Cars and EV chargers get hacked in four main ways. Attackers abuse the cloud services and apps that send commands to cars, which is where most remote car flaws of the last few years have been found. They trick keyless entry into thinking the key is close, which is how many keyless cars are stolen. They break into infotainment and telematics units and try to reach the networks that control the car. And they attack chargers, the firmware inside them and the backends that manage them. Almost every documented case was fixed by an update from the company that wrote the software, and most of the risk sits in that software supply chain.
That is the short answer. This guide is the long one. It explains what software runs in a modern car and an EV charger, and where an attacker can reach it. It walks through the research that shaped car security, from the 2015 Jeep Cherokee remote attack by Charlie Miller and Chris Valasek to Sam Curry's automaker API research and the 2024 Kia web portal flaw. It covers keyless relay theft and CAN injection theft at a conceptual level, the infotainment and Bluetooth flaws found in head units, and three years of Pwn2Own Automotive results. It then turns to chargers: the Open Charge Point Protocol (OCPP), charger firmware CVEs, the 2026 wave of charging backend flaws, and attacks on the charging cable link itself. It ends with the rules that now govern car software (UN R155, UN R156 and ISO/SAE 21434), hardening checklists for owners, fleets and makers, and the supply chain inside every car and charger.
It is written from the attacker's point of view for defenders. You will find how each attack works and why, what the attacker needs, the impact, how to detect it and how to defend against it. You will not find exploit code, tool command lines or step-by-step instructions. Where a tool or a contest is named, it is named as context for authorised testing.
A note on the law. Testing cars, chargers or their cloud services that you do not own, or have no written permission to test, is illegal in most countries. Everything in this guide is for protecting your own vehicles and chargers, your organisation's fleet and charging sites with authorisation, or the products you build. The researchers cited here worked on their own vehicles, inside sanctioned contests or through coordinated disclosure.
This post is part of a device security series. The companion guides on how Bluetooth gets hacked and how to secure it and how phones get hacked and how to secure them cover the two technologies that now double as your car key, and the guide on how Wi-Fi gets hacked covers the network most home chargers sit on. Our posts on fabless chip designers and the software supply chain and how a laptop is built explain where the chips and SDKs inside cars and chargers come from.
Contents
- How does a connected car work, and what software does it run?
- What is the attack surface of a car and its charger?
- What does an attacker need to hack a car?
- Which car hacks changed the industry?
- How does keyless relay theft work?
- How are connected car apps and APIs attacked?
- How are infotainment systems and Bluetooth attacked?
- How do EV chargers work, and what software do they run?
- How do EV chargers get hacked?
- How should you weigh these risks?
- How do you detect attacks on cars and chargers?
- What do UN R155, UN R156 and ISO/SAE 21434 require?
- Why is car and charger security a supply chain problem?
- How should car owners and home charger owners protect themselves?
- How should fleets and charge point operators secure cars and chargers?
- How should carmakers and charger makers build secure products?
- How Safeguard helps teams that build car and charger software
- The takeaway
- Frequently asked questions
- Sources
How does a connected car work, and what software does it run?
A connected car is a network of dozens of small computers called electronic control units (ECUs), joined by internal networks such as the CAN bus. A gateway separates them into domains, an infotainment head unit runs a full operating system, a telematics unit links the car to the cloud, and every one of them runs firmware.
Start with the vocabulary, because the rest of this guide depends on it.
An electronic control unit (ECU) is an embedded computer that runs one job or a group of jobs: engine or motor control, battery management, brakes, steering assistance, airbags, door locks, lights, seats, the instrument cluster. Wikipedia's article on the CAN bus notes that a modern car may have as many as 70 of them. Each runs firmware, usually written by a supplier rather than the carmaker, and each has a software version that has to be tracked and updated.
The Controller Area Network (CAN bus) is how most of those ECUs talk. Bosch began developing CAN in 1983 and released it in 1986. It is a shared bus: every ECU on a segment sees every message, and each message carries an identifier that says what it is, not who sent it. That design is efficient and dependable, and it has a property that matters for security. In Wikipedia's words, "CAN is a low-level protocol and does not support any security features intrinsically," and "there is also no encryption in standard CAN implementations." A receiving ECU cannot tell, from the frame alone, whether a "brake" or "unlock" message came from the right ECU or from something that should not be there. CAN FD (flexible data rate), released by Bosch in 2012, carries more data per frame and runs faster but keeps the same basic trust model.
Cars also use LIN, a slower and cheaper bus for simple sensors and switches, and increasingly automotive Ethernet for cameras, displays and the high-bandwidth links in newer "zonal" designs, where a few powerful computers replace many small ECUs.
The central gateway sits between network segments, which carmakers group into domains such as powertrain, chassis, body and driver assistance. A good gateway forwards only the messages each domain needs and blocks the rest. As you will see, the gap between a gateway that filters and a gateway that just forwards has been decisive in more than one attack.
The infotainment head unit (IVI) is the screen in the dashboard. It runs a full operating system: often Linux, QNX or Android. Android Automotive is "an operating system and platform running directly on the in-vehicle hardware," which differs from Android Auto, a platform running on the user's phone and projected to the car's screen. Automotive Grade Linux, a Linux Foundation project, provides a shared Linux platform for infotainment, instrument clusters, head-up displays and telematics. The head unit handles Bluetooth, Wi-Fi, USB media, navigation, apps and often voice assistants. It is the most feature-rich and most exposed computer in the car.
The telematics control unit (TCU) carries the cellular modem, the GNSS receiver and usually an embedded SIM. It is the car's link to the carmaker's cloud: remote unlock, remote start, location, emergency calls and over-the-air updates all flow through it.
Over-the-air (OTA) updates let the carmaker ship new software for ECUs without a workshop visit. That is the single most important defensive capability a modern car has, and also a high-value target, because whoever controls the update pipeline controls the fleet.
Keys have become computers too. A passive keyless entry fob wakes up when the car sends a short-range signal and answers with a cryptographic response. A phone-as-a-key system uses Bluetooth Low Energy, ultra-wideband (UWB) or NFC. Behind both sits an immobiliser that decides whether the powertrain may start.
The OBD-II port under the dashboard gives diagnostic tools access to the car's networks. Wikipedia notes that OBD-II has been mandatory for cars and light trucks sold in the United States since model year 1996, and CAN is one of the protocols it uses.
Figure 1. A connected car is dozens of computers, plus a cloud side and a charging side.
Electric vehicles add a battery management system, an onboard charger and a charge port controller that talks to the charging station over the cable. They also add a second ecosystem outside the car: the charger, its firmware and the cloud platform that runs it. We cover that later in its own section.
The cloud side is part of the car
The most important shift of the last decade is that a car's security now extends far outside the car. A modern vehicle trusts commands that arrive from the carmaker's cloud through the TCU. The cloud in turn trusts the owner's app, dealer portals, fleet management software and partner integrations. If any of those trust the wrong person, the car will unlock for them. That is why the most consequential remote car flaws since 2023 were web application bugs, not radio tricks.
What is the attack surface of a car and its charger?
The attack surface is every path into the car's software: remote paths through apps, APIs, telematics, updates and charging backends; radio paths through keys, Bluetooth and Wi-Fi; and physical paths through the OBD-II port, USB, exposed wiring and the charge cable. Remote paths are rarer but reach whole fleets at once.
It helps to group the entry points by how close the attacker has to be, because that decides how many vehicles a single flaw can reach.
Fully remote. Automaker APIs and the owner app; dealer, partner and fleet portals; the telematics cellular link; the OTA update pipeline, including signing keys and supplier build systems; and charging platform backends. A flaw here can affect every vehicle or charger on the platform.
Radio range. Keyless entry fobs, phone-as-a-key systems, and the Bluetooth and Wi-Fi radios in the head unit. These need the attacker within metres or tens of metres. Each attack hits one vehicle at a time, but the flaw may exist in millions.
