Cybersecurity

Penetration testing: test your defences before someone else does

A penetration test does not draw up a list of theoretical flaws: it shows, with supporting evidence, what a third party can actually obtain. We carry it out on a written scope, once a letter of authorisation has been signed.

Network rack: switches and tidy cables, a single orange cable among the others (computer-generated image).

The subject

Why, and for whom

Three points of entry put to the test: two hold, the third gives way. That is the finding, with its evidence.

A configuration audit tells you what ought to be a problem. A penetration test goes one step further: it actually attempts to get in, to reach data or to take control of a service or a device, in order to establish what is possible and how far it goes.

It covers what is online just as much as what can be held in the hand or travels over the air: an application, a shop-floor network, a circuit board, an enclosure, a radio link.

It is a supervised exercise. Gaining access to a system without authorisation is a criminal offence: nothing therefore starts until a signed letter of authorisation has set out the targets, the dates, the hours and the limits. The tests are conducted so as not to interrupt your services.

  • Software vendors and companies that put an application, a customer portal or an API online.
  • Industrial companies whose PLCs and shop-floor networks are now connected to the information system.
  • Manufacturers and integrators of electronic or communicating equipment who want to know what their hardware gives away to anyone who has it in their hands or is within radio range.
  • Subcontractors asked by a customer, an insurer or a call for tenders for a recent penetration test report.
  • Companies that want to check that a fix or a new version has not reopened a breach.

Scope

What we test

Web and mobile applications
Authentication, sessions, access rights, handling of user input: the defects that make it possible to see or change what should not be seen or changed, in the browser and in the mobile application alike.
  • HTTP / HTTPS
  • WebSocket
  • OAuth 2
  • OpenID Connect
  • SAML
  • JWT
  • Android
  • iOS
APIs
The interfaces used by your mobile applications and by your partners, often less well protected than the website itself.
  • REST
  • SOAP
  • GraphQL
Exposed services and network
Remote access, email, administration interfaces, directories: what a third party can reach from the Internet or from your internal network, and what they can do with it.
  • SSH
  • FTP / SFTP
  • TLS
  • DNS
  • SNMP
  • SMB
  • RDP
  • LDAP
  • Kerberos
  • Active Directory
  • IPsec
  • OpenVPN
  • WireGuard
  • 802.1X
  • NTP
Delivery pipeline and containers
Code repositories, integration and deployment pipelines, images and orchestrators: what access to the delivery pipeline makes it possible to reach in production.
  • CI/CD
  • Docker
  • Kubernetes
Industrial control systems and networks
PLCs, gateways and supervisory systems: protocols designed for closed networks, often without authentication, and now connected to the rest of the company.
  • Modbus RTU / TCP
  • EtherNet/IP (CIP)
  • PROFINET
  • PROFIBUS
  • OPC UA
  • S7
  • BACnet
  • DNP3
  • IEC 60870-5-104
  • IEC 61850
  • MQTT
Fieldbuses and embedded networks
The buses that link control units, sensors and actuators inside a machine or a vehicle: what a device connected to the bus can read, inject or disrupt.
  • CAN
  • CANopen
  • J1939
  • UDS
  • LIN
  • embedded Ethernet
  • RS-232
  • RS-485
Hardware and firmware
Circuit boards, maintenance ports, memories and internal buses: what a person who has the device in their hands can read, extract or modify, down to the firmware and its boot chain.
  • UART
  • JTAG
  • SWD
  • SPI
  • I2C
  • USB
  • eMMC / NAND / NOR
  • U-Boot
  • secure boot
  • TPM
  • signed updates
  • embedded Linux
  • microcontrollers
  • FPGA
Side channels and fault injection
What a component leaks through its power consumption or its emissions, and what a disturbance triggered at the right moment can make it accept.
  • power analysis
  • electromagnetic emissions
  • fault injection
Short-range radio
The wireless links of your equipment, from common protocols to proprietary links analysed with software-defined radio: what can be listened to, replayed or spoofed without any contact with the device.
  • Wi-Fi (802.11)
  • Bluetooth / BLE
  • Zigbee
  • Thread (802.15.4)
  • Z-Wave
  • NFC
  • RFID 125 kHz, 13.56 MHz, UHF
  • 433 / 868 MHz
  • software-defined radio (SDR)
Mobile networks, long-range links and positioning
The modems and receivers that connect a device to its operator: what they accept from the network, and what a falsified signal would lead them to believe.
  • 2G / 3G / 4G / 5G
  • NB-IoT
  • LTE-M
  • LoRa / LoRaWAN
  • Sigfox
  • TETRA
  • DMR
  • GNSS (GPS, Galileo)
Microwave frequencies
Above one gigahertz: microwave links, satellite links, radars and the upper wireless bands. Links long considered safe because they are hard to intercept, and which still often travel unencrypted.
  • microwave links
  • satellite links
  • radar
  • 5 and 6 GHz bands
  • 60 GHz
Methodological frameworks
The tests rely on public frameworks: the findings remain comparable from one test to the next, and readable by your customers as well as by your auditors.
  • OWASP (web, API, mobile, IoT)
  • IEC 62443
  • ETSI EN 303 645
  • NIST
  • CIS
Two levels of knowledge
With no prior information at all, like any third party, or with what an ordinary user has – an account, a user manual, a unit of the product – to measure what a customer or an employee could obtain.
Retest
After your fixes, we replay the findings to confirm that they are closed.

