Railways get hacked in five main ways. Attackers send commands over old radio systems that never checked who was talking, which is how trains were stopped in Poland in 2023 and in Taiwan in 2026. They hit the IT around the railway with ransomware, which takes down ticket machines, departure boards and planning systems. They compromise a supplier the railway depends on, which is how Danish trains stopped in 2022. They abuse station and onboard Wi-Fi and the providers that run it. And, rarely, the software inside the trains themselves turns out to contain behaviour nobody asked for, as researchers reported in Polish Newag trains in 2023. In nearly every documented case the trains stopped. They did not collide, because rail signalling is built to fail safe.
That is the short answer. This guide is the long one. It explains what software a modern railway runs, from interlockings and the European Train Control System (ETCS) to metro signalling, train radio, onboard computers, ticketing and station IT. It walks through the documented incidents technically, without guesswork about who was behind them. It covers CLC/TS 50701 and IEC 62443, and ends with hardening checklists for passengers, for operators and infrastructure managers, and for rolling stock and signalling makers, plus the software supply chain where a growing share of rail risk now sits.
It is written from the attacker's point of view for defenders. Each attack class explains how it works and why, what the attacker needs, real examples, the impact, how to detect it and how to defend against it. You will not find exploit code, radio parameters, tool command lines or step-by-step instructions.
A note on the law. Interfering with railway radio, signalling, trains or station systems that you do not own, or have no written permission to test, is illegal in most countries, and in many it is a serious criminal offence because it endangers the public. Everything here is for protecting railways you are responsible for, or the products you build for them. The researchers cited worked under contract, inside universities, or through coordinated disclosure.
This post is part of a device security series. The guides on how Wi-Fi gets hacked and how routers get hacked cover the networks that stations and trains depend on, the guide on how laptops get hacked covers the engineering laptops that touch depot and signalling equipment, and how cars and EV chargers get hacked covers the rest of the transport picture. Our posts on fabless chip designers and the software supply chain and how a laptop is built explain where the chips and SDKs inside rail controllers come from.
Contents
- How does a modern railway work, and what software does it run?
- Why can't a cyberattack simply crash a train?
- What is the attack surface of a railway?
- What does an attacker need to disrupt a railway?
- Which rail incidents should defenders learn from?
- How do radio commands stop trains?
- How secure are ETCS, ERTMS and GSM-R?
- How is CBTC in metros exposed?
- What happened with the Newag trains?
- How does ransomware hit rail operators and ticketing?
- How did a supplier's test environment stop Danish trains?
- How exposed are station Wi-Fi, displays and passenger systems?
- How should you weigh these risks?
- How do you detect attacks on a railway?
- What do CLC/TS 50701 and IEC 62443 require?
- Why is rail security a software supply chain problem?
- How should passengers protect themselves?
- How should operators and infrastructure managers secure a railway?
- How should rolling stock and signalling makers build secure products?
- How Safeguard helps teams that build and run rail software
- The takeaway
- Frequently asked questions
- Sources
How does a modern railway work, and what software does it run?
A modern railway is a stack of computer systems. Control centres route trains. Interlockings prove that each route is safe before a signal clears. Train protection such as ETCS or CBTC supervises speed. Radio links connect trains to the track. Each train carries its own network of controllers, and stations run ticketing, displays and Wi-Fi. Almost every layer runs software.
It helps to walk through the layers from the track up, because each has a different owner, a different lifespan and a different security story.
Figure 1. The five layers of a modern railway, and the main systems in each.
Signalling and interlocking
The oldest safety idea on the railway is the interlocking: an arrangement of signal apparatus that prevents conflicting movements through junctions and crossings. In the words of the standard definition, an interlocking "is designed so that it is impossible to display a signal to proceed unless the route to be used is proven safe" (Wikipedia, Interlocking). The points must be set and locked, and the track ahead must be clear, before the signal can show proceed. Interlockings have evolved from mechanical lever frames to relay logic and, since the 1980s, to solid state and computer based (electronic) interlockings. Signalling equipment, the same source notes, is expensive partly "because of its specialized nature and fail-safe design."
Around the interlocking sit the trackside devices it controls or reads: points machines that move switches, signals, train detection (track circuits or axle counters that tell the interlocking whether a section is occupied) and level crossings. In modern installations each of these is driven by an object controller, a small computer in a lineside cabinet. Above the interlocking sits signalling control, the signaller's workstation, and above that traffic management, which plans and regulates the timetable.
ETCS and ERTMS
The European Rail Traffic Management System (ERTMS) is the system of standards for interoperable signalling in Europe, and it has two main parts: the European Train Control System (ETCS) for signalling and train protection, and GSM-R for communication (Wikipedia, ERTMS). ETCS has trackside and onboard equipment. On board, the European Vital Computer (EVC) "monitors the train's compliance with Movement Authorities and other operational constraints, and intervenes if necessary to ensure safety by making an emergency brake application or otherwise overriding the driver's controls" (Wikipedia, ETCS). A Juridical Recording Unit logs what happened, much like a flight recorder.
On the track, Eurobalises are antenna devices mounted between the rails that send fixed data (gradients, speed limits) or switchable data from the signalling to passing trains. In ETCS Level 2, train movements are monitored continually by a Radio Block Centre (RBC), which the ETCS article describes as "a specialised computing device with specification safety integrity level 4 (SIL)" that generates movement authorities and sends them to trains over GSM-R. Safety integrity level 4 is the highest level of safety assurance in the railway safety standards. With Level 2, most lineside signals can be removed, and the cab display becomes the signal.
CBTC in metros
Metros mostly use communications-based train control (CBTC). The IEEE 1474 standard defines it as "a continuous, automatic train control system utilizing high-resolution train location determination, independent from track circuits; continuous, high-capacity, bidirectional train-to-wayside data communications; and trainborne and wayside processors" that provide automatic train protection, with optional automatic train operation and supervision (Wikipedia, CBTC). Trains report their exact position, speed and direction by radio, and wayside equipment tells each train the point it must never pass. That is what lets metros run trains closer together, and what makes the radio link central to the system.
GSM-R and FRMCS
GSM-R is a railway version of 2G mobile technology, specified around 2000, that carries voice between drivers and signallers and data for ETCS Level 2 (Wikipedia, GSM-R). Because GSM technology is approaching end of life around 2030, the rail sector is moving to the Future Railway Mobile Communication System (FRMCS), based on 5G New Radio, with dedicated frequency bands allocated in Europe in 2020 and rollout planned from 2025 onwards (Wikipedia, FRMCS). Many networks still also run older analogue radio, especially for shunting and on lines not yet converted.
Train control and management systems and onboard networks
Inside each train, a train control and management system (TCMS) coordinates traction, brakes, doors, heating and ventilation, lighting, diagnostics and the driver's displays. Many fleets use the train communication network (TCN) defined in IEC 61375, with a Multifunction Vehicle Bus (MVB) inside each vehicle and a Wire Train Bus (WTB) between vehicles; newer trains add an Ethernet Consist Network (ECN) and Ethernet Train Backbone (ETB) (Wikipedia, Train communication network). The controllers on these networks are programmable logic controllers and embedded computers running real-time operating systems or embedded Linux, plus application logic written by the train maker or its suppliers. Many trains also carry a cellular modem for remote diagnostics and fleet telemetry.
Passenger systems, ticketing and station IT
Then there is everything passengers see: passenger information systems driving displays, announcements and departure boards from real-time train data; CCTV; passenger Wi-Fi with its own cellular backhaul; and ticketing, from machines and gates to back office servers, booking sites and apps. Behind those sit ordinary enterprise systems, from email and ERP to maintenance management and the remote access suppliers use.
The thing to take from this tour is ownership. Infrastructure managers typically own the track, interlockings and RBCs, operators own or lease the trains, makers wrote most of the safety software and support it for decades, and ticketing, Wi-Fi and IT are often run by yet more providers. Every boundary is a place where security responsibility can fall through.