Physical access. The OBD-II port inside the cabin, USB ports and media, wiring reachable from outside the car such as the CAN wiring behind a headlight, and the charge cable and charger. These need hands on the car or the charging bay.
Figure 2. Every way into a car and its charger, grouped by how close the attacker has to be.
The pattern across the public record is clear. Physical and radio attacks are the ones used to steal individual cars. Remote attacks are the ones that change an industry, because a single bug in a cloud service or a telematics design can reach a whole fleet.
What does an attacker need to hack a car?
It depends on the path. Relay theft needs a pair of radio relay devices and a car parked near its key. API attacks need web skills and an account. Remote attacks on the vehicle itself have needed a car to study, extensive reverse engineering and a chain of several flaws.
Here is what each class of attack has needed in the documented cases.
- Keyless relay theft needs two radio devices and a car whose key is within some tens of metres, for example inside a house near the front door. No knowledge of the car's cryptography is required, because the attack carries the genuine signals rather than breaking them.
- CAN injection theft needs physical access to a wire that carries CAN traffic from outside the cabin, and a device built to send the right messages for that model.
- Cloud and API attacks need web application security skills, a browser, an account with the carmaker's app or portal, and sometimes a target's VIN, licence plate, email address or phone number.
- Infotainment attacks need Bluetooth or Wi-Fi range or a USB port, plus a vulnerability in the head unit's software. Turning that into control of driving functions needs a further way across the gateway.
- Remote attacks on vehicle control have needed the most: a vehicle to study, deep knowledge of its internals, a remotely reachable service, a way to move from the head unit or TCU to the control networks, and a gateway or ECU that accepts the result.
- Charger attacks range from network access to the charger's local interface, to physical access to a USB or service port, to nothing more than a charger's identifier for backends that never check who is connecting.
The skill and cost gap between those rows explains the risk ranking later in this guide. Cheap attacks on individual cars and cheap attacks on cloud authorisation are where the real-world harm has been.
Which car hacks changed the industry?
The 2015 Jeep Cherokee remote attack forced the first cybersecurity recall, of 1.4 million vehicles. Keen Lab's 2016 Tesla research showed a fix could ship over the air within days. Research on telematics dongles, BMW telematics units and, since 2023, automaker web APIs moved the industry's focus from the radio to the cloud.
The timeline below shows the cases this guide covers, for cars and chargers together.
Figure 3. Fifteen years of public car and charger research, from relay attacks in 2011 to the 2026 charging backend CVEs.
The 2015 Jeep Cherokee remote attack
In 2015, Charlie Miller and Chris Valasek published Remote Exploitation of an Unaltered Passenger Vehicle, describing how they took control of functions in a 2014 Jeep Cherokee from a laptop far away. It is the case most people mean when they say "car hacking," and it is worth understanding as a chain of trust failures rather than a single bug.
How it worked, conceptually. The Uconnect head unit in the Jeep had cellular connectivity through Sprint's network. A message service on the head unit, the D-Bus inter-process communication system, was reachable from the cellular network and accepted commands without authentication. The paper notes that "Sprint does not block this type of traffic between devices on their network", so any device on the carrier's network could reach any vulnerable head unit. From the head unit, the researchers could talk to a separate Renesas V850 chip that connects to the CAN bus. That chip's firmware could be replaced with modified firmware, which then sent CAN messages. ECUs on the vehicle networks acted on those messages, and the paper documents effects on functions including steering and braking under specific conditions.
What the attacker needed. A vehicle of the same model to study, a great deal of reverse engineering, a device on the same cellular network, and the five flaws in the chain. It was a research effort, not a script.
Impact. Chrysler recalled 1.4 million vehicles with 8.4-inch Uconnect touchscreens, including 2013 to 2015 Ram pickups and Dodge Vipers and 2014 to 2015 Jeep Grand Cherokees and Cherokees. Owners received a USB device to update the software. The carrier also blocked the traffic: Fiat Chrysler asked Sprint for a network fix "blocking access to the specific port used to penetrate the Jeep's computer systems," as Transport Topics reported.
What the case taught. Every link in the chain is now a standard defence. Filter traffic on the carrier network so devices cannot reach each other. Require authentication on every service. Separate infotainment from control networks with a gateway that filters, not just forwards. Verify a cryptographic signature before accepting any firmware. Authenticate messages on the vehicle network where it matters. And have a way to update cars without mailing out USB sticks.
Figure 4. The 2015 Jeep research as five links in a chain. Breaking any one would have stopped it.
Telematics dongles plugged into the OBD-II port
The same summer, a team from the University of California San Diego presented Fast and Vulnerable: A Story of Telematic Failures at USENIX WOOT 2015. They studied an aftermarket telematics unit that plugs into the OBD-II port, the kind of dongle insurers and fleet services use to track driving. Their abstract states: "We show that these devices can be discovered, targeted, and compromised by a remote attacker." The Register reported the team's demonstration of "a complete remote compromise via SMS," with brakes and other functions manipulated at low speed, and that fixes were released.
The lesson is uncomfortable for owners. Anything you plug into the OBD-II port sits on the car's diagnostic network, and its security becomes the car's security. Because such devices are aftermarket, the carmaker has limited ability to fix them.
Keen Lab and Tesla (2016)
In September 2016, Tencent's Keen Security Lab reported that it had "successfully implemented remote, aka none physical contact, control on Tesla Model S in both Parking and Driving Mode," by compromising the CAN bus. Tesla's product security team confirmed the issues. Electrek reported the next day that Tesla had "already deployed an over-the-air software update (v7.1, 2.36.31)," and that, according to Tesla, the issue was "only triggered when the web browser is used."
The defensive point is the speed. A year after Chrysler mailed USB sticks, a carmaker with a working OTA pipeline shipped a fix to its fleet within days of disclosure.
Keen Lab and BMW telematics (2018)
In May 2018, Keen Lab published an experimental security assessment of BMW cars. It reported 14 vulnerabilities with local and remote access vectors, seven of which received CVE numbers. Some were in the head unit; others were in the telematics unit. The CVE record for CVE-2018-9311 describes how the telematics control unit "in certain BMW vehicles allows remote attacks over a cellular network," for vehicles produced from 2012 through 2018. BMW confirmed the vulnerabilities were in "the infotainment and T-Box components," and Keen Lab reports that BMW began rolling out mitigations for the cellular attack path from April 2018 as remote configuration updates.
The case shows the TCU is as much a target as the head unit, and that a fix can sometimes land in the backend and the network configuration rather than in every car.
How does keyless relay theft work?
Keyless relay theft uses two radio devices to stretch the link between a car and its key fob. One sits near the car, the other near the key inside the house. The car hears its genuine key answer and unlocks, because it checks that the key responds, not how far away it is.
Passive keyless entry and start (PKES) is convenient: walk up, the car unlocks; press start, it goes. Under the hood, the car sends a short-range low-frequency wake-up signal, the fob answers over a longer-range radio band with a cryptographic response, and the car accepts it.
Why it works
The cryptography in these systems can be sound. The problem is an assumption: that if the fob answers, the fob must be close. Radio signals can be captured and retransmitted in real time. If one device sits by the car and a second sits within range of the fob, the pair can carry the conversation across the gap. The car is not hearing a copy or a forgery. It is hearing its own key, just from further away than the designers assumed.
This was demonstrated in public research long before relay theft became a mainstream crime. In their NDSS 2011 paper, Aurélien Francillon, Boris Danev and Srdjan Capkun showed relay attacks on passive keyless entry systems in 10 models from 8 manufacturers. Their slides note that the relay was tested at distances of up to 50 metres and that "authentication do not prevent it."
Figure 5. Keyless relay theft, conceptually. The car checks that the key answers, not how far away it is.