Deliverables

What you receive

  • An executive summary: what was obtained, and what it means for the company.
  • A technical report: for each finding, the path followed, the evidence, the severity and the recommended fix.
  • An oral debrief for your technical teams.
  • A retest report after the fixes.

Process

How the engagement runs

  1. Scoping

    Targets, objectives, level of knowledge and schedule are settled together; for a device, the number of units made available.

  2. Letter of authorisation

    Signed before any test: it lists the targets, the dates, the limits and the emergency contacts.

  3. Tests

    Carried out on the authorised scope only, at the agreed times and in the agreed places.

  4. Debrief

    Handover of the reports and presentation of the findings to your teams.

  5. Retest

    Once the fixes have been made, we check that the findings are closed.

Frequently asked questions

What we are asked most often

What is the difference between a penetration test and an external exposure audit?

The exposure audit identifies and checks what is visible from the outside, across your entire perimeter. The penetration test concentrates on a specific target and seeks to go all the way: get in, reach data, obtain privileges. The first draws the map, the second tests one point.

Can the tests interrupt our services?

They are designed to avoid it: no saturation testing, no radio jamming, no destructive action on a live system, agreed hours and an emergency contact on each side. The most sensitive tests can be carried out on a pre-production environment.

What do we need to provide for a test on hardware?

Units of the device in the version to be tested and, if you wish, its documentation. Some checks – opening an enclosure, working on a board, reading a memory – can leave a unit unusable: they are carried out only with your written agreement, on units supplied for that purpose.

How are radio tests carried out?

On your equipment, at a place agreed during scoping. Most of the work consists of listening to and analysing the exchanges. Tests that involve transmitting on a band subject to authorisation – satellite navigation, mobile networks – are carried out only over a cabled link or in a shielded enclosure, never over the air.

Can an industrial installation be tested while in operation?

With caution: a PLC can stop on a single unexpected request. Active tests are preferably carried out on a mock-up, on a test bench or during a planned shutdown. On an installation in operation, we confine ourselves to listening and analysis, unless you give your written agreement.

How does it differ from the connected product security review?

The review supports the design: it starts from the architecture and the documentation, and leads to recommendations for your development teams. The penetration test starts from the finished device and looks for what a third party can obtain from it. The two complement each other.

What happens if you discover a serious flaw during the test?

You are informed without waiting for the report, through the channel agreed during scoping, with what is needed to fix it urgently.

How often should a test be repeated?

After each major change to the application or the infrastructure, and at regular intervals for the most exposed services. Some customers and some frameworks require an annual test.

How is the price set?

At a fixed price, once the scope has been defined. You receive a quote; payment is by bank transfer on invoice.

Let’s talk about what you need.

Describe your situation in a few lines: we will come back to you with a scoping proposal and a fixed-price quote.