Why can't a cyberattack simply crash a train?
Rail signalling is designed so that any fault, doubt or lost message drives the system to its most restrictive state. An interlocking will not clear a signal unless it can prove the route is safe, train protection brakes a train that exceeds its authority, and the safe state of a train is stopped. So most attacks on rail systems end in stopped trains and lost money, not collisions.
This is the single most important fact for keeping rail cybersecurity in proportion, and it is worth understanding why.
Figure 6. Why attacks on rail safety systems usually trade availability for safety rather than causing collisions.
The engineering principle is called fail-safe. A relay interlocking is wired so that a broken wire or a lost power supply drops the relay into the position that shows a red signal. A computer based interlocking typically runs its vital logic on redundant hardware that must agree before anything moves, and treats missing or malformed input as unsafe. ETCS onboard equipment applies the brakes if it loses its movement authority or cannot trust its data. GSM-R and CBTC radio links are monitored, and when communication is lost the train slows or stops. On top of the technology sit operating rules and trained staff: signallers can authorise movements by written order, drivers can proceed at caution, and whole routes can be worked by phone if the electronics fail.
That design creates an asymmetry that shapes every incident in this guide. An attacker who interferes with safety systems usually triggers the safe state. Trains stop. Timetables collapse, passengers are stranded, freight is late and the operator loses money and reputation. Those are real harms, but they are availability harms. To cause a collision, an attacker would need to defeat the vital logic and the train protection and the human checks at the same time, and those layers were designed by people who assumed that equipment, software and humans all fail.
There are two honest caveats. First, the layers can be thin. Tram networks and some older or secondary lines have less protection than a mainline with ETCS, and the 2008 Lodz tram incident later in this guide shows what that means. Second, availability matters a great deal. A railway that cannot run is a serious problem for a city or a country, so "only" stopping trains is not a reason for complacency. It is a reason to aim defences where attackers can actually hurt you: the non-vital systems, the radio, the IT and the suppliers that keep trains running.
What is the attack surface of a railway?
A railway's attack surface spans remote entry points such as booking sites, corporate IT and supplier remote access; radio entry points such as legacy analogue radio, end of train links, GSM-R, CBTC radio and Wi-Fi; and physical entry points such as depot laptops, onboard ports, trackside cabinets and ticket machines. Most real incidents came through remote and radio paths.
Figure 2. Every way into a railway, grouped by how close the attacker has to be.
Remote, over the internet. Booking sites, mobile apps and their APIs are internet facing by design. Corporate email and IT are where most ransomware starts. Supplier and maintainer remote access is the bridge from the internet into operational systems, and it is often the least visible. Cloud tools that staff rely on, from rostering apps to the app Danish drivers used for safety information, are part of the surface even though the railway does not host them. Hosted portals, such as the landing page for station Wi-Fi, are run by providers with their own admin accounts.
Within radio range. Railways use a lot of radio. Legacy analogue systems carry voice and, in some countries, emergency stop signals. Freight trains use radio links between the locomotive and the device on the last wagon. GSM-R carries ETCS data. CBTC trains talk to wayside radios, often in licence free bands. Stations and trains offer Wi-Fi. An attacker in range can listen, jam or, where messages are not authenticated, transmit.
Physical access or insider. Depots are full of laptops and diagnostic ports used to maintain trains. Trains have network ports and modems. Lineside cabinets hold object controllers and network equipment. Ticket machines and gates sit in public spaces. USB media and engineering tools move between networks. Physical access is harder to get and easier to trace, which is why it features less often in incidents, but it is how targeted tampering would most likely happen.
What does an attacker need to disrupt a railway?
Most documented rail disruptions needed very little: cheap radio equipment and public knowledge for legacy radio commands, or ordinary ransomware and an unpatched server for IT outages. Deeper attacks on signalling or train software need insider knowledge, engineering access or a foothold in a supplier, which is why they are rarer.
| Attack class | What the attacker needs | Typical impact | Documented example |
|---|---|---|---|
| Legacy radio commands | Inexpensive radio, knowledge of a published or recovered signal, being within range | Unplanned stops, delays for hours | Poland 2023, Taiwan 2026 |
| Unauthenticated train links | Software defined radio, protocol knowledge, proximity | Unwanted braking commands | CVE-2025-1727 (research) |
| Ransomware on IT and ticketing | Phishing or an unpatched internet facing server | Ticket machines, boards, planning down for days | Muni 2016, Deutsche Bahn 2017, TTC 2021 |
| Supplier compromise | A foothold in a provider's environment | Operations halted if the supplier's service stops | DSB and Supeo 2022 |
| Wi-Fi and portal abuse | Provider admin credentials or a misconfigured database | Defacement, data exposure | Network Rail 2024, C3UK 2020 |
| Train software manipulation | Access to the software build or the vehicle | Trains that will not start | Newag 2023 (reported, disputed) |
| Signalling compromise | Engineering access, insider knowledge, bypassing vital logic | Bounded by fail-safe design; mostly stops | No public mainline case |
The pattern matters. The cheapest attacks hit the systems that were never designed with an adversary in mind: analogue radio from an era when only railway staff had transmitters, and IT systems that grew up separately from the safety world.
Which rail incidents should defenders learn from?
The incidents worth studying are the 2008 Lodz tram points, the 2016 San Francisco Muni ransomware, WannaCry at Deutsche Bahn in 2017, the 2022 Danish halt caused by a supplier, the 2023 Polish radio-stop events, the 2023 Newag research, the 2024 Network Rail Wi-Fi defacement, and the 2026 Taiwan High Speed Rail radio alarm. Each teaches a different lesson.
Figure 3. Documented rail incidents and research, 2008 to 2026.
The rest of this guide takes them by attack class rather than by date, because the defences group that way. Two items on the timeline are not attacks at all. The 2017 EuroRadio paper is academic research, and the June 2026 GSM-R outage in Germany had, according to reporting at the time, no evidence of a cyberattack. Both are included because they show where the dependencies are.
How do radio commands stop trains?
Several rail radio systems accept commands without checking who sent them. In Poland in 2023, unauthorised "radio-stop" signals on the analogue train radio halted about 20 trains. In Taiwan in 2026, a student imitated the high speed railway's radio signalling and triggered an emergency alarm that stopped four trains. The fix is authenticated radio, plus monitoring and direction finding.
Poland, August 2023
On the night of 25 to 26 August 2023, unauthorised "radio-stop" signals were broadcast in the West Pomeranian region around Szczecin, and the Polish infrastructure manager PKP PLK reported delays on more than 20 connections (Notes from Poland). Further unauthorised signals were reported elsewhere over the weekend, and on the Sunday police detained two men, aged 24 and 29, in Białystok, accused of using the signal three times (The Record). PKP PLK said the situation did not pose a threat to passengers and affected only train schedules.
How it works, conceptually. Poland's railway used an analogue VHF radio system with a RADIOSTOP function. As Hackaday explained, transmitting a basic tone sequence makes any equipped train receiving it apply emergency braking (Hackaday). The researcher Łukasz Olejnik told Wired that the system "lacks encryption or authentication for those commands," so anyone with inexpensive radio equipment could trigger it (Schneier on Security, quoting Wired). CERT Polska noted that the analogue design is deliberate: any train or post can signal an emergency stop to all trains within a small radius (The Record).
Why it works. This was a safety feature doing exactly what it was designed to do. If a driver sees an obstruction, being able to stop every nearby train instantly, without a login, is valuable. The design assumption was that only railway staff would have the equipment and the knowledge. Both assumptions stopped being true long ago, and Notes from Poland reported that the instructions were public.