Phone as a key has the same problem
Phone-as-a-key systems that use Bluetooth Low Energy to sense proximity face the same physics. In 2022, NCC Group showed a Bluetooth LE relay against Tesla Model 3 and Model Y. BleepingComputer reported that the researchers could "unlock and operate the vehicle" with the phone 25 metres away, and that Tesla responded that "relay attacks are a known limitation of the passive entry system." The recommended mitigations were to turn off passive entry or enable PIN to Drive. Our Bluetooth security guide covers this case and the newer Bluetooth Channel Sounding feature in more depth.
CAN injection: theft without the key at all
A different technique appeared in public research in 2023. Ken Tindell of Canis Automotive Labs described CAN injection after investigating a car theft. The thieves pulled the bumper away and reached the CAN bus "from the headlight connector." A device then sent messages that impersonated the smart key receiver. The car's gateway forwarded those messages to the powertrain network, where the engine ECU deactivated the immobiliser. The device was disguised inside a Bluetooth speaker case.
Why it works is the CAN property covered earlier: a receiving ECU cannot tell who sent a message. Tindell's write-up proposes two defences. In the short term, ECUs can check whether recent bus errors make a message implausible. In the long term, CAN messages should be authenticated with keys specific to each vehicle, which Tindell calls a zero trust approach to CAN.
What stops relay and injection theft
- Distance measurement. Ultra-wideband keys measure how long a signal takes to travel. Bosch describes this as a "distance-bounding mechanism" that "prevents relay attacks by ensuring that the key is physically present near the vehicle." Research has explored attacks on distance measurement too, so treat UWB as raising the bar substantially rather than ending the problem.
- Motion-sensing fobs that go to sleep after a period without movement. A key lying still on a hall table stops answering.
- A second factor to drive, such as a PIN to Drive feature, so unlocking the door is not enough to move the car.
- Turning keyless entry off if the car allows it, or storing the fob in a signal-blocking pouch away from outside walls.
- Authenticated in-vehicle messages and gateways that do not forward key and immobiliser messages from the wrong network segment. This is the maker's job, and it is the real fix for CAN injection.
- A visible physical barrier, such as a steering lock, which raises effort for any theft method.
Detection for owners is mostly after the fact: the car's app showing an unlock or start you did not make, or a dashboard fault after headlight damage. For makers, it is event logging in the gateway and the key ECU, and an anomaly signal when key messages arrive from a segment where they never should.
How are connected car apps and APIs attacked?
Connected car apps send commands through the carmaker's cloud, which relays them to the car. Since 2023, researchers have repeatedly found that those cloud services did not check whether the person asking had any right to that vehicle, allowing remote unlock, start and tracking from a VIN, an email address or a licence plate.
When you tap "unlock" in a car app, the app calls the carmaker's API. Backend services check your session, look up the vehicle, and pass a command to a telematics service, which sends it over cellular to the car's TCU. The car generally executes what the TCU tells it. So the security of the whole chain rests on one question asked in the cloud: is this user allowed to do this to this vehicle?
Sam Curry's 2023 automaker research
In January 2023, Sam Curry and a group of researchers published Web Hackers vs. The Auto Industry, covering work done during 2022 across carmakers and their suppliers. The issues fell into a few classes:
- Remote commands from a VIN. For Kia, Honda, Infiniti, Nissan and Acura, the researchers reported remote lock and unlock, engine start and stop, and vehicle location using only a vehicle identification number, along with exposure of personal data and account takeover.
- Remote commands from an email address, for Hyundai and Genesis.
- Single sign-on misconfiguration. At BMW and Rolls Royce, SSO flaws led to employee account takeover and access to internal dealer portals. At Mercedes-Benz, a misconfigured SSO gave access to internal tools including source repositories and chat.
- Account takeover and data exposure at Ferrari, Ford, Porsche, Jaguar Land Rover and Toyota Financial, among others.
- Supplier platforms. Telematics provider Spireon exposed administrative access to about 15.5 million devices. SiriusXM Connected Vehicle Services and the digital licence plate company Reviver were also affected.
The post records fixes for several, including BMW and Rolls Royce ("have since been fixed"), Nissan and Infiniti, and Reviver, which it says was "patched in under 24 hours."
The 2024 Kia web portal issue
In September 2024, Curry and colleagues published Hacking Kia: Remotely Controlling Cars With Just a License Plate. They found that a weakness in Kia's dealer infrastructure let an attacker, with only a licence plate, take over remote access to "any hardware-equipped vehicle" in about 30 seconds, whether or not the owner had an active Kia Connect subscription. The attacker could obtain the owner's name, phone number, email and address, and silently add themselves as a second user. The demonstrated commands included lock, unlock, start, stop, honk and locate, across model years 2014 to 2025.
Kia told the researchers in August 2024 that it had remediated the vulnerability. The write-up states that "the Kia team has validated this was never exploited maliciously," and that the proof-of-concept tool "was never released."
The 2025 Subaru Starlink issue
In January 2025, Curry published Hacking Subaru. An administrative panel for Subaru's Starlink connected services gave remote start, stop, lock, unlock and locate for any targeted vehicle, plus up to a year of location history and customer details. A target could be found by last name and ZIP code, email, phone number or licence plate. Subaru fixed the issue within 24 hours of the report, and the post says it was "never exploited maliciously."
Figure 6. From app tap to unlocked door. Most remote car flaws since 2023 were in the cloud.
Why this keeps happening
The common thread is broken object level authorisation: the backend checks that you are logged in, but not that the vehicle you are acting on is yours. Dealer and support portals make it worse, because by design they need to act on many vehicles, so a flaw in their authorisation reaches every car the portal can see. Employee single sign-on adds another path, because a takeover of one employee account can open internal tools.
Impact. Location tracking, unlocking, remote start and exposure of owners' personal data, at fleet scale. In the published cases the companies fixed the issues quickly, and several stated they were not exploited maliciously.
Detection. For carmakers: log every remote command and every location lookup with the user, the vehicle and the source; alert on one account touching many VINs, on dealer accounts acting outside their dealership, and on spikes of commands. For owners: watch the app's activity history and the list of authorised users and drivers.
Defence. Check authorisation on every request for this user and this vehicle, in the backend, not the app. Treat dealer, partner and employee portals as high-risk systems with strong sign-in, least privilege and session limits. Rate-limit and log remote commands. And run a public bug bounty: every one of these cases came from researchers who reported what they found.
How are infotainment systems and Bluetooth attacked?
Head units run full operating systems with Bluetooth, Wi-Fi, USB and apps, so they inherit the bugs of those stacks. Researchers have shown remote code execution through head unit Bluetooth stacks and dozens of zero-days in commercial head units at Pwn2Own. The key question is whether the head unit can reach control networks.
PerfektBlue (2025)
In July 2025, PCA Cyber Security published PerfektBlue, a set of vulnerabilities in OpenSynergy's BlueSDK, a commercial Bluetooth stack used in car head units. The Hacker News reported four CVEs:
- CVE-2024-45434 (CVSS 8.0), a use-after-free in the AVRCP profile, which handles media controls;
- CVE-2024-45431 (CVSS 3.5), improper validation of an L2CAP channel's remote identifier;
- CVE-2024-45433 (CVSS 5.7), incorrect function termination in RFCOMM;
- CVE-2024-45432 (CVSS 5.7), a function call with an incorrect parameter in RFCOMM.
Affected head units came from Mercedes-Benz, Volkswagen and Skoda, plus a fourth unnamed maker. The flaws were disclosed to the vendor in May 2024 and patches rolled out from September 2024. Volkswagen said it was "addressing the security gap with software updates" and that the issues concerned Bluetooth and infotainment rather than vehicle safety. That gap between the vendor's patch and the public advisory is typical of automotive supply chains: the fix travels from the stack vendor to the Tier 1 supplier to the carmaker to the car, and each step takes time. Our Bluetooth guide covers PerfektBlue in more detail.
Head units at Pwn2Own Automotive
The Zero Day Initiative has run Pwn2Own Automotive in Tokyo every January since 2024, a sanctioned contest where researchers attack production products and the vendors receive the vulnerabilities to fix. Head units have been targets every year.
