering
• SS-03 Deepfake Live — real-time voice cloning and video avatar
• SS-04 Multi-Persona — simultaneous fake identities for consensus manufacturing
• SS-05 Resistance Destroy — systematic resistance destruction across all channels
• SS-06 Psych Exploit — full psychological exploitation using all six influence techniques

Pipeline Integration

SPECTER SOCIAL completes the Red Specter offensive pipeline. Every other tool attacks infrastructure. SPECTER SOCIAL attacks the human. The last gap is closed.

Tool Integration

• GLASS — intercept target’s AI-mediated traffic, inject payloads into observed sessions
• FORGE — generate personalised lures and phishing content using the mutation engine
• NEMESIS — adaptive planning engine — SPECTER SOCIAL inherits the reasoning loop
• ARSENAL — probe any AI interfaces the target uses
• AI Shield — validate that AI Shield detects the social engineering vectors

Evidence & Cryptography

Report Output

Every campaign produces a full report: executive summary, interaction timeline across all channels, psychological vulnerability map, trust trajectory, findings with ATT&CK mapping, statistics, and recommendations.

• Executive summary — high-level campaign outcome and objective status
• Interaction timeline — every interaction across all channels, chronologically
• Psychological assessment — target vulnerability map, influence susceptibility, trust trajectory
• Findings — all vulnerabilities with MITRE ATT&CK mapping
• Trust trajectory — trust level progression across the campaign
• Statistics — reasoning loops, channels used, attacks deployed, success rates
• Recommendations — security awareness training, MFA, verification procedures, channel-specific mitigations

Disclaimer