What the attacker needs and the impact. Radio range, inexpensive equipment and public knowledge; some commentators argued it should not be called hacking at all. The result was stopped trains and hours of delay, with no injuries, because the emergency stop put trains into their safe state.
Detection and defence. Hackaday reported that Poland planned to move to GSM-R by 2025, keeping the old radio only for shunting and edge cases. In the meantime the practical defences are monitoring the radio channel for stop commands that do not match any incident report, direction finding to locate unauthorised transmitters quickly, and procedures that let signallers confirm or clear a stop fast. Long term, any command that can stop a train should be authenticated, or at least attributable to a known radio.
Taiwan High Speed Rail, April 2026
During the Qingming holiday in April 2026, four Taiwan High Speed Rail trains were stopped for nearly an hour after a general emergency alarm was triggered on the railway's radio system. Security Affairs reported that a 23-year-old university student and radio enthusiast had used radio equipment and software tools bought online to imitate the railway's own radio signalling, and that local reports said the same system parameters had been in use for 19 years without being changed (Security Affairs).
Two details stand out for defenders. First, staff noticed quickly that the alarm did not match any assigned radio device and concluded it came from an unauthorised source. That is good detection: the system knew which radios were legitimate. Second, investigators traced the transmitter using station CCTV and radio network logs and recovered handheld radios, a software defined radio receiver and a laptop. Logging made attribution possible. The weakness was parameters that never rotated, so once recovered they worked indefinitely.
End of train and head of train links: CVE-2025-1727
Long freight trains in North America use a device on the last wagon, the end of train (EoT) unit, linked by radio to the head of train (HoT) unit in the locomotive. The CVE record for CVE-2025-1727, published in July 2025 and credited to Neil Smith and Eric Reuter, says the remote linking protocol "relies on a BCH checksum for packet creation" and that it is possible to create these packets with a software defined radio "and issue brake control commands to the EoT device, disrupting operations or potentially overwhelming the brake systems." It is scored CVSS 3.1 8.1 (High), with an adjacent network attack vector, and classed as CWE-1390, weak authentication.
Why it works. A checksum detects accidental corruption. It does not prove who sent a message. Anyone who knows the format can compute a valid checksum. The CVE record notes that the Association of American Railroads is pursuing new equipment and protocols to replace the traditional devices. Until then, the defence is operational: awareness, monitoring for unexpected brake commands, and procedures for crews.
Lodz trams, January 2008
The oldest case is a reminder of what happens when the safety layers are thin. In January 2008, a 14-year-old in Lodz, Poland, allegedly modified a TV remote control so it could change tram track points, derailing four trams; twelve people were injured in one incident (The Register). Police said he had trespassed in tram depots to gather the information he needed. The points evidently answered to a wireless signal that a homemade transmitter could imitate, with nothing like a mainline interlocking proving the route before the tram moved. The lesson is not that trains are fragile. It is that safety depends on the interlocking and protection layers, and systems without them need compensating controls.
How secure are ETCS, ERTMS and GSM-R?
ETCS Level 2 protects safety messages with three layers: GSM-R radio encryption, a EuroRadio message authentication code, and checks in the ETCS application. Research in 2017 found weaknesses in all three, including a MAC collision attack, but judged it not an immediate danger at current data volumes. The bigger practical risks are jamming, key management and the move to FRMCS.
Figure 5. The three protocol layers between an ETCS Level 2 train and the Radio Block Centre.
The three layers
In ETCS Level 2, the train's Euroradio unit keeps two simultaneous GSM-R connections to the RBC (Wikipedia, ETCS). GSM-R provides encryption between the train and the base station, as any GSM network does. On top of that, the EuroRadio protocol adds a message authentication code (MAC), a cryptographic checksum computed with a secret key, so that the RBC and the train can each verify that a message came from the other and was not altered. The ETCS application messages ride on top: movement authorities, position reports, emergency messages. The RBC, the ETCS article notes, "receives cryptographic keys from trains passing in," and keys are distributed by a key management process run by the railway.
What the research found
In 2017, Tom Chothia, Mihai Ordean, Joeri de Ruiter and Richard J. Thomas, from the University of Birmingham and Radboud University, published a cryptographic analysis of these protocols at ACM AsiaCCS (University of Birmingham). Their abstract says GSM-R "has the same well known weaknesses as the GSM protocol," presents "a new collision attack against the EuroRadio protocol," and shows that, combined with design weaknesses in the application protocol, an attacker who observes a MAC collision could forge train control messages. They demonstrated it with train control messages they generated themselves (OpenAlex record).
The same abstract is careful about the risk. Because ERTMS sends small amounts of data in short sessions, "this attack does not present an immediate danger." They calculated that an attacker monitoring all backend control centres in a country the size of the UK for 45 days would have about a 1 percent chance of being able to take control of a train, and warned that the picture changes if EuroRadio carries much more data in future. That is exactly the kind of result defenders want early: a design weakness found by researchers, with time to fix it in the next generation.
Jamming and dependency
The easier attack on any radio based signalling is jamming, and its effect is the safe state. Lose the radio link and ETCS Level 2 trains brake to a stop. The scale of that dependency was shown, without any attacker, on 23 June 2026, when a nationwide GSM-R failure led Deutsche Bahn to hold all trains across Germany at stations. Reporting at the time found no evidence of a cyberattack or physical damage, and the cause had not been publicly disclosed (Security Affairs). The same article notes that Deutsche Bahn has contracted for a 5G FRMCS replacement. Nothing about that night was malicious, but it shows what an attacker who could take down GSM-R would achieve: a halt, not a crash, and a very expensive one.
What FRMCS changes
FRMCS moves rail communications to 5G, which brings modern cryptography and mutual authentication at the radio layer, more bandwidth, and IP based services. It also brings an IP network with far more software in it, more applications sharing the link, and in some countries the option of using public mobile networks alongside dedicated ones. The FRMCS article notes Finland's interest in using all three public network operators across its track (Wikipedia, FRMCS). Each of those choices is reasonable. Each also adds software, suppliers and interfaces that need the security engineering this guide describes.
Defences for operators. Treat key management for EuroRadio as a security process, with documented key generation, distribution, rotation and revocation. Monitor GSM-R for interference and unexplained loss of connections along the line. Plan and rehearse degraded working for radio loss. For FRMCS projects, write security requirements into the tender from day one rather than retrofitting them.
How is CBTC in metros exposed?
CBTC depends on continuous radio between trains and wayside equipment, and many systems use licence free bands at 2.4 GHz and 5.8 GHz shared with Wi-Fi. Interference or jamming forces trains into degraded modes, and weak segmentation could expose zone controllers. The defences are dedicated spectrum where possible, authenticated and encrypted links, and practised recovery procedures.
There are fewer public CBTC incidents to learn from than mainline ones, so this section is conceptual. The CBTC article notes that with services emerging over "open ISM radio bands (i.e. 2.4 GHz and 5.8 GHz)" and the potential disruption of critical CBTC services, there is pressure in the international community to reserve spectrum for urban rail (Wikipedia, CBTC). The same article points out that one of the major hazards in CBTC is human error in applying recovery procedures when the system becomes unavailable.
How it is exposed. CBTC radio is a data network carrying position reports and movement limits. If it is disrupted, trains fall back to slower, manually supervised modes, and service capacity drops sharply. If it is poorly segmented from other networks in the metro, the wayside zone controllers and automatic train supervision servers become reachable from places they should not be. Station networks, CCTV and passenger Wi-Fi often share physical infrastructure in tunnels and stations, which makes logical separation essential.
What the attacker needs. For disruption, proximity and a transmitter. For anything beyond that, network access to the signalling segment, which is why segmentation and authentication matter more than the radio itself.
Defences. Use dedicated or protected spectrum where regulators allow it, and spread the radio design so that one source of interference cannot take out a line. Require authentication and encryption on train to wayside links, with keys managed like any other critical credential. Keep CBTC networks in their own zones, separate from station and passenger networks, with monitored conduits. And train staff on recovery procedures until they are routine, because the CBTC article is right that recovery is where the real risk sits.