- 2024. On day one, Synacktiv compromised the Tesla modem with a three-bug chain and won $100,000. Teams also compromised an Alpine Halo9 head unit, a Sony XAV-AX5500 and a Pioneer DMH-WT7600NEX. On day two, Synacktiv compromised Tesla's infotainment system with a two-bug chain for another $100,000.
- 2025. The final results include successful attacks on Sony, Alpine and Kenwood head units.
- 2026. On day one, Synacktiv demonstrated a USB-based attack on Tesla infotainment, combining an information leak and an out-of-bounds write. Teams compromised Sony XAV-9500ES, Alpine iLX-F511 and Kenwood DNR1007XR units across the event.
Many aftermarket head units are standalone and do not reach the car's control networks, so the impact of these results varies. But they show the code quality of head unit software, which often shares stacks and components with factory units.
USB and media
A head unit parses everything plugged into it: music files, video, playlists, firmware update packages. Each parser is attack surface, and the 2026 Tesla result shows that USB remains a realistic path. For owners the rule is simple: do not plug unknown drives into the car. For makers, it means treating media parsers as untrusted input handlers, sandboxing them and fuzzing them.
The bridge question
A compromised head unit can leak contacts, messages and location history synced from your phone, which is a privacy harm on its own. Whether it can do more depends on the architecture. If the head unit can send messages onto a CAN segment that reaches driving functions, and the gateway forwards them, infotainment flaws become safety flaws. If the gateway filters strictly and control ECUs authenticate the messages that matter, the damage stays inside the head unit. That architectural line, drawn and enforced, is worth more than any single patch.
How do EV chargers work, and what software do they run?
An EV charger is an embedded computer with a power stage. It talks to the car over the cable, and to a cloud backend called a charging station management system (CSMS) using OCPP over a WebSocket. The charger's firmware, its local interfaces and the backend are each attack surface.
The charging world has its own vocabulary.
EVSE (electric vehicle supply equipment) is the formal term for the charger. Home chargers, often called wallboxes, run embedded Linux or a real-time operating system, connect over Wi-Fi, Ethernet or cellular, and usually have a phone app, Bluetooth for setup, a web or service interface and sometimes a USB port. Public DC fast chargers add payment terminals, screens and more complex power electronics.
The charging station management system (CSMS), also called the backend or central system, is the cloud service that authorises sessions, starts and stops charging, sets power limits, collects meter data, bills drivers and pushes firmware updates. A charge point operator may run its own CSMS or buy one as a service.
The Open Charge Point Protocol (OCPP) is, in the Open Charge Alliance's words, "the global open communication protocol between charging stations and charging management systems." OCPP 1.6 was released in 2015, OCPP 2.0.1 in 2020 and OCPP 2.1 in 2025. OCPP 2.0.1 added device management, security features, smart charging and support for ISO 15118, and the Alliance states it has been approved as an IEC standard, IEC 63584. The widely deployed JSON flavours of OCPP run over a WebSocket, a long-lived connection that the charger opens to the backend.
ISO 15118 governs the high-level conversation between car and charger over the charging cable, including Plug and Charge, where the car presents a certificate so the driver does not need an app or card. On CCS fast chargers, this conversation runs over the power line itself using HomePlug Green PHY, after a pairing step called SLAC (Signal Level Attenuation Characterization) that works out which car is on which cable.
Figure 7. How EV charging works, and where each link has been attacked in public research.
OCPP security profiles
The OCPP security specification defines three security profiles for how a charger and backend authenticate each other. The WinCC OA OCPP documentation summarises them, and the Open Charge Alliance's OCPP 1.6 security whitepaper explains how the OCPP 2.0 security features apply to OCPP 1.6 deployments.
| Profile | Charger proves who it is with | Backend proves who it is | Traffic encrypted | Use |
|---|---|---|---|---|
| 1. Unsecured transport with basic authentication | A password (HTTP Basic) | No | No | Only inside an already protected network |
| 2. TLS with basic authentication | A password, sent inside TLS | Yes, by TLS server certificate | Yes | A reasonable minimum for internet-connected chargers |
| 3. TLS with client-side certificates | Its own certificate (mutual TLS) | Yes | Yes | The strongest; each charger has its own key |
| No authentication at all | Nothing but its charger ID | Varies | Varies | The flaw behind the 2026 backend CVEs below |
Profile 1, in WinCC OA's words, "does not include authentication for the Central System, or measures to set up a secure communication channel." Vendor guidance, such as Ampcontrol's, describes profile 2 as encrypting OCPP traffic with TLS and profile 3 as adding mutual authentication. If you operate chargers, the profile in use on each one is one of the first facts worth knowing.
How do EV chargers get hacked?
Chargers get hacked through firmware bugs in their local and network services, through cloud APIs that let one user control another's charger, through backends that accept any client claiming a charger's ID, and through attacks on the cable link that disrupt or intercept charging sessions. Most documented flaws were fixed by firmware or backend updates.
Charger firmware vulnerabilities
Chargers are embedded devices with network services, web interfaces, update mechanisms and OCPP clients, and they show the same classes of bugs as routers and smart home devices: command injection, memory corruption and missing authentication. These are well-documented examples:
| Charger | CVE | What it allowed | Fix |
|---|---|---|---|
| Phoenix Contact CHARX SEC-3000 series | CVE-2024-25995 | Unauthenticated remote configuration change leading to code execution, root rights or denial of service | Firmware 1.5.1, per VDE-2024-011 |
| ChargePoint Home Flex | CVE-2024-23971 | Network-adjacent, unauthenticated command injection through OCPP message handling, running as root | Vendor states it was patched in April 2024, per ZDI-24-1053 |
| Tesla Wall Connector | CVE-2025-8321 | Physically present attacker could run code because the firmware update had no anti-downgrade protection | Firmware 24.44.3, per ZDI-25-712 |
| Autel MaxiCharger AC Wallbox Commercial | CVE-2025-5828 | Physically present attacker could run code through a USB buffer overflow | Firmware V1.39.51 (US) and V1.56.51 (EU), per ZDI-25-347 |
| Autel MaxiCharger Single | CVE-2026-8984 | Unauthenticated remote code execution through a service on the charger's network interface | Affects firmware through V1.03.51, per the CVE record |
| Autel MaxiCharger Single | CVE-2026-8986 | Command injection when processing an OCPP diagnostics request from a malicious or compromised backend | Affects firmware through V1.03.51, per the CVE record |
Two patterns stand out. First, several of these bugs live in the OCPP handling code, which means the backend, or anyone who can impersonate it, is part of the attack surface of every charger. CVE-2026-8986 is explicit: "a malicious or compromised OCPP server" can run commands on the charger. Second, physical attacks matter for chargers in a way they rarely do for phones, because public chargers stand unattended in car parks. Our analysis of the Autel MaxiCharger port 9002 flaw goes deeper on one of these.
Earlier research showed the same weaknesses. In 2018, Kaspersky examined the ChargePoint Home and found stack buffer overflows in its web interface, reachable by someone on the owner's Wi-Fi. The impact included raising or stopping charging current, with the risk that a higher current could overload a home's wiring. The vulnerabilities were patched.
Pwn2Own Automotive results for chargers
Chargers have been a full category at Pwn2Own Automotive since the first event.
- 2024. On day three, teams compromised the ChargePoint Home Flex, Ubiquiti Connect EV Station, JuiceBox 40, Phoenix Contact CHARX SEC-3100 and EMPORIA Level 2 charger. The event as a whole awarded $1,323,750 for 49 unique zero-days, and Synacktiv won Master of Pwn.
- 2025. The final results include attacks on ChargePoint, Autel MaxiCharger, Ubiquiti and WOLFBOX chargers and the Tesla Wall Connector. ZDI "awarded $886,250 for 49 0-days," and Sina Kheirkhah of Summoning Team won Master of Pwn.
