At 18:40 on Friday, an operations lead looks at the cash dashboard. A payment queue is larger than usual, an administrator account has just been used from a new location, and the ledger no longer matches the bank file. The week has been bad. The first question is not whether the platform is secure. It is whether anyone can move money, change the record or delay discovery while the team investigates.
Define what must be protected
Client funds security starts with a map of custody and control. List every place money or the authority to move it exists: bank and payment accounts, wallets where relevant, trading and back-office systems, payment processor consoles, treasury tools, and the credentials or keys used to access them. Include people and suppliers with administrative access.
For each asset, name an owner, the authorised actions, and the system that records the authoritative balance. Separate client money from operating funds in both account structure and ledger treatment. Technical controls support this separation, but do not replace legal segregation, reconciliation or jurisdiction-specific client money rules.
Draw the movement path from deposit to withdrawal. Mark where instructions are created, approved, sent and reconciled. A diagram that shows only the trading platform misses payment portals, spreadsheets and emergency access. Those less visible components often have broad permissions and weaker monitoring.
Map money paths and failure points
Trace a normal deposit, withdrawal, refund and correction. For each path, identify the data source, the person or service that can change the instruction, and the independent check before funds leave. The goal is to find single points of control, not to make every step manual.
- Keep payment initiation separate from approval. Use two authorised people for releases above a defined threshold, with separate credentials and no shared sessions.
- Restrict bank beneficiary changes to a controlled workflow. Verify changes through a known channel, not contact details supplied in the change request.
- Reconcile the internal ledger against bank or processor records at least daily. Alert on missing files, duplicate references and unexplained balance differences.
- Keep production payment credentials out of developer workstations, source repositories and general-purpose automation accounts.
Thresholds should reflect normal volumes and the cost of delay. A broker might require dual approval for every manual payment, or set a low threshold for routine withdrawals and a second threshold for exceptional transfers.
Document the rule, test that the system enforces it, and review exceptions monthly. More approval steps can slow a legitimate payout, so measure queue time as well as control failures.
Make access hard to misuse
Use phishing-resistant multi-factor authentication for privileged and payment roles where the systems support it. Hardware security keys or passkeys are stronger than one-time codes that can be phished. Remove standing administrator rights where practical. Grant just-in-time access for a defined task, time-limit it, and log both the request and approval.
Separate identity administration, payment approval and ledger administration. A single compromised account should not be able to change a beneficiary, approve the transfer and alter the record that would expose it.
Service accounts need named owners, narrowly scoped permissions, stored secrets in a vault and rotation after staff or supplier changes. Do not rely on password rotation as a substitute for limiting access.
Review privileged and payment access at least quarterly, and within one working day of a role change or departure. Disable dormant accounts after 30 days unless a documented exception applies.
Alert on new administrator creation, MFA removal, unusual login locations, access outside normal hours and changes to payment limits. Send logs to a separate account or environment so an administrator on the trading platform cannot quietly erase the evidence.
These controls apply across the connected stack, not just the execution platform. Trading Technology's broker infrastructure coverage tracks the systems and providers that commonly sit around it.
Make transactions fail safe
Build safeguards into the payment path. Use allow-listed beneficiary accounts, per-user and per-day limits, and a cooling period for new or changed destinations. A delay of 4-24 hours can give an operations team time to verify a high-risk change, though it may be unsuitable for every withdrawal flow. Define exceptions and record who authorised them.
Make payment requests idempotent: retrying a request after a timeout must not send a second payment. Use unique transaction references and reconcile the response from the bank or processor against the original instruction. Queue systems should have a clear state for pending, rejected, sent and confirmed payments. Operators must be able to stop a queue without deleting records or creating duplicate work when service resumes.
Protect data in transit with current TLS and encrypt sensitive records at rest. Keep encryption keys separate from the data they protect, restrict key administration, and alert on key use outside expected services.
Test a key rotation in a non-production environment before a scheduled production change. A rushed rotation can make records inaccessible even when no attacker is involved.
Detect incidents and recover cleanly
Monitoring should join payment events, identity events and ledger changes. Set alerts for a burst of withdrawals, a new beneficiary followed by a large payment, repeated failed approvals, disabled logging or a reconciliation mismatch. Route critical alerts to a monitored channel with a named on-call owner. A dashboard that no one is responsible for watching is not an alerting control.
Write a response procedure that lets staff freeze outbound payments, revoke sessions and credentials, contact banks or processors, preserve logs and continue essential trading operations safely. Define who can invoke each action and who can release the freeze. A practical initial target is to contain a suspected unauthorised payment path within 15 minutes, then investigate without deleting or overwriting evidence.
Keep backups encrypted and immutable, with credentials separate from production administration. Restore into an isolated environment and verify balances, permissions and audit records before reconnecting services. The figures below are typical market ranges in USD for a small-to-mid-sized broker or prop firm; scope, data volume and internal staffing can change costs materially.
| Control | Typical setup cost | Typical recurring cost | Scope notes |
|---|---|---|---|
| Independent penetration test | $15,000-$50,000 per engagement | Usually annual or after major changes | Cost varies with application count, API scope and retesting. |
| Managed detection and response | $5,000-$25,000 onboarding | $8,000-$30,000 per month | Coverage and log volume drive the range; confirm 24-hour response terms. |
| Immutable backup and recovery | $2,000-$15,000 setup | $1,000-$10,000 per month | Storage volume, retention and isolated recovery capacity affect cost. |
| Hardware security module or managed key service | $5,000-$40,000 setup | $1,000-$8,000 per month | Deployment model, key count and high-availability requirements matter. |
Run the controls as an operating routine
Security decays when ownership is unclear. Assign an accountable owner to each payment path and control, set a review date, and retain evidence of tests and exceptions. A short monthly review can cover unresolved reconciliation breaks, privileged access changes, payment-limit overrides, backup status and open security findings. Track failures as operational issues with deadlines, not as a list for an annual audit.
Use this checklist to establish a baseline. Record the result and the person responsible for any gap:
- Inventory every account, service and credential that can hold funds or initiate, approve or record a movement.
- Test that one person cannot independently change a beneficiary, release a high-value payment and alter its ledger evidence.
- Confirm that all privileged users have strong MFA, named accounts and time-limited access where supported.
- Run a daily reconciliation test using a deliberately missing, duplicated or mismatched transaction reference.
- Exercise the payment freeze procedure, including bank or processor contacts, at least twice a year.
- Restore a backup in isolation and verify both balances and audit logs before approving recovery.
Set recovery targets according to the ledger and payment service, rather than copying a generic platform target. For many firms, a ledger recovery point measured in minutes and a service recovery time measured in hours is a sensible starting discussion, but the acceptable interruption depends on payment cut-offs, market hours and the ability to reconcile offline. Test the targets with the teams who would have to meet them.
Check how many days have passed since the last successful isolated full restore. The number to beat is 30
Frequently Asked Questions
What is the most important security control for client funds?
Start by separating payment initiation, approval and ledger administration so one compromised account cannot complete and conceal a transfer. Pair that separation with strong MFA, transaction limits and independent reconciliation.
How often should a broker reconcile client funds?
A daily reconciliation is a practical baseline for bank and processor records against the internal ledger. Raise alerts for missing files, duplicates and unexplained differences, and define a named owner and resolution deadline for each break.
How often should a trading firm test its backups?
Test full restores into an isolated environment regularly, not just backup job completion. A 30-day interval is a useful operational target for many firms, with more frequent tests for systems whose loss would stop payments or reliable balance reporting.
Platforms · Software · Providers · Markets