What happened with the Newag trains?
In 2023, the Polish research team Dragon Sector, hired by an independent maintenance company, reported that software in Newag Impuls trains contained logic that locked trains with bogus error codes after idle periods, after certain dates, or when GPS showed them at third-party workshops. Newag strongly denied introducing such software. The case shows why operators need visibility into the software in the trains they buy.
What was reported
In May 2022, the independent train maintenance company Serwis Pojazdów Szynowych (SPS) hired three members of the Dragon Sector team, Jakub Stępniewicz, Sergiusz Bazański and Michał Kowalczyk, to investigate Newag Impuls 45WE trains that would not start after being serviced in third-party workshops (The Register). Kaspersky's summary of their later talk says the operator Koleje Dolnośląskie (Lower Silesian Railways) had found that several trains would not start after SPS serviced them, even though diagnostics showed no faults (Kaspersky).
The team reverse engineered the trains' control software by comparing firmware from trains that had locked with firmware from trains still running. Bazański wrote that "the PLC [programmable logic controller] code actually contained logic that would lock up the train with bogus error codes after some date, or if the train wasn't running for a given time," and that one version contained GPS coordinates matching third-party workshops (The Register). Kaspersky's summary lists several mechanisms: a lock if a train spent more than 10 days in certain geographic areas that matched independent repair shops, a lock triggered by a change in a part's serial number, a date based trigger, and the most common, a lock after a train had been parked for a set number of days. One train had a device labelled UDP<->CAN Converter connected to a GSM modem that received lock status from the onboard computer (Kaspersky). Of about 30 trains from several operators that the team examined, 24 contained such mechanisms. The researchers also reported an undocumented key combination in the cab that cleared the lock, and Wikipedia notes that a later software update removed that ability (Wikipedia, Newag).
The team presented at the Oh My H@ck conference in Warsaw in December 2023, and in detail at the 37th Chaos Communication Congress, in a talk titled "Breaking 'DRM' in Polish trains" that covered how a train control and monitoring system works and how they analysed its TriCore machine code (media.ccc.de). CERT Polska confirmed to The Register that the team had disclosed their results to it more than a year earlier.
The manufacturer's position
Newag "emphatically denied that it installed such software," attributing any issues to unknown hackers and calling the claims "a slander from our competition, which is conducting an illegal black PR campaign against us" (The Register). Wikipedia records that Newag said there was no proof it was the author of the software or of the modifications, and that it would take legal action against SPS and the researchers (Wikipedia, Newag). We describe the technical claims as reported, and note that the manufacturer disputes them.
Why this matters for every railway
Set aside who wrote the code. Four lessons apply to any fleet.
- Operators often cannot see inside the trains they own. A railway can buy a fleet, run it for years and not know what logic its controllers contain, or what conditions can stop it.
- Availability is a security property. A train that refuses to start is out of service whether the cause is malice, a bug or a commercial dispute.
- Remote links into trains need governance. A modem that reports lock status can, in principle, carry other commands. Every off-board link should be documented, authenticated and owned.
- Independent analysis has value. It took reverse engineering to find out what the software did. Contracts that give operators access to software documentation, SBOMs and the right to independent review would make that easier.
Detection. Trains that fail after maintenance with no physical cause, especially in patterns tied to location, dates or idle time, deserve a software investigation, not just a hardware one. Diagnostics logs, controller parameter dumps and comparisons between units are the starting point.
Defence. Write software transparency into rolling stock contracts: full documentation of every condition that can inhibit traction or doors, SBOMs for every controller, a statement that no undocumented lock or remote functions exist, escrow or access arrangements for long-term support, and the right to commission independent security review.
How does ransomware hit rail operators and ticketing?
Ransomware hits the IT around the railway, not the signalling: ticket machines, departure boards, booking, planning and staff systems. San Francisco's Muni gave free rides in 2016, WannaCry took down Deutsche Bahn departure boards in 2017, and Toronto's transit commission lost platform screens and apps in 2021. Segmentation, patching and offline backups are the defences.
Ransomware is the most common real-world cyber threat to rail operators, and it almost never touches train control. What it does is take away the systems that let a railway sell tickets, inform passengers and plan its day. The cases are consistent.
San Francisco Muni, November 2016. Riders found handmade "Out of Service" and "Metro Free" signs on station ticket machines after ransomware hit the San Francisco Municipal Transportation Agency, and terminals displayed "You are Hacked. ALL Data Encrypted" (Krebs on Security). The attacker demanded 100 Bitcoin. Krebs's reporting suggested the agency was not specifically targeted: the attacker scanned the internet for vulnerable systems, most commonly exploiting a deserialisation flaw in Oracle WebLogic server products for which Oracle had issued a patch in November 2015.
Deutsche Bahn, May 2017. In the global WannaCry outbreak, around 450 Deutsche Bahn computers were infected, causing departure boards at many stations, video surveillance systems and a regional control centre in Hanover to fail (Wikipedia (German), WannaCry). WannaCry spread through a Windows networking flaw that had been patched before the outbreak. Display systems running old, unpatched Windows on flat networks were exactly what it found.
ADIF, July 2020. Spain's railway infrastructure manager, responsible for track, signalling and stations, confirmed a ransomware attack in which attackers claimed to have stolen 800 GB of data. ADIF said "the infrastructure has not been affected at any time" (Security Affairs).
Merseyrail, April 2021. The UK operator confirmed a cyber attack after the attackers used a compromised director's email account to announce it to staff and journalists, and claimed to have stolen employee and customer data (Security Affairs).
Toronto Transit Commission, October 2021. Ransomware disrupted internal email, the Wheel-Trans booking service, vehicle information on platform screens, apps and the TTC Vision communication system, and operators were forced to communicate by a conventional radio system (Security Affairs). That fallback to radio is a good example of a rehearsed degraded mode doing its job.
Transport for London, September 2024, and Keio, September 2026. TfL's attack mainly affected back office systems at its headquarters, with no impact on services at the time (Security Affairs). Keio Corporation, a Tokyo private railway, confirmed a ransomware attack on 26 September 2026 that disrupted business systems and led it to shut down its network while investigating (Security Affairs).
DDoS on booking. Not every outage needs malware. In February 2026 a large distributed denial of service attack disrupted Deutsche Bahn's information and booking systems, including bahn.de and the DB Navigator app, for several hours (Security Affairs).
Why ransomware works on railways
Rail operators run large estates of Windows machines that are hard to patch: ticket machines, display controllers, station servers and depot PCs built into equipment with long lifecycles. Many of those devices sit on networks that were flat when they were installed. And railways are attractive to extortion because every hour of outage is visible to the public.
Detection and defence
Detect mass file encryption or renaming, new scheduled tasks and services across many machines, unusual use of admin tools, and large outbound data transfers, which usually precede encryption. Defend by separating ticketing, passenger information and station systems from corporate IT and from each other; patching internet facing systems first; requiring multi factor authentication on email and remote access; keeping offline, tested backups and known good images for ticket machines and display controllers so they can be rebuilt quickly; and rehearsing degraded operation, from paper tickets to manual announcements. Our guide to the first hour of an OT cyber incident covers response steps that translate well to rail.
How did a supplier's test environment stop Danish trains?
On 29 October 2022, Danish trains stopped for hours because Supeo, a supplier whose app gave drivers safety-critical information, shut down its systems after criminals compromised its test environment. DSB itself was not attacked. The lesson is that a supplier's outage can halt operations even when the operator's own systems are untouched.