- 2026. The day three results include the Grizzl-E Smart 40A, the Alpitronic HYC50 fast charger and the Autel MaxiCharger AC Elite Home 40A. ZDI reports that "$1,047,000 USD was awarded for 76 unique 0-day vulnerabilities," and Fuzzware.io won Master of Pwn.
Figure 8. Pwn2Own Automotive, three years of results across Tesla systems, head units and chargers.
Pwn2Own results are a lower bound on what exists, not a measure of what is exploited in the wild. They are useful because each one goes to the vendor, which then has to ship a fix, and because they show that well-resourced teams find serious bugs in production chargers every single year.
Cloud APIs that hand out control of chargers
In July 2021, Pen Test Partners published research on smart chargers after roughly 18 months of work, reporting vulnerabilities "that allowed account hijack of millions of smart EV chargers." The issues were mostly in vendor cloud APIs rather than the hardware:
- one platform's API skipped real credential checks and used predictable serial numbers, with what the researchers described as "roughly 2.9 million devices on it";
- Wallbox "had two separate insecure direct object references in their API which allowed account hijack," fixed within days;
- EVBox let users change their own role through a profile API, a privilege escalation "acknowledged and fixed in around 24 hours";
- ChargePoint exposed a GraphQL endpoint that allowed introspection, fixed in about a day.
These are the charger equivalent of the automaker API flaws above. A charger that trusts its cloud, and a cloud that does not check ownership, means anyone with an account may be able to control someone else's charger.
Backends that never ask the charger who it is
In 2023, researchers at SaiFlow described a weakness in how OCPP 1.6J backends accept connections. The Hacker News quoted them: "The OCPP standard doesn't define how a CSMS should accept new connections." Backends that identify a charger only by its ID can be fooled by a second connection using the same ID. Depending on the backend, that either drops the real charger, which is a denial of service, or diverts responses, including driver and session data, to the impostor. The researchers' mitigations were to require charger credentials, as OCPP 2.0.1 does, and to test both connections and drop the one that does not respond.
In February and March 2026, the CVE Program published a wave of records showing that this was not theoretical. Charging platforms including EV2GO, Chargemap, Mobiliti, Everon, Mobility46, swtchenergy.com, EV Energy and ePower received CVEs with the same description: "WebSocket endpoints lack proper authentication mechanisms, enabling attackers to perform unauthorized station impersonation and manipulate data sent to the backend." An unauthenticated attacker could connect to the OCPP endpoint "using a known or discovered charging station identifier." A related record for EV Energy, CVE-2026-26290, describes session "hijacking or shadowing, where the most recent connection displaces the legitimate charging station." Our write-up of CVE-2026-24731 and EV charger backends covers the EV2GO case and why charger identifiers are easy to discover.
Impact. False meter readings and session data in billing systems, denial of service for real chargers, exposure of session and driver data, and confusion in the smart charging signals that operators use to manage load.
Detection. Alert on two simultaneous connections for one charger ID, on a charger connecting from an unexpected network or country, and on impossible meter values or session patterns.
Defence. Use security profile 2 at minimum and profile 3 where chargers support it, with a unique credential or certificate per charger. Never treat the charger ID as a secret.
Attacks on the cable link
The conversation between car and charger has been attacked too, though with narrower impact.
Brokenwire. Researchers from the University of Oxford and armasuisse S+T presented Brokenwire at NDSS 2023, a wireless attack that aborts CCS fast-charging sessions. It exploits the collision-avoidance behaviour required by the HomePlug Green PHY, DIN 70121 and ISO 15118 standards. The team tested it against 8 vehicles and 20 chargers, at distances up to 47 metres with less than 1 watt of power. It disrupts charging; it does not take over the car or the charger. Its significance is for anyone who depends on fast charging, such as fleets and emergency services: a cheap radio could stop a whole bank of chargers.
SLAC spoofing. The CVE record for CVE-2025-12357 states that "by manipulating the Signal Level Attenuation Characterization (SLAC) protocol with spoofed measurements, an attacker can stage a man-in-the-middle attack between an electric vehicle and chargers that comply with the ISO 15118-2 part." It is exploitable only at close range. The longer-term defence is TLS with certificate checks on the car-to-charger link, as ISO 15118-20 specifies.
Screens, signage and platform breaches
Not every charging incident is technical. In April 2022, Electrive reported that inappropriate content appeared on public EV charging signage on the Isle of Wight; the local council said "a third-party web address displayed on our electric vehicle (EV) signage appears to have been hacked." The chargers were taken out of service.
Charging platforms also hold driver accounts and payment relationships, so they are ordinary targets for data theft. In September 2025, The Register reported that Digital Charging Solutions, which runs charging services for carmaker brands, found that a customer service provider had accessed data "without a valid reason"; it said a single-digit number of cases were confirmed and that "billing is secure." In March 2026, The Register reported a ransomware attack on charger maker ELECQ's cloud platform, in which databases were encrypted and copied and customer contact details exposed; the company said devices were not affected.
How should you weigh these risks?
Fix first what is cheap for attackers and reaches many vehicles: cloud authorisation flaws, charging backends that do not authenticate chargers, keyless relay exposure and charger account takeover. Plan and monitor the harder chains that reach vehicle control, which need more skill but carry higher impact.
The matrix below is a qualitative reading of the public research, not a scoring standard. It places each attack class by the effort and skill it has needed and by its impact on owners and fleets.
Figure 9. Risk versus attacker effort for the attack classes in this guide.
Two conclusions follow. For owners, the realistic threat is theft and tracking, not someone steering the car remotely; the defences in the owner checklist below are aimed squarely at that. For makers and operators, the cloud deserves at least as much security attention as the vehicle. The flaws that could reach the most cars and chargers in the last three years were authorisation bugs in web services, and they are cheap to find for anyone who looks.
How do you detect attacks on cars and chargers?
Owners detect attacks mostly through their app's activity log and unfamiliar users or devices. Fleets and charge point operators need logs from telematics and the charging backend. Makers need vehicle security monitoring that correlates ECU, gateway and cloud events across the fleet.
Signals for owners
- Unlocks, starts, climate commands or location lookups in the app history that you did not make.
- Unknown phones, keys or users listed as authorised drivers in the app or the car's settings.
- Unknown Bluetooth devices paired to the head unit.
- Warning lights or electrical faults after damage near a headlight or bumper, which can be a sign of tampering with CAN wiring.
- For a home charger: sessions you did not start, changed current limits, or an app account change you did not request.
Signals for fleets and charge point operators
- Telematics events out of pattern: remote commands outside working hours, vehicles reporting from unexpected locations, a sudden change in the list of users with app access.
- In the CSMS: two connections for the same charger ID, chargers connecting from unexpected networks, chargers that silently fall back to an older protocol or weaker security profile, meter values that are physically impossible, and firmware versions that change without a scheduled update.
- In the charger's own logs, where available: security events such as failed logins, firmware update attempts and tamper alerts.
Signals for makers
Carmakers increasingly run a vehicle security operations centre (VSOC), which collects security events from vehicles, the telematics backend and the app platform. Useful detections include key and immobiliser messages arriving from the wrong network segment, unexpected diagnostic sessions, failed message authentication on in-vehicle networks, ECU software versions that do not match the update records, and cloud-side anomalies such as one account acting on many VINs. Platforms such as Upstream, which describes itself as a cybersecurity and intelligence platform built on mobility data, focus on this monitoring layer. Detection on the vehicle side is only as good as the inventory behind it: you cannot spot an unexpected software version if you do not know which version should be there.
What do UN R155, UN R156 and ISO/SAE 21434 require?
UN R155 makes an audited cybersecurity management system a condition of vehicle type approval; UN R156 does the same for a software update management system. ISO/SAE 21434 and ISO 24089 are the engineering standards most makers use to meet them. In the EU they applied to new vehicle types from July 2022 and all new vehicles from July 2024.
