Most smart home devices are not hacked through clever exploits. They are hacked through a default password nobody changed, a cloud account protected by a reused password, a peer-to-peer video service that trusts the wrong party, or firmware that stopped getting updates years ago. Each of those has a clear fix, and this guide walks through every one of them.
Security cameras, video doorbells, smart TVs, smart speakers, smart plugs and baby monitors are small Linux or real-time computers with microphones, cameras and a permanent connection to a cloud service. They sit inside your home network. When one is compromised, the attacker can see or hear inside the house, take over the account, or use the device as a stepping stone to your laptop and phone.
This guide is written for three groups of readers. Home users who want to know what is actually risky and what to change this weekend. IT and security teams in businesses that have cameras, smart TVs and speakers in offices, shops and meeting rooms. And the engineers and product security teams who build these devices and their apps, who will find the supply chain and compliance sections most useful. Jargon is defined on first use.
It is a defender's guide written from the attacker's point of view. For each attack class you will find how it works and why, what an attacker needs, documented real cases with CVE numbers where they exist, the impact, how to detect it and exactly how to defend against it. There are no exploit steps, payloads or tool command lines. A note on the law: testing a device you do not own, or do not have written permission to test, is illegal in most countries. Tools are named here only as context for authorised testing of your own equipment.
This post is part of a device security series. The companion guides cover how Wi-Fi gets hacked and how to secure it and how routers get hacked. For where device hardware and firmware come from, see how a laptop is built and fabless chip designers and the software supply chain.
Contents
- How do smart home devices get hacked? The short answer
- How do smart home devices work, and what software do they run?
- What is the attack surface of a smart home device?
- What are the main ways smart home devices get hacked?
- Attack 1: Why are default passwords still the biggest smart home risk?
- Attack 2: How do attackers take over smart home cloud accounts?
- Attack 3: Why are P2P and cloud relay services a weak point for cameras?
- Attack 4: How do firmware vulnerabilities in shared components spread?
- Attack 5: How do companion apps and cloud backends leak access?
- Attack 6: How do smart devices expose your privacy even when nobody hacks them?
- Attack 7: How is a smart device used as a foothold into the home network?
- Attack 8: What happens to devices that stop getting updates?
- What are the risks for each type of smart home device?
- How can you tell if a smart home device has been hacked?
- How should you segment a home network for smart devices?
- Do Matter and Thread make smart home devices secure?
- What do the UK PSTI Act and the EU Cyber Resilience Act require?
- Hardening checklists for home users, businesses and manufacturers
- Where does the smart device software supply chain break?
- How Safeguard helps smart device makers
- The takeaway
- Frequently asked questions
- Sources
How do smart home devices get hacked? The short answer
Smart home devices get hacked in eight main ways: default or hardcoded passwords, cloud account takeover through reused passwords, flawed peer-to-peer video relays, vulnerabilities in shared firmware components, bugs in companion apps and cloud backends, privacy leaks built into the product, use as a foothold into the home network, and devices abandoned without updates.
The pattern behind all eight is the same. A smart device is cheap to make, so it reuses the same software components across many brands. It talks to a cloud service you do not control. Its owner rarely logs in to it, so nobody notices when something changes. And it often outlives the company's interest in patching it.
That also tells you where the defences sit:
- The account is the front door for most people. A unique password and two-factor authentication stop the most common real-world attack.
- The device needs current firmware, no default credentials and as few open services as possible.
- The network decides how far an attacker gets if a device falls. A separate network segment for smart devices limits the damage.
- The maker decides how long the device is patched, which components it ships and whether vulnerabilities are fixed quickly.
The rest of the guide goes deep on each attack and each layer.
How do smart home devices work, and what software do they run?
A smart home device is a small computer built around a system on chip (SoC), running embedded Linux or a real-time operating system, with vendor and third-party libraries on top, a local web or control service, and an agent that keeps a permanent connection to the maker's cloud. A phone app controls it through that cloud.
Here is what that looks like for each device type in this guide.
- Security cameras and video doorbells pair an image sensor and microphone with a video encoder on the SoC. Most run embedded Linux. They stream video to the maker's cloud for recording and to your phone for live view, often through a peer-to-peer (P2P) connection service so that you can watch without opening ports on your router. Doorbells add a button, a speaker and sometimes a chime unit.
- Smart TVs are full computers. They run a large operating system such as Tizen, webOS, Android TV, Google TV, Fire OS or Roku OS, with an app store, a browser engine, media decoders and advertising and analytics services. Our post on how a smart TV is made covers the hardware.