On the morning of Saturday 29 October 2022, trains across Denmark stopped. DSB, the largest train operator, explained that the cause was an error in a safety-critical IT system, the Digital Backpack 2 (Den Digitale Rygsæk 2), which lets train drivers access operationally critical information on an iPhone or iPad. The app was provided by Supeo, a company that supplies enterprise asset management software to railway operators and infrastructure managers (Security Affairs).
DSB's head of security, Carsten Dam Sonderbo-Jacobsen, said: "We were contacted by our subcontractor who told us that their testing environment had been compromised by criminal hackers." He told Reuters, "It hasn't targeted infrastructure or DSB, it was economic crime" (Bitdefender). Supeo shut down its servers in response, drivers lost access to the information they needed, and DSB stopped trains until about 1 p.m., with long-distance services affected into the next day.
Why it happened. Nothing in DSB's network was compromised, and nothing in the signalling was touched. A safety procedure required drivers to have certain information, the information lived in a supplier's service, and the supplier turned that service off to contain an attack on a different environment. The safe decision, stopping trains, was the right one. But a single supplier had become a single point of failure.
What it teaches.
- Map operational dependencies on suppliers, including apps and cloud services, not just equipment. Ask which supplier outages would stop trains.
- Test environments are production for attackers. They are often less protected and share credentials, code or networks with live systems. Suppliers should segment and monitor them as carefully as production, and operators should ask about it.
- Plan for the supplier going dark. For each critical service, know the fallback: cached data on the device, a paper procedure, an alternative channel to drivers.
- Contract for it. Incident notification times, the right to know what is affected, and continuity requirements belong in supplier contracts.
How exposed are station Wi-Fi, displays and passenger systems?
Station Wi-Fi and passenger systems are often run by third parties and are internet facing. In 2020 a station Wi-Fi provider left a database of about 10,000 users' travel details exposed online, and in 2024 an unauthorised change from a provider's administrator account made station Wi-Fi landing pages across the UK display an offensive message. Neither touched train operations.
C3UK, 2020
In March 2020 the BBC reported that the email addresses and travel details of about 10,000 people who used free Wi-Fi at UK railway stations had been exposed online. The provider C3UK confirmed that a backup database, containing 146 million records, had been left on Amazon Web Services storage without password protection; it was found by researcher Jeremiah Fowler (BBC News). Fowler said the data appeared searchable by username, which could reveal individuals' regular travel patterns, and that it also showed the software types of connecting devices. C3UK described it as a low-risk potential vulnerability and secured it.
Network Rail, 2024
In late September 2024, passengers logging into Wi-Fi at 19 Network Rail managed stations, including London Euston, Liverpool Lime Street, Edinburgh Waverley and Glasgow Central, saw a landing page with a message about terror attacks in Europe (ITPro). The Wi-Fi service was run by Telent, and Telent said that "an unauthorised change was made to the Network Rail landing page from a legitimate Global Reach administrator account," Global Reach being the provider of the landing page. Telent said no personal data was affected, and the service was suspended while it was fixed.
Why it works. Captive portals and their management consoles are web applications run by specialist providers. One admin account at the provider can change what passengers see at every station that provider serves. It is a supply chain issue in miniature.
Displays and announcements
Passenger information systems turned up in several ransomware cases above: WannaCry took out Deutsche Bahn departure boards, and the TTC lost vehicle information on platform screens. Displays are attractive targets for disruption and for defacement because the public sees them. They are usually Windows or Linux PCs or embedded players fed from a central content system, which makes that content system the thing to protect.
Defences
For passengers, the advice is in the checklist later in this guide. For operators and providers: hold Wi-Fi and portal providers to the same access controls you would apply internally (named admin accounts, multi factor login, change approval and alerting on portal changes); collect only the data the service needs and keep it encrypted and access controlled; keep passenger Wi-Fi on a separate network with its own backhaul, never bridged to operational or onboard control networks; and protect display content systems with change control, so an unapproved message is caught before it reaches a screen. Our Wi-Fi security guide covers the wireless side in depth.
How should you weigh these risks?
Fix first what is cheap for attackers and costly for you: ransomware on IT, ticketing and information systems; supplier compromise; and legacy radio commands. Deep attacks on signalling are rarer and bounded by fail-safe design. Wi-Fi and DDoS are mostly reputational. Cryptographic weaknesses belong in the design of the next generation.
Figure 7. Rail attack classes placed by attacker effort and likely impact on service.
The placement in Figure 7 is a judgement from the documented cases, not a scored model. Three points shape it. First, the incidents that have actually cost railways the most time were ransomware and supplier outages, and they needed no rail knowledge at all. Second, the radio cases show that a cheap attack can stop trains when a legacy system trusts anything it hears. Third, research on EuroRadio and on train protection shows real design weaknesses, but the researchers themselves assessed them as hard to exploit today, and fail-safe design bounds the outcome. That points budgets at segmentation, backups, supplier access, radio monitoring and the migration away from unauthenticated radio, while new designs build in authentication from the start.
How do you detect attacks on a railway?
Watch five places: the radio spectrum for alarms or stops that match no assigned radio; control and signalling networks for new devices and unplanned changes; trains and depots for faults after maintenance with no physical cause; supplier remote access for logins outside agreed windows; and IT, ticketing and passenger systems for mass encryption, unapproved content and unusual exports.
Figure 10. Where to look for signs of attack on a railway, and the signals that should page someone.
Radio spectrum. The Taiwan case shows the value of knowing which radios are legitimate: staff spotted an alarm that matched no assigned device. Railways can monitor emergency and stop channels, log every command with its source where the system allows it, and use direction finding to locate unauthorised transmitters. For GSM-R and CBTC, monitor interference levels and unexpected connection losses along the line.
Control and signalling networks. Signalling networks are stable, which makes anomalies easy to see. Passive network monitoring, built for industrial protocols, can learn what normal looks like and flag a new device, a new connection, or an engineering command outside an approved work order. Configuration and firmware baselines for interlockings, object controllers and RBCs let you detect unplanned changes.
Onboard and depot. The Newag case is a detection lesson: a pattern of trains that failed after maintenance, with no physical cause, was the signal. Log diagnostics sessions and tie them to job tickets. Watch onboard modem traffic for unexpected destinations or volumes.
Supplier and remote access. Every supplier session into operational systems should be logged, recorded and tied to an approved job. Alert on logins outside agreed windows, new or reactivated admin accounts, and access from unexpected locations.
IT, ticketing and passenger systems. Use endpoint detection on Windows estates that can run it, alert on mass encryption or renaming, and put change detection on display content systems and Wi-Fi portals so that unapproved content triggers an alert rather than a passenger's photo on social media.
Detection only pays off if someone acts on it, so agree in advance who decides to isolate a system and when to fall back to manual working.
What do CLC/TS 50701 and IEC 62443 require?
CLC/TS 50701 is CENELEC's 2021 technical specification for railway cybersecurity. It adapts IEC 62443-3-3 and the EU common safety method for risk assessment to rolling stock, signalling and fixed installations, and covers risk assessment, security design, assurance, vulnerability management and patch management. IEC 62443 supplies the zones and conduits model and security levels it builds on.
CLC/TS 50701
CENELEC announced CLC/TS 50701, "Railway applications, Cybersecurity," in June 2021, developed by its committee CLC/TC 9X for electrical and electronic applications for railways (CEN-CENELEC). According to the announcement, it aims to handle cybersecurity "in a unified way for the railway sector," and it applies to the communications, signalling and processing domain, to rolling stock and to fixed installations. It takes safety aspects into account through EN 50126, the railway standard for reliability, availability, maintainability and safety, and draws on IEC 62443-3-3 and the common safety method for risk assessment (CSM-RA), adapting them to rail.
The topics it covers, per CEN-CENELEC, are:
- a railway system overview and cybersecurity across the application life cycle;
- risk assessment;
- security design;
- cybersecurity assurance and system acceptance;
- vulnerability management;
- security patch management.