Car security was once voluntary. In major markets it is now a condition of selling a car.
UN Regulation 155 (R155) requires carmakers to operate a cybersecurity management system (CSMS) covering how they identify, assess and treat cyber risk across the vehicle's life, including in their supply chain, and to show it is working for each vehicle type. UN Regulation 156 (R156) requires a software update management system (SUMS). As the itemis glossary entry puts it, R156 "makes an audited Software Update Management System (SUMS) a prerequisite for type approval," and manufacturers must know which software version runs on which vehicle type at any time. Type-approval-relevant software configurations are recorded with an RXSWIN, a software identification number, and each update has to be checked for whether it changes an approved state.
The two are twins: "same origin, same deadlines, same approval logic," in itemis's words. Both applied to new vehicle types from July 2022 and to all newly produced vehicles, including existing types, from July 2024. Upstream's 2024 analysis notes the same July 2024 milestone for R155. They apply in the contracting states of the UN's 1958 vehicle agreement that adopt them, which itemis lists as including the EU, the UK, Japan and South Korea. They do not apply in the United States, but because so many carmakers sell into Europe, they have become a global benchmark.
ISO/SAE 21434:2021, "Road vehicles: Cybersecurity engineering," is the engineering standard most makers use to meet R155. It covers the whole life of a vehicle's electrical and electronic systems, "from concept and product development through production, operation, maintenance, and decommissioning," as Keyfactor's summary puts it, and its core method is threat analysis and risk assessment (TARA). Keyfactor describes it as "the recognized engineering framework supporting UNECE UN R155." ISO 24089, "Road vehicles: Software update engineering," plays the same role for R156.
Figure 10. Two regulations, two engineering standards, and one software inventory underneath.
What this means in practice
Three practical consequences follow for anyone in the supply chain.
- Suppliers are inside the scope. A carmaker's CSMS has to cover the risks that come from its suppliers, so Tier 1 and Tier 2 suppliers are asked for evidence: their own processes, vulnerability handling commitments and, increasingly, software bills of materials.
- Monitoring does not end at sale. Makers have to watch for new vulnerabilities that affect vehicles already on the road and respond to them, which is impossible without an accurate inventory of what software is in each vehicle.
- Updates need records. Every update needs a record of what changed, for which vehicles, and how it was verified.
Chargers have their own standards: OCPP 2.0.1 between charger and backend, published as IEC 63584, and ISO 15118 between car and charger. Products with digital elements sold in the EU also face the Cyber Resilience Act, which our guide to CRA SBOM obligations explains. For a plain language walk through the car rules, see connected vehicle software, R155, R156 and ISO 21434 in plain terms and our post on R155's software supply chain requirements for automakers.
Why is car and charger security a supply chain problem?
A car or charger is assembled from software written by dozens of companies: chip vendors' SDKs, Tier 1 suppliers' ECU firmware, commercial stacks such as Bluetooth, open source operating systems and libraries, and cloud services. A flaw in one shared component appears in many brands, and the fix has to travel back through every link.
Look at the cases in this guide through that lens.
- PerfektBlue was a flaw in one supplier's Bluetooth stack that appeared in head units from several carmakers. The fix had to go from OpenSynergy to the head unit suppliers to the carmakers to the cars, which took months.
- The 2015 Jeep chain ran through a head unit, a cellular module, a separate chip with its own firmware and a carrier network, each from a different company.
- The 2023 API research reached suppliers such as Spireon and SiriusXM Connected Vehicle Services, whose platforms serve many brands.
- Charger firmware is built on embedded Linux distributions, web servers, MQTT and OCPP libraries, and chip vendor SDKs. The CHARX SEC advisory includes an out-of-bounds write in the MQTT stack, CVE-2024-26001, for example.
- Charging backends are often bought as a service. The 2026 OCPP CVEs hit several platforms with the same flaw class, and every operator using those platforms inherited the exposure.
The chain
A typical vehicle's software moves through several hands. Chip vendors ship SDKs, boot loaders and low-level drivers. Tier 2 suppliers add stacks and middleware: Bluetooth, TCP/IP, cryptography, AUTOSAR basic software. Tier 1 suppliers build complete ECUs, head units and telematics units on top. Open source sits inside many layers: Linux and its userland, Android Automotive, Automotive Grade Linux, open source libraries in apps and backends. The carmaker integrates everything, signs it and ships updates. And the cloud side has its own stack of open source packages, third-party services and suppliers.
Each hand-off is a place where a vulnerable component can enter, and where knowledge of what is inside can be lost. That is why the single most useful artefact in automotive security is an accurate software bill of materials (SBOM): a machine-readable list of every component and version in each ECU image, head unit build, charger firmware and backend service. Our guide to SBOMs for automotive ECU firmware and embedded software covers how to build one.
Secure updates are a supply chain control
OTA updates are the defence that makes everything else fixable, and they are also a supply chain risk: whoever can sign an update can change every car. Uptane is "an open software update strategy" first developed with automotive makers and suppliers, designed so that the compromise of a single server or key does not hand an attacker the fleet. It was adopted into IEEE's Industry Standards and Technology Organization in 2018 and later affiliated with the Linux Foundation as a Joint Development Foundation project. The Uptane site lists adopters including Automotive Grade Linux, through its aktualizr client, and several update platform vendors, and states that it "has been adopted by several major automakers."
On the vehicle networks, AUTOSAR, the industry's standard software architecture for ECUs, specifies Secure Onboard Communication (SecOC), which adds a message authentication code and a freshness value to messages so receivers can reject forged or replayed ones. That is the long-term answer to CAN injection, and it depends on keys being provisioned and managed across suppliers.
Why it is hard
- Binary deliveries. Carmakers often receive ECU firmware as compiled images, not source, so they cannot easily see what is inside.
- Long lives. Cars stay on the road for well over a decade. Components that were current at launch become years out of date, and some suppliers stop supporting them.
- Many variants. One model can ship many software configurations by market, trim and year, each needing its own inventory.
- Slow fixes. A patch has to be built by the supplier, integrated and validated by the carmaker, approved under R156 where relevant, and delivered over the air or at a dealer.
How should car owners and home charger owners protect themselves?
Install updates, secure the car's app account, reduce keyless relay exposure, avoid untrusted OBD-II dongles and USB media, remove old users and devices, and keep a home charger updated on its own network. These steps address theft, tracking and account takeover, which are the realistic threats.
For your car
- Install vehicle and app updates when offered, including those that arrive over the air. Most fixes in this guide reached owners this way.
- Protect the car account. Use a unique, strong password and turn on two-factor sign-in if the app offers it. The car account can unlock and locate your car, so treat it like a bank account.
- Review authorised users and keys in the app and the car, and remove anyone or any phone you do not recognise.
- Reduce relay exposure. Keep fobs away from outside walls and doors, use a signal-blocking pouch, and turn off keyless entry if your car allows it and you do not need it.
- Turn on PIN to Drive or an equivalent second factor if your car offers it.
- Be careful with OBD-II dongles. Insurance and tracking dongles sit on the car's diagnostic network. Use only those from your insurer or a reputable supplier, and unplug ones you do not need.
- Do not plug unknown USB drives into the head unit.
- Pair only your own phones and remove old pairings. The car stores contacts and messages from phones that have been paired.
- Before you sell or return a car, factory reset the head unit, remove the car from your app account and ask the dealer to remove you as an owner. When you buy a used car, make sure the previous owner no longer has app access.
- Use a visible physical lock if your model is a frequent theft target.
For your home charger
- Keep its firmware updated, and prefer a charger whose maker publishes security updates and a disclosure policy. Our guide to how long smart devices get security updates shows how to check before you buy.
- Put it on a separate network, such as a guest or IoT network, as our router and smart home guides describe.
- Secure the charger app account with a unique password and two-factor sign-in.
- Have it installed by a qualified electrician with correct protection on the circuit, so a software fault cannot push more current than the wiring can carry.
