Security: Security Tiers

Security Tiers Reference

Wardex automatically adapts its security posture based on the value at risk. Small transactions pass through quickly; large operations get full scrutiny.

---

Tier Overview

Tier	ID	Value Range (USD)	Block Threshold	Human Required	Time Lock	On-Chain Proof
Audit	tier-0-audit	$0 - $1	Never (100)	No	No	No
Co-pilot	tier-1-copilot	$1 - $100	Never (100)	No	No	No
Guardian	tier-2-guardian	$100 - $10,000	Score > 70	On block only	No	No
Fortress	tier-3-fortress	> $10,000	Score > 30	Always	15 minutes	Yes
Freeze	frozen	Any	Everything	Required to resume	N/A	N/A

---

Tier 0: Audit

Range: < $1 USD (gas-only transactions, dust amounts)

Setting	Value
Mode	audit
Block threshold	100 (never blocks)
Human approval	No
Operator notification	No

Behavior: Logs the evaluation but never blocks. Even critical findings (prompt injection, denylisted address) are logged but not acted upon. This is a design choice — blocking dust transactions creates friction without preventing meaningful harm.

Use case: Gas-only operations, very small test transactions.

---

Tier 1: Co-pilot

Range: $1 - $100 USD

Setting	Value
Mode	copilot
Block threshold	100 (advisory only)
Human approval	No
Operator notification	No

Behavior: Full evaluation pipeline runs, but the verdict is always approve or advise — never block. Advisory verdicts include risk scores and reasons for operator review, but don't stop the transaction.

Exception: Critical findings (denylisted address, prompt injection) in Co-pilot tier are elevated to block by the policy engine. The blockThreshold of 100 applies to the composite score, but individual critical findings trigger an override.

Use case: Small trades, token swaps, regular DeFi operations.

---

Tier 2: Guardian

Range: $100 - $10,000 USD

Setting	Value
Mode	guardian
Block threshold	70
Human approval	On block only
Operator notification	Yes

Behavior: Full evaluation pipeline with enforcement. Transactions scoring above 70 are blocked. The operator is notified of all evaluations. Human approval is only required when a transaction is actually blocked (not for every evaluation).

What gets blocked at this tier:

• Composite risk score > 70
• Denylisted address (critical finding override)
• Detected prompt injection (critical finding override)
• SELFDESTRUCT contract (critical finding override)
• Unsafe delegatecall (critical finding override)

Use case: Medium DeFi operations, token transfers, contract interactions.

---

Tier 3: Fortress

Range: > $10,000 USD

Setting	Value
Mode	fortress
Block threshold	30
Human approval	Always
Time lock	900 seconds (15 minutes)
Operator notification	Yes
On-chain proof	Yes

Behavior: Maximum security. All transactions are blocked first and require explicit human approval. A 15-minute time lock provides a cooling-off period. The evaluation proof hash is recorded for on-chain submission.

Use case: Large operations, treasury movements, significant token approvals.

---

Freeze Mode

Trigger: 5 or more blocked transactions in the last 10 evaluations.

When Wardex detects a possible active attack (repeated blocks suggest someone is persistently trying malicious transactions), it automatically freezes all operations.

Setting	Value
All transactions	Blocked with decision: 'freeze'
Human required	Yes, must call wardex.unfreeze()
Callbacks	onFreeze and onThreat fire

Recovery: A human operator must call wardex.unfreeze() after investigating.

// Automatic freeze fires these callbacks
wardex = createWardex({
  // ...
  onFreeze: (event) => {
    sendAlert(`FROZEN: ${event.reason}`);
  },
});

// Manual freeze
wardex.freeze('Investigating suspicious activity');

// After investigation
wardex.unfreeze();

---

Critical Finding Overrides

Certain findings override the tier's normal block threshold:

Finding	Override Behavior
DENYLISTED_ADDRESS	Blocks at Guardian+ regardless of composite score
PROMPT_INJECTION	Blocks at Guardian+ regardless of composite score
SELFDESTRUCT_DETECTED	Blocks at Guardian+ regardless of composite score
DELEGATECALL_DETECTED	Blocks at Guardian+ regardless of composite score
INFINITE_APPROVAL	Blocks at Guardian+ regardless of composite score

These overrides do NOT apply at Audit or Co-pilot tiers (by design — dust transactions should not be blocked).

---

Customizing Tiers

Change Thresholds

const policy = defaultPolicy();

// Fortress at $5K instead of $10K
policy.tiers[3].triggers.minValueAtRiskUsd = 5000;

// Guardian blocks at 50 instead of 70
policy.tiers[2].enforcement.blockThreshold = 50;

Add Protocol-Specific Tiers

const uniswapTier: SecurityTierConfig = {
  id: 'uniswap-trading',
  name: 'Uniswap Trading',
  triggers: {
    targetAddresses: ['0x68b3465833fb72A70ecDF485E0e4C7bD8665Fc45'],
    maxValueAtRiskUsd: 50_000,
  },
  enforcement: {
    mode: 'guardian',
    blockThreshold: 80,
    requireHumanApproval: false,
    notifyOperator: true,
    requireOnChainProof: false,
  },
};

policy.tiers.splice(3, 0, uniswapTier);

ETH Price Considerations

Tier thresholds are in USD. At different ETH prices, the same amount of ETH falls into different tiers:

ETH Amount	At $2,000/ETH	At $3,500/ETH	At $5,000/ETH
0.001 ETH	$2 (Co-pilot)	$3.50 (Co-pilot)	$5 (Co-pilot)
0.05 ETH	$100 (Guardian)	$175 (Guardian)	$250 (Guardian)
1 ETH	$2,000 (Guardian)	$3,500 (Guardian)	$5,000 (Guardian)
5 ETH	$10,000 (Fortress)	$17,500 (Fortress)	$25,000 (Fortress)

---

What's Next?

• Custom Policies — Full policy customization guide
• Threat Model — What Wardex defends against
• Attack Vectors — The 7 attack categories