It is a technical specification, not a law; it becomes binding when written into a tender or contract. Our explainer on what CLC/TS 50701 asks of suppliers turns those topics into a supplier gap checklist.
IEC 62443
IEC 62443 is the international series for the security of industrial automation and control systems. Three ideas from it carry straight into rail.
Zones and conduits. Assets with common security requirements are grouped into zones; conduits group the elements that allow communication between zones and provide the security functions that control it (Wikipedia, IEC 62443). IEC 62443-3-2 sets out how to derive a target security level for each zone and conduit from a risk assessment, including partitioning by separating business and control system assets, safety-related assets, temporarily connected devices, wireless devices and devices connected via external networks.
Security levels. Requirements are graded in four security levels, from SL 1, protection against unintentional or accidental misuse, to SL 4, protection against intentional misuse with sophisticated means, extensive resources and high motivation.
Secure development. IEC 62443-4-1 defines a secure product development lifecycle for suppliers, including how they handle vulnerabilities and publish updates, and IEC 62443-4-2 sets technical requirements for components. Our post on IEC 62443-4-1 goes into the eight practice areas.
Figure 9. An illustrative zones and conduits model for a train.
Figure 9 maps those ideas onto a train, from the safety zone around ETCS and brake control down to the internet facing passenger zone. The conduits are where the security work happens: fixed, validated interfaces into the safety zone, one-way data from control to displays, an authenticated and encrypted tunnel from the gateway to the operator's back office, and no conduit at all from the passenger zone to anything that moves the train. Our post on IEC 62443 roles, zones, conduits and safety explains how to run that process.
Why is rail security a software supply chain problem?
A railway runs software written by signalling suppliers, rolling stock makers, ticketing and passenger information vendors, IT providers and the open source projects inside all of them. The Danish halt, the Newag case, the Network Rail Wi-Fi defacement and the LNER and Ferrovie dello Stato data exposures all started outside the railway's own systems.
Figure 8. Who writes the software a railway runs, and where documented incidents started.
Look back at the incidents. The Danish trains stopped because of a supplier's test environment. The Newag questions arose in software delivered with the trains. The Network Rail message came from a portal provider's admin account. In 2025, LNER reported unauthorised access to files managed by a third-party supplier, involving customer contact details and some past journey information, with no impact on train operations or ticket sales (Security Affairs). In the same year, data belonging to Italy's Ferrovie dello Stato group was leaked after a breach at its IT provider Almaviva (Security Affairs).
The chain
- Open source and components. Real-time operating systems, embedded Linux, bootloaders, network stacks, cryptographic libraries and the SDKs that come with processors sit inside interlockings, object controllers, onboard computers, ticket machines and display players.
- Product makers. Signalling and ETCS suppliers, rolling stock and TCMS makers, and ticketing and passenger information vendors integrate those components with their own application code.
- Integrators and service providers. Depot maintainers, IT and managed service providers, and Wi-Fi and portal operators configure, host and support the products.
- Railway duty holders. Train operators and infrastructure managers run the result and carry the responsibility for safety and service.
Why it is hard
Rail software lives for decades, outliving several generations of the open source components inside it. Safety approval makes change expensive, so updates are batched and slow. Products arrive as compiled firmware and images, often without an SBOM. And responsibility is spread across contracts written before cybersecurity requirements were common.
What good looks like
- SBOMs for every product and every release, delivered by suppliers and kept current, so that when a vulnerability is announced in a library an operator can find every affected interlocking, train and ticket machine in minutes. Our post on SBOMs for railway signalling machines and firmware covers what to ask for.
- Vulnerability management that understands safety. Not every vulnerability in a safety system needs an emergency patch. Reachability and exposure decide urgency, and compensating controls (segmentation, monitoring) can carry the risk until the next approved release. Our guide to patching OT and ICS safely explains the trade-offs.
- Supplier security clauses covering development practices, disclosure, incident notification, remote access, test environments and long-term patch support.
- Independent verification of what is delivered, including binary analysis of firmware and images where source is not available.
Our guide to railway signalling supply chain risk to safety goes deeper on the signalling side.
How should passengers protect themselves?
Passengers face little direct risk from rail cyber incidents beyond delays. The practical steps are to treat station and train Wi-Fi as public networks, use official apps and booking sites only, never enter payment details on a captive portal, check odd notices on screens or portals against official channels, and report strange behaviour to staff.
Figure 11. What each group can do this quarter.
- Treat station and onboard Wi-Fi as public. Rely on encrypted connections (HTTPS, a trusted VPN) rather than on the network, and turn off file sharing. The Wi-Fi guide in this series explains why.
- Book through official channels. Use the operator's own app or site, or a well-known retailer. Phishing that follows rail data breaches can use journey details to look convincing, which is why LNER customers were told to watch for it after the 2025 incident.
- Never enter card details on a captive portal. A Wi-Fi landing page has no reason to ask for payment information unless you are knowingly buying access from a provider you recognise.
- Check odd messages. If a departure board, app or Wi-Fi page shows something strange, check the operator's official channels before acting on it.
- Report it. Tell staff about unusual behaviour on screens, ticket machines or Wi-Fi pages. Operators want to know.
During an outage, follow staff instructions. Trains stopping is the system working as designed.
How should operators and infrastructure managers secure a railway?
Start with an asset and software inventory, then segment operational systems with the Purdue model and IEC 62443 zones and conduits, lock down supplier remote access, monitor networks and radio, keep offline tested backups and known good images, rehearse manual fallback, and write security requirements, SBOMs and incident notification into every supplier contract.
1. Asset and software inventory
You cannot protect what you cannot list. Build and keep a live inventory of every system in the railway: interlockings, object controllers, RBCs, CBTC equipment, radio infrastructure, onboard controllers by fleet and unit, ticket machines, display controllers, servers and network devices. For each, record the supplier, the software and firmware versions, the components inside (from SBOMs or binary analysis), the network location, the owner and the support status. Our post on building an OT asset inventory covers practical methods, including passive discovery that is safe on live networks.
2. Segmentation on the Purdue model, zones and conduits
Figure 4. Where railway systems sit in the Purdue model, with passenger Wi-Fi outside and supplier access through the buffer zone.
The Purdue model is a reference architecture from the 1990s that organises industrial systems into levels, from the physical process at Level 0 up to enterprise systems at Levels 4 and 5 (Wikipedia, Purdue Enterprise Reference Architecture). Mapped to rail, as in Figure 4, points, signals, train detection, brakes and traction are Level 0; interlockings, object controllers and onboard ETCS and TCMS are Level 1; signaller workstations, RBCs and CBTC zone controllers are Level 2; traffic management and the control centre are Level 3; and planning, maintenance management, booking and email sit above a buffer zone (often called a DMZ) between IT and operational technology.
The rules that follow are simple to state. No direct path from Levels 4 and 5 to Levels 0 to 2. All exchange between IT and operations goes through the buffer zone, using jump hosts, one-way gateways where data only needs to flow outward, and staging servers for patches and antivirus updates. Passenger Wi-Fi sits outside entirely. Then refine with IEC 62443 zones and conduits, which let you group by security requirement rather than strictly by level, as Figure 9 did for a train. Our posts on the Purdue model and where it falls short and how to segment an OT network go into implementation.
3. Secure remote access
Suppliers need access to support their systems, and that access is one of the most common paths into operational networks across industries. Make it narrow: named individual accounts, no shared vendor logins; multi factor authentication; access through a jump host in the buffer zone, never direct; sessions approved per job and time bound; sessions recorded; and accounts disabled when the contract or the person leaves. Our list of questions to ask ICS vendors and service providers about remote access is a good starting point for those conversations.
4. Monitoring
Deploy passive monitoring on signalling and control networks, endpoint detection where the estate supports it, and radio monitoring, all feeding a security operations function with a direct line to the control centre.