At public chargers, the practical risk to you is mostly to your payment and account data. Use the operator's official app or a contactless card, and be wary of QR codes stuck onto chargers that lead to unfamiliar payment pages.
How should fleets and charge point operators secure cars and chargers?
Fleets and operators need an inventory of every vehicle, telematics unit and charger with its software version, strong charger to backend authentication, monitoring for impersonation and odd commands, segmented networks, control over who can reach fleet portals, and contracts that require updates and SBOMs from suppliers.
- Inventory everything. Every vehicle, telematics device, charger and backend, with model, firmware version and who supplies updates. You cannot respond to a CVE in a charger model you do not know you run.
- Require OCPP security profile 2 or 3 with a unique credential or certificate per charger, and ask your CSMS provider how it handles duplicate connections and charger authentication. After the 2026 CVEs, this is the question to ask every backend vendor.
- Monitor the backend for duplicate charger connections, downgrades to weaker profiles, unexpected firmware changes and implausible meter data.
- Segment networks. Chargers belong on their own network segment with only the outbound connections they need. Depot and office networks should not be reachable from a charger.
- Control fleet and dealer portal access. Fleet management systems can locate and command many vehicles. Use strong sign-in, least privilege and regular access reviews, and remove leavers promptly.
- Review aftermarket telematics. Know which dongles and trackers are in your vehicles, who wrote their software and how they are updated.
- Contract for security. Require suppliers of vehicles, chargers and backends to provide security updates for a defined period, a vulnerability disclosure process and SBOMs. Our post on third-party risk with Safeguard TPRM shows how to track those commitments.
- Plan for outages. Brokenwire-style disruption, a backend ransomware attack or a bad firmware update can all stop charging. Know how drivers will charge and how vehicles will be dispatched if the platform is down.
- Secure physical access to public chargers: tamper-evident enclosures, closed service ports and alerts when a cabinet is opened.
How should carmakers and charger makers build secure products?
Makers should threat model to ISO/SAE 21434, authorise every cloud request for the specific vehicle and user, separate infotainment from control networks, authenticate critical in-vehicle messages, sign and verify every update, ship SBOMs, run disclosure and bug bounty programmes, and be able to fix third-party components fast.
Vehicle architecture
- Threat model each system with a TARA under ISO/SAE 21434, including the paths that cross from infotainment and telematics into control domains.
- Make the gateway a firewall, not a router: forward only the messages each domain needs, and log the rest.
- Authenticate critical messages with AUTOSAR SecOC or an equivalent, especially immobiliser, key and safety-relevant messages, so CAN injection fails.
- Lock down diagnostics. Require authentication for diagnostic sessions and firmware programming, and treat the OBD-II port as untrusted.
- Use distance-bounded keys (UWB) and motion-sensing fobs, and offer a PIN to Drive option.
Cloud and apps
- Authorise every API request in the backend for this user and this vehicle. Test for broken object level authorisation as a release gate, not a pentest afterthought.
- Treat dealer, partner and employee portals as crown jewels: strong sign-in, least privilege, short sessions and full audit logs.
- Rate-limit, log and alert on remote commands and location lookups.
- Protect signing keys for updates in hardware security modules, separate from build systems.
Chargers and backends
- Default to OCPP security profile 3 where hardware allows, and never accept a charger on its ID alone.
- Treat every OCPP message as untrusted input in the charger, including those from your own backend; CVE-2024-23971 and CVE-2026-8986 were both in OCPP handling.
- Close or authenticate local services and debug ports, and add anti-downgrade protection to firmware updates, as the Tesla Wall Connector fix did.
- Use ISO 15118-20 with TLS for Plug and Charge, so the car-to-charger link is authenticated.
Updates, supply chain and disclosure
- Sign every update and verify it on the device, with a design such as Uptane that survives the compromise of one key or server.
- Build and keep an SBOM for every ECU image, head unit build, charger firmware and backend, and require SBOMs from suppliers.
- Watch for new vulnerabilities in what you have already shipped, and rank them by whether the vulnerable code is reachable in your product.
- Run a vulnerability disclosure programme and a bug bounty. Almost every case in this guide reached the vendor through a researcher.
- Plan how to fix third-party components when the supplier is slow.
Figure 11. Hardening checklists for the people who own, operate and build cars and chargers.
How Safeguard helps teams that build car and charger software
Safeguard covers the software supply chain inside a car or charger product in one platform: SBOMs for firmware, apps and backends, 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.
Safeguard does not sit on the CAN bus or in the charger's power stage. It sits where most long-lived exposure in this guide comes from: the firmware, stacks, open source components, apps and cloud services that carmakers, suppliers, charger makers and charging platforms build and maintain. Teams that take this seriously often end up with one tool for source scanning, another for binary analysis, another for SBOM management, a questionnaire tool for suppliers and a GRC platform for audits. 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 car or charger product, 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 automotive and charging teams that covers the owner and charger apps on both mobile platforms, desktop diagnostic and fleet tools, Linux-based images for head units, telematics gateways and charger controllers, and the containers behind charging backends. Raw firmware images for microcontroller ECUs 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 head unit image or charger firmware 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. That is how a team tells the OCPP parser flaw that needs fixing this week apart from an unused function in a bundled library, and it gives an R155 risk treatment decision the evidence it needs.
Scan the code you own
For ECU application code, head unit services, charger firmware, mobile 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. Charger flaws such as the hard-coded authentication token in the Autel MaxiCharger Single (CVE-2026-8983) and the OCPP command injections above are classes that static analysis and secrets scanning exist to catch. And because a leaked OTA or firmware signing key turns one bug into a fleet-wide one, secrets in firmware repositories are worth catching early.
SBOMs your customers, carmakers and auditors can use
Every project produces an SBOM you can export in CycloneDX or SPDX, among other formats. That is what a carmaker will ask a Tier 1 supplier for, what a charge point operator will ask a charger maker for, and what lets a team answer "do we ship BlueSDK, which OCPP library, which kernel 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 a supplier waiting on an upstream release, or a carmaker waiting on a supplier, that is the difference between shipping the fix in the next update and waiting a quarter. 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 teams working under R155, R156 and ISO/SAE 21434, or preparing for the EU Cyber Resilience Act, the SBOMs, vulnerability monitoring records, risk decisions and fix trail are the raw material that type approval audits, customer reviews and regulators ask for.
Deploy where your code lives
ECU source code, signing infrastructure and charger firmware 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
Automotive and device teams often combine several tools: a source SCA tool such as Snyk or Black Duck, a product security platform built for connected device makers such as Cybellum, which says it "brings the entire product security workflow into one dedicated platform," a vehicle security operations platform such as Upstream, and a GRC tool for audits. Those are capable products, and a vehicle SOC does a job that Safeguard does not: watching live vehicle and fleet telemetry. Safeguard's case is consolidation and remediation across the software itself: one inventory across source, binaries and images; one vulnerability model with reachability; fix pull requests and patched packages instead of tickets; and supplier risk and compliance evidence from the same data. Safeguard can also work alongside an existing product security or VSOC platform and take over SBOM delivery and fixing. For more context, see software supply chain security for automotive, ISO/SAE 21434 compliance for automotive software suppliers and our firmware supply chain security guide.
The takeaway
Car hacking has a cinematic reputation, and the 2015 Jeep research earned it. But the attacks that have reached real owners and real fleets since then are less dramatic. A keyless fob answering from the hallway. A message forged on wiring behind a headlight. A web portal that never asked whether the person sending the unlock command owned the car. A charging backend that trusted any client that knew a charger's ID.
Nearly all of them have been fixed, and the fixes share a shape. Someone found the flaw, told the company that wrote the software, and that company shipped an update that had to travel through suppliers, platforms and over-the-air pipelines to reach the car or charger. Owners can shorten that journey by installing updates and securing their accounts. Fleets and operators can shorten it by knowing exactly what they run. And makers can shorten it most of all, by knowing what is inside every ECU, head unit, charger and backend they ship, and by being able to fix it without waiting for everyone upstream.
