Proxies for Pentest Teams: Plan and Record Your Egress
Plan proxy egress for authorized assessments: which source address the client sees for each test, how it is recorded, and which product fits which check.
If you run authorized assessments, the proxy question is narrow: which source address will the client's systems see for each part of the test, and does your authorization cover it. Settle that in writing before configuring anything. A proxy changes the source address a target sees; it does not change what you are permitted to test, and an exit the client didn't expect can trip their incident response or invalidate an agreed whitelist.
The legitimate uses are specific: testing source-IP restrictions and geographic access rules, checking how defensive logging attributes traffic, and giving a distributed team a fixed, disclosed egress. Testing systems you are not authorized to assess is outside Node4's acceptable use policy whatever the routing.
The egress record in the engagement file
Every egress you plan to use belongs in the rules of engagement, next to scope and the testing window. For example, a record like this one, for an illustrative engagement, makes the routing auditable:
engagement: ACME-EXT-114
authorization: signed rules of engagement v2, client security lead named in section 1
scope:
in: [app.acme.example, api.acme.example]
out: [sso.identity-vendor.example, payments.psp.example]
window: "Oct 6 22:00 to Oct 7 04:00 UTC"
egress:
- id: E1
product: dedicated datacenter (static)
address: 198.51.100.17
country: US
used_for: baseline and authenticated testing
disclosed_to_client: yes, whitelisted on the WAF for the window
- id: E2
product: residential, country GB, one session per test case
address: varies, logged per request
used_for: check that the admin portal refuses non-GB sources (RoE 4.2)
disclosed_to_client: yes, as "GB residential, addresses in report"
rate_limit: 5 requests per second per egress
stop_condition: any 5xx spike on app.acme.example, or a call from the client contactThen make every request log line carry its egress ID and the exit address actually observed. That gives you two things. The client's security team can separate your traffic from a real attack during the window, and every finding that depends on source address ("the admin portal is reachable from outside the UK") comes with the evidence of where the request originated.
If a planned test needs an egress that isn't in the record, stop and get it approved. Source variation added after approval is how a proxy turns a clean engagement into a confused one.
Which egress fits which test
| Test | Egress | Why |
|---|---|---|
| Source-IP whitelist and baseline testing | Dedicated datacenter (details, prices) | One fixed address you can disclose in advance and the client can whitelist |
| Geographic access rule (US, Italy or Spain) | Datacenter exit in that country | Node4-owned IP blocks in those three countries |
| Geographic access rule (other countries) | Residential with a country segment | Partner network exits in 160+ countries |
| Does a session survive a source change? | Two pinned sessions, switched deliberately | You control exactly when the address changes |
| Non-HTTP tool traffic | SOCKS5 on a paid plan | TCP only: Node4's SOCKS5 implements CONNECT, with no UDP |
The last row has a practical consequence. Tools that send raw packets, run UDP scans or need inbound connections will not work through these proxies; they need a network path of their own, agreed with the client. The HTTP vs SOCKS5 guide lists exactly what each protocol carries.
Interception and upstream routing are different jobs. An intercepting proxy such as Burp or ZAP inspects and modifies your requests; an upstream proxy chooses the source address. Chain them (client to interception proxy to upstream proxy) and confirm that the interception tool's upstream setting, not the operating system's, is what routes the traffic.
Verifying the route before testing
Send one request to an endpoint you control and read its server log. The address in that log is the one the client will see; a configuration screen only tells you what you intended.
curl -q --noproxy '' --proto '=https' \
--proxy "http://USERNAME:PASSWORD@HOST:PORT" \
--connect-timeout 10 --max-time 30 \
--write-out 'http_code=%{http_code}\n' \
https://canary.your-team.example/egress-checkThe empty --noproxy stops curl from bypassing the proxy for hosts on a local exception list, and -q stops a personal .curlrc from changing the route. Repeat the check whenever you switch egress, and record the result against the egress ID. Keep proxy credentials out of shared notes and shell history; they belong in the team's secret store, not the evidence folder.
Session timing during authenticated tests
Rotating gateways assign a new exit to each HTTPS tunnel, and keep it for every request through that tunnel; plain-HTTP requests each get their own. A session ID in the username holds one exit across connections, which is what an authenticated test case needs.
The window is product-specific. On Rotating Datacenter and Rotating Unmetered plans, a session ID holds its exit for up to 30 minutes from first assignment; activity does not extend it, so a long authenticated workflow can change address partway. Residential sessions are best effort, with a 30-minute ceiling; if the device behind the exit goes offline first, requests on that session fail, so switch to a new session ID when they keep failing. For any test where an unplanned address change would confuse the result, use the dedicated static address (E1 above). Use pinned sessions only where the change itself is under test, and log the timestamp of each switch. The sticky versus rotating sessions guide covers how to spot an unplanned change.
Where to start
Draft the egress section of your next engagement using the record above, get it signed with the rest of the rules of engagement, and run the route check from each egress before the window opens. If the client wants a fixed address to whitelist, start with the smallest dedicated datacenter plan, $7.45 a month for 5 proxies, in the country they expect, and disclose only the address you will test from.