5. Backups and known good images
Keep offline, tested backups of configuration for interlockings, RBCs, CBTC and network equipment, and known good images for ticket machines, display controllers and station servers, so a ransomware outage becomes a rebuild measured in hours rather than weeks. Test restores on a schedule. Store copies where ransomware cannot reach them.
6. Incident response and manual fallback
Write and rehearse playbooks for the scenarios in this guide: loss of a supplier service, ransomware in ticketing or passenger information, radio interference, suspicious behaviour on a fleet. Define who can decide to isolate a system and when to move to manual working. Practise paper ticketing, manual announcements and degraded signalling procedures, as the TTC's fallback to radio in 2021 shows is worth doing.
7. Supplier governance
Put security into every contract: secure development practices aligned with IEC 62443-4-1, SBOMs per release, vulnerability disclosure and notification timelines, remote access rules, test environment segregation, continuity of critical services, long-term patch support, and the right to independent review. Ask the Danish question for every supplier: if this company went dark tomorrow, which trains would stop?
8. Legacy radio
Where analogue or unauthenticated radio functions can stop trains or issue commands, plan their retirement or authentication, and in the meantime add monitoring, logging of command sources, direction finding and fast procedures to confirm or clear an unexpected stop.
How should rolling stock and signalling makers build secure products?
Build to IEC 62443-4-1 and CLC/TS 50701 from the first design review: authenticate every message that can affect movement, document every condition that can inhibit a train, deliver an SBOM with every release, sign and verify updates, run coordinated vulnerability disclosure, and support patches for the whole life of the fleet.
Design
- Authenticate anything that can stop, start or steer. The Polish, Taiwanese and end of train cases all came down to commands accepted without proof of origin. New designs should use message authentication with managed, rotatable keys on radio links and, where practical, on onboard buses carrying safety-relevant commands.
- Zone the train. Apply the zones and conduits model in Figure 9 at design time, with the passenger network physically or cryptographically isolated from control and safety zones.
- Govern off-board links. Every modem and telemetry path should be documented, authenticated in both directions, limited to the functions it needs, and visible to the operator.
- No hidden behaviour. Document every condition that can inhibit traction, doors or brakes, and every remote function. Operators should never discover them by reverse engineering.
Development
- Secure development lifecycle. IEC 62443-4-1 sets out the practices: security requirements, secure design, secure coding, security testing, vulnerability handling and update management.
- Know your components. Generate an SBOM for every controller, image and application at build time, including transitive dependencies and the processor vendor's SDK.
- Scan what you write and what you ship. Static analysis, software composition analysis and secrets scanning on source; binary analysis on delivered images, including supplier components you only receive compiled.
- Protect signing keys. Firmware and update signing keys are among the most valuable assets a rolling stock or signalling maker holds. Keep them in hardware security modules with strict access and logging.
Support
- Coordinated vulnerability disclosure. Publish how researchers can report issues, respond quickly, and credit them.
- Patch for the fleet's life. Plan how security updates will be produced, safety assessed and delivered for 30 years, including when a component's upstream project has moved on.
- Give operators what they need. SBOMs, vulnerability statements explaining whether a given CVE is exploitable in your product, and clear guidance on compensating controls while an update goes through approval.
How Safeguard helps teams that build and run rail software
Safeguard covers the software supply chain inside rail products in one platform: SBOMs for firmware, images and apps, binary and image scanning, vulnerabilities ranked by reachability, SCA, SAST and secrets scanning on code you own, fix pull requests gated by your CI, Gold patched packages, supplier risk and compliance evidence, deployed as SaaS, in your VPC or on premises.
Safeguard does not sit on the GSM-R network, in the interlocking or on the train bus, and it does not replace OT network monitoring. It sits where much of the long-lived exposure in this guide comes from: the firmware, operating systems, open source components and applications that signalling suppliers, rolling stock makers, ticketing vendors and railway IT teams build and maintain. Instead of one tool for source scanning, another for binary analysis, another for SBOMs, a supplier questionnaire tool and a GRC platform, Safeguard does that work end to end, with one inventory and one vulnerability model, from the first SBOM to the merged fix and the audit record.
Figure 12. How Safeguard fits a railway software supply chain, from inputs to fixes and evidence.
See inside the binaries and images you ship and receive
In the Safeguard console, Integrations > Upload App Binary accepts compiled software: Windows EXE, MSI and DLL files, Android APKs and iOS IPAs, VM and disk images in QCOW2, VMDK, VHD and OVA formats, and container tarballs. Safeguard unpacks the file, identifies embedded libraries and their versions, maps them to known vulnerabilities and records licences, and the result becomes a project with a full SBOM. For rail teams that covers passenger and staff apps, Windows software on ticket machines, display controllers and engineering workstations, VM images for station and control centre servers, and the containers behind booking services. Raw firmware images for embedded controllers may need to be extracted into their component files first, and the usual limits of binary analysis apply: heavy obfuscation, encryption or unusual packing can give incomplete results. Our guide to scanning app binaries with Safeguard and our explainer on binary composition analysis go further.
Rank vulnerabilities by what can actually be reached
A controller image or station server can contain hundreds of components, and many vulnerabilities in them sit in code your product never calls. Safeguard's reachability analysis ranks the vulnerabilities your code can actually reach above those it cannot, and adds EPSS (the Exploit Prediction Scoring System) and known exploited context. For rail, where every change to safety-related software has a cost, that ranking is what tells you which vulnerability justifies an out-of-cycle release and which can wait for the next approved baseline behind compensating controls.
Scan the code you own
For signalling applications, TCMS logic, ticketing systems, passenger information software, apps and backend APIs that you write yourself, Safeguard runs software composition analysis (SCA) across the full transitive dependency tree, static application security testing (SAST) on your own code, and secrets scanning that catches signing keys, API tokens and hard-coded credentials before they ship. A leaked firmware signing key turns one product flaw into a fleet-wide one, so catching secrets in repositories early matters.
SBOMs your customers and auditors can use
Every project produces an SBOM you can export in CycloneDX or SPDX, among other formats: what infrastructure managers and operators ask suppliers for, and what lets a team answer "which of our products ship this library version" in minutes. Safeguard's Third-Party Risk Manager lets you request SBOMs from your own suppliers, score them and monitor them continuously, so the supplier side of the chain is visible too. See SBOMs for firmware and IoT devices for the details.
Fix it without waiting for upstream
This is where Safeguard differs most from a scanner. For code in your repositories, Griffin, Safeguard's remediation model, opens a fix pull request that your own CI builds and tests before anyone merges it, and your policies decide what can merge. For open source components where upstream has not shipped a fix, or where the fix sits in a version you cannot move to without a larger change, Safeguard Gold provides patched, hardened builds of those packages, and Safeguard can fork and patch the vulnerable package so your next build pulls a fixed version. For rail products with long lives and slow approval cycles, that matters: a minimal, targeted fix to the component you already ship is far easier to assess and approve than a jump to a new major version. Closed source components still have to come from their vendor, but the open source around them no longer has to wait.
Compliance evidence from the same data
Safeguard's compliance suite collects evidence for frameworks such as SOC 2 and ISO 27001 from the same scans, policies and fix history. For suppliers answering CLC/TS 50701 or IEC 62443 requirements in a tender, the same SBOMs, vulnerability records and fix trail support the vulnerability and patch management answers.
Deploy where your code lives
Signalling source code, train software and signing infrastructure are sensitive. Safeguard runs as SaaS, in your own VPC, or fully on premises. It can run in Zero AI mode with no language model in the loop, or with a model you choose.