Try it yourself
Click through these interactive Safeguard demos to see the steps described in this guide.
Test yourself
Why was the 2015 Jeep Cherokee attack such a landmark?
It led to the first cybersecurity recall, covering 1.4 million vehicles. The attack chained five flaws, starting from a head unit service reachable over the cellular network with no authentication, and the carrier also blocked the traffic.
How does keyless relay theft work?
Two radio devices stretch the link between the car and its key fob. One sits near the car and the other near the key inside the house. The car unlocks because it checks that the key answers, not how far away it is.
Name two defences against relay theft.
Ultra-wideband keys measure signal travel time to confirm the key is physically near the car. Motion-sensing fobs go to sleep when still, and a PIN to Drive feature means unlocking the door is not enough to move the car.
What went wrong in the connected car app and API cases, such as Sam Curry's 2023 research?
The carmakers' cloud services did not check whether the person asking had any right to that vehicle. Researchers reported remote unlock, start and location from just a VIN or an email address for several brands.
What is the right defence for those app and API flaws?
Check authorisation on every request for this user and this vehicle, in the backend and not the app. Also treat dealer, partner and employee portals as high-risk systems, rate-limit and log remote commands, and run a public bug bounty.
What is the difference between OCPP security profile 2 and profile 3 for EV chargers?
Profile 2 uses TLS with a password, and the post calls it a reasonable minimum for internet-connected chargers. Profile 3 adds a certificate per charger (mutual TLS), which is the strongest option.
Frequently asked questions
Can a car be hacked remotely?
Yes, but it is rare and has required serious effort when it targets the car's driving functions. Researchers have done it, most famously with a 2014 Jeep Cherokee in 2015 and a Tesla Model S in 2016, and both were fixed. The more common remote risk today is the carmaker's cloud: flaws in apps and web portals have allowed remote unlock, start and tracking. Updates and a secure car account are the main defences.
How do thieves steal keyless cars, and how do I stop it?
The most documented method is relay theft: two radio devices carry the signal between the car and the key fob inside your home, so the car unlocks and starts as if the key were next to it. A newer method injects forged messages into wiring reachable from outside the car. Keep fobs away from doors or in a signal-blocking pouch, turn off keyless entry or enable PIN to Drive if offered, and consider a steering lock.
Can someone hack my car through its app?
Not usually through the app on your phone, but through the cloud services behind it. Researchers found in 2023, 2024 and 2025 that several carmakers' and suppliers' backends did not check ownership properly, allowing commands from a VIN, email address or licence plate. Those flaws were fixed. Protect your car account with a unique password and two-factor sign-in, and check the app's list of users and activity.
Can my car be hacked through Bluetooth or USB?
It is possible. The 2025 PerfektBlue research found Bluetooth stack flaws in head units from several carmakers, and Pwn2Own Automotive has shown head unit attacks over USB. Most such attacks need the attacker close by and stay within the infotainment system unless the car's gateway lets them reach control networks. Install updates, pair only your own phones and do not plug unknown drives into the car.
Are public EV chargers safe to use?
For most drivers, yes. The documented charger attacks targeted the charger's firmware, its backend, its signage or the charging session, and were fixed by updates. The published research we cite shows disruption of charging sessions and close-range interception, not a public charger taking over a car's driving systems. Your main practical risk is to payment and account data, so use official apps or contactless cards and avoid QR codes stuck onto chargers.
How do I secure a home EV charger?
Keep its firmware updated, buy from a maker that publishes security updates, put it on a separate guest or IoT network, and protect the charger app account with a unique password and two-factor sign-in. Have it installed by a qualified electrician so the circuit is protected regardless of the software. Check the app occasionally for sessions or setting changes you did not make.
What are UN R155 and UN R156?
They are UN vehicle regulations that make cybersecurity and software updates part of type approval. R155 requires carmakers to run an audited cybersecurity management system, and R156 requires an audited software update management system. In the EU they applied to new vehicle types from July 2022 and to all newly produced vehicles from July 2024. ISO/SAE 21434 and ISO 24089 are the engineering standards most makers use to meet them.
What is an automotive SBOM and why does it matter?
A software bill of materials is a machine-readable list of every software component and version in a product, such as an ECU image, a head unit build or charger firmware. It matters because cars and chargers are built from many suppliers' code, and when a flaw like PerfektBlue appears in a shared component, an SBOM tells a maker in minutes which products ship it. R155 monitoring and supplier management depend on that inventory.
Sources
- Miller and Valasek, Remote Exploitation of an Unaltered Passenger Vehicle (2015)
- Computerworld, Chrysler recalls 1.4M vehicles after Jeep hack (July 2015)
- Transport Topics, Hacked Jeep exposes weak underbelly of high-tech vehicles (August 2015)
- Foster, Prudhomme, Koscher and Savage, Fast and Vulnerable: A Story of Telematic Failures, USENIX WOOT 2015
- The Register, Corvette telematics dongle research (August 2015)
- Tencent Keen Security Lab, Remote attack to Tesla cars (September 2016)
- Electrek, Tesla pushed a fix (September 2016)
- Tencent Keen Security Lab, Experimental Security Assessment of BMW Cars (May 2018)
- CVE Program, CVE-2018-9311
- Francillon, Danev and Capkun, Relay Attacks on Passive Keyless Entry and Start Systems in Modern Cars, NDSS 2011, and slides
- BleepingComputer, Tesla Model 3 and Y Bluetooth LE relay (May 2022)
- Bosch Mobility, Perfectly Keyless
- Ken Tindell, CAN Injection: keyless car theft (April 2023)
- Sam Curry, Web Hackers vs. The Auto Industry (January 2023)
- Sam Curry, Hacking Kia (September 2024)
- Sam Curry, Hacking Subaru (January 2025)
- The Hacker News, PerfektBlue Bluetooth vulnerabilities (July 2025)
- Zero Day Initiative, Pwn2Own Automotive 2024 day one, day two and day three results
- Zero Day Initiative, Pwn2Own Automotive 2025 final results
- Zero Day Initiative, Pwn2Own Automotive 2026 day one and day three and Master of Pwn results
- Wikipedia, CAN bus
- Android Open Source Project, What is Android Automotive
- Automotive Grade Linux, About
- AUTOSAR, Specification of Secure Onboard Communication
- Uptane, About and Adoptions
- itemis, UNECE R156
- Keyfactor, ISO/SAE 21434
- Upstream, The impact of UNECE R155 continues to expand (April 2024)
- Open Charge Alliance, Open Charge Point Protocol and OCPP 1.6 Security Whitepaper, 4th edition
- WinCC OA documentation, OCPP security
- Ampcontrol, OCPP security and security profiles
- The Hacker News, SaiFlow OCPP research (February 2023)
- Pen Test Partners, Smart car chargers: plug n play for hackers? (July 2021)
- Kaspersky Securelist, Remotely controlled EV home chargers (December 2018)
- CERT@VDE, VDE-2024-011, Phoenix Contact CHARX SEC
- Zero Day Initiative advisories ZDI-24-1053, ZDI-25-712 and ZDI-25-347
- CVE Program records: CVE-2026-8983, CVE-2026-8984, CVE-2026-8986, CVE-2026-24731, CVE-2026-25851, CVE-2026-26051, CVE-2026-26288, CVE-2026-27028, CVE-2026-27767, CVE-2026-27772, CVE-2026-26290, CVE-2026-22552 and CVE-2025-12357
- Köhler, Baker, Strohmeier and Martinovic, Brokenwire: Wireless Disruption of CCS Electric Vehicle Charging, NDSS 2023
- Electrive, Charging station hacking on the Isle of Wight (April 2022)
- The Register, Digital Charging Solutions data breach (September 2025)
- The Register, Ransomware hits EV charger maker (March 2026)
- Cybellum and Upstream homepages