How this compares
Rail organisations often combine an OT network monitoring platform such as Nozomi Networks or Claroty, a rail-specific security platform such as Cylus, a source SCA tool such as Snyk or Black Duck, and a GRC tool. Those are capable products, and live monitoring of signalling and train networks is a job Safeguard does not do. Safeguard's case is consolidation and remediation across the software itself: one inventory across source, binaries and images, reachability-ranked vulnerabilities, fix pull requests and patched packages instead of tickets, and supplier risk and compliance evidence from the same data. It works alongside a monitoring platform. For more context, see critical infrastructure software supply chain security and our firmware supply chain security guide.
The takeaway
Rail cybersecurity attracts dramatic headlines, but the record is calmer than they suggest. In the documented cases, trains stopped. A tone on an old radio channel did what it was designed to do. A forged alarm did the same. Ransomware took away ticket machines and departure boards. A supplier's outage left drivers without information they needed, so trains waited. Signalling did its job and failed safe every time.
That is not a reason to relax. It is a map. The exposure sits in the systems that were never designed with an adversary in mind, in legacy radio that trusts anyone, in IT estates that are hard to patch, and above all in the chain of suppliers whose software and services a railway now depends on. The defences are known: inventory, segmentation, controlled supplier access, monitoring, backups, rehearsed fallback, authenticated designs and real visibility into the software inside every train and every interlocking. Railways that do those things will keep their worst days to delays.
Try it yourself
Click through these interactive Safeguard demos to see the steps described in this guide.
Test yourself
Why can't a cyberattack simply crash a train?
Rail signalling is designed so that any fault, doubt or lost message drives the system to its most restrictive state. An interlocking will not clear a signal unless it can prove the route is safe, so most attacks end in stopped trains and lost money, not collisions.
What happened in Poland in 2023 with train radio?
Unauthorised radio-stop signals on the analogue train radio halted about 20 trains. The radio system accepted commands without checking who sent them, and the fix is authenticated radio plus monitoring and direction finding.
Why did Danish trains stop on 29 October 2022 even though DSB was not attacked?
Supeo, a supplier whose app gave drivers safety-critical information, shut down its systems after criminals compromised its test environment. A supplier outage can halt operations even when the operator's own systems are untouched.
What did ransomware hit in cases like San Francisco Muni and Deutsche Bahn?
Ransomware hits the IT around the railway, not the signalling. Muni gave free rides in 2016, and WannaCry took down Deutsche Bahn departure boards in 2017.
What was the Newag trains case about?
In 2023 the research team Dragon Sector reported that software in Newag Impuls trains contained logic that locked trains with bogus error codes. Newag strongly denied introducing such software. The case shows why operators need visibility into the software in the trains they buy.
What is CLC/TS 50701?
It is CENELEC's 2021 technical specification for railway cybersecurity. It adapts IEC 62443-3-3 and the EU common safety method for risk assessment to rolling stock, signalling and fixed installations.
Frequently asked questions
Can a train be hacked?
Yes, parts of a railway can be attacked, and documented incidents show trains being stopped by unauthorised radio commands and rail IT being hit by ransomware. But mainline signalling is built to fail safe, so the realistic outcome of an attack on train control is that trains stop. The more common harm is disruption to ticketing, passenger information and planning systems.
Can hackers cause a train crash?
It is very unlikely on a modern mainline. Interlockings will not clear a signal unless the route is proven safe, train protection such as ETCS brakes a train that exceeds its authority, and lost data or radio drives systems to stop. An attacker would need to defeat several independent safety layers at once. Simpler systems, such as the Lodz trams in 2008, have fewer layers.
How did hackers stop trains in Poland in 2023?
Poland's analogue train radio had a "radio-stop" function: a short tone sequence makes equipped trains apply emergency braking, with no authentication, so any train or post can stop nearby trains in an emergency. In August 2023 unauthorised signals using that function stopped about 20 trains near Szczecin and others elsewhere. Two men were detained. No one was hurt.
What happened with the Newag trains and Dragon Sector?
In 2023, Dragon Sector researchers hired by the independent maintainer SPS reported that software in Newag Impuls trains locked them with bogus error codes after idle periods, on certain dates or when GPS placed them at third-party workshops. They presented the work at 37C3. Newag strongly denied introducing such software and said it would take legal action.
Is station and train Wi-Fi safe to use?
Treat it as a public network. Use HTTPS or a trusted VPN, avoid sensitive tasks if you can, and never enter payment details on a captive portal. Incidents include a 2020 exposure of station Wi-Fi users' travel data and a 2024 defacement of UK station Wi-Fi landing pages through a provider's admin account. Neither affected trains.
How does ransomware affect railways?
Ransomware usually hits IT rather than signalling: ticket machines, departure boards, booking, email and planning. San Francisco's Muni gave free rides in 2016, WannaCry took down Deutsche Bahn departure boards in 2017, and Toronto's transit commission lost platform screens and apps in 2021. The defences are segmentation, patching, multi factor login, offline backups and rehearsed manual working.
What is CLC/TS 50701?
CLC/TS 50701, "Railway applications, Cybersecurity," is a CENELEC technical specification published in 2021. It applies to signalling, rolling stock and fixed installations, adapts IEC 62443-3-3 and the common safety method for risk assessment to rail, considers safety through EN 50126, and covers risk assessment, security design, assurance, vulnerability management and patch management. It becomes binding through contracts.
How does IEC 62443 apply to railways?
IEC 62443 provides the zones and conduits model, four security levels and secure development requirements that rail cybersecurity builds on. Operators use it to segment signalling, train and passenger networks and control traffic between them. Suppliers use IEC 62443-4-1 for secure development and 4-2 for component requirements. CLC/TS 50701 adapts it to rail.
Sources
- The Record, Two suspects arrested following Poland railway hack
- Notes from Poland, Policeman among those arrested over disruption of Poland's train system
- Hackaday, Polish railways fall victim to cheap radio attack
- Schneier on Security, Remotely stopping Polish trains
- Security Affairs, Taiwan High-Speed Rail emergency braking hack
- CVE Program, CVE-2025-1727
- The Register, Polish teen derails tram after hacking train network
- Chothia, Ordean, de Ruiter and Thomas, An Attack Against Message Authentication in the ERTMS Train to Trackside Communication Protocols, ACM AsiaCCS 2017, and its OpenAlex record
- Security Affairs, One railway radio outage stopped trains across Germany
- Kaspersky, Hacking a train: a 37C3 talk
- media.ccc.de, Breaking "DRM" in Polish trains
- The Register, Polish train maker denies claims its software bricked rolling stock
- Wikipedia, Newag
- Security Affairs, A cyberattack blocked the trains in Denmark
- Bitdefender, Denmark's train network frozen due to cyberattack on subcontractor
- Krebs on Security, San Francisco rail system hacker hacked
- Wikipedia (German), WannaCry, and Wikipedia, WannaCry ransomware attack
- Security Affairs, ADIF infected with ransomware
- Security Affairs, Merseyrail hit by ransomware
- Security Affairs, Toronto Transit Commission hit by a ransomware attack
- Security Affairs, Transport for London dealing with an ongoing cyberattack
- Security Affairs, Deutsche Bahn hit by a DDoS attack
- Security Affairs, Keio Corporation and Tokyo Metro disclose security breaches
- Security Affairs, LNER discloses a data breach
- Security Affairs, Ferrovie dello Stato data leak via Almaviva
- BBC News, Rail station wi-fi provider exposed traveller data
- ITPro, Network Rail confirms cyber attack on Wi-Fi systems at UK train stations
- CEN-CENELEC, A major step for railways cybersecurity: the new CLC/TS 50701
- Wikipedia, IEC 62443
- Wikipedia, Purdue Enterprise Reference Architecture
- Wikipedia, European Train Control System
- Wikipedia, European Rail Traffic Management System
- Wikipedia, GSM-R
- Wikipedia, FRMCS
- Wikipedia, Communications-based train control
- Wikipedia, Interlocking
- Wikipedia, Train communication network
- Cylus, homepage; Nozomi Networks, homepage; Claroty, homepage