- Smart speakers and displays add microphone arrays and a low-power wake word processor that listens locally for the trigger phrase, then stream audio to the cloud for speech recognition. They often also act as hubs for other devices. See how a smart speaker is made.
- Smart plugs and switches are the simplest. A Wi-Fi or Thread microcontroller drives a relay. Many run a small RTOS (real-time operating system) and expose a local control protocol such as UPnP (Universal Plug and Play) as well as a cloud connection.
- Baby monitors are cameras with two-way audio, temperature sensors and sometimes breathing or sleep tracking. Their software stack is close to a security camera, often built on the same P2P SDKs (software development kits).
Figure 1. The software stack inside a typical camera, doorbell or plug. Most of the code is written by someone other than the brand on the box.
Reading that stack from the bottom up:
- Hardware and boot. The SoC, flash memory and radios. A bootloader such as U-Boot loads the operating system. Secure boot, where the chip only runs signed code, is present on some devices and absent on many cheap ones.
- Operating system. Usually a Linux kernel with a small userland (often BusyBox, a single binary that provides many standard tools), or an RTOS on microcontroller devices.
- Open source libraries. TLS libraries such as OpenSSL, mbedTLS or wolfSSL, curl, web servers, JSON parsers, media libraries and compression code. These are where many shared vulnerabilities live.
- Vendor SDKs. Chip vendor board support packages, Wi-Fi drivers, and third-party SDKs for P2P video, cloud connection, voice and analytics. The brand usually receives these as binaries.
- Device services. A local web interface, RTSP (Real Time Streaming Protocol) video, UPnP, a setup access point, an OTA (over-the-air) update client and the cloud agent.
- Cloud and app. The maker's backend APIs, storage for recordings and thumbnails, the account system, and Android and iOS companion apps.
Every layer has its own suppliers and its own update cycle. That is why a single vulnerability in a P2P SDK can sit in cameras from dozens of brands, and why the brand on the box often cannot fix it alone. The supply chain section later in this guide comes back to that.
What is the attack surface of a smart home device?
The attack surface of a smart home device is every way to reach its code or its data: the cloud account, the maker's cloud and P2P relay, the companion app, the local network services, the radio links used during setup, the firmware update channel, the physical ports on the board and the third-party code inside it.
Figure 2. The eight places attackers reach a smart home device. Most real incidents start at the top two: the account and the cloud.
The surface splits into remote, nearby and physical parts. The difference matters because it tells you who can attack.
| Surface | Who can reach it | Typical weakness |
|---|---|---|
| Cloud account | Anyone on the internet | Reused password, no two-factor authentication |
| Maker's cloud and P2P relay | Anyone on the internet | Weak device authentication, predictable device IDs, missing encryption |
| Companion app | Anyone who can get a malicious app or link onto the phone | Exported app components, insecure deep links, leaked API keys |
| Local network services | Anyone on the same network, or the internet if exposed | Default credentials, command injection in web interfaces, UPnP |
| Radio and setup mode | Anyone within Wi-Fi or Bluetooth range | Unencrypted setup traffic, open setup access points |
| Firmware update channel | Whoever controls the update server or the network path | Unsigned updates, update over plain HTTP |
| Physical ports | Anyone holding the device | UART or JTAG debug ports, unencrypted flash |
| Supply chain components | The component's own attackers | Vulnerable SDKs and libraries shipped in many products |
Remote surfaces are the ones that matter most for a home. An attacker on another continent can try a password list against your camera account or probe an exposed web interface. Nearby and physical attacks need someone close to your house, which is a real but much smaller group.
What are the main ways smart home devices get hacked?
The eight attack classes below cover nearly every publicly documented smart home compromise. Default passwords and account takeover need the least skill and cause most real incidents. P2P, firmware and app flaws need more skill but can hit millions of devices at once. Abandonment turns every unpatched flaw into a permanent one.
Figure 3. A qualitative view of attacker effort against scale of impact. It is a judgement drawn from the cases in this guide, not a measured dataset.
The timeline below shows that none of this is new. The same classes have repeated for a decade.
Figure 4. Ten years of smart home security incidents and research, with the regulatory milestones that followed.
Each of the following sections uses the same structure: how it works, what the attacker needs, real cases, impact, detection and defence.
Attack 1: Why are default passwords still the biggest smart home risk?
Default passwords are dangerous because they are identical across thousands of devices and published in manuals, so one list unlocks a whole product line. The Mirai botnet used a fixed set of 64 common default logins and grew to about 600,000 infected devices in 2016. Universal default passwords are now banned in the UK.
How it works and why
A device ships with a username and password set at the factory, such as admin and admin, or a password printed on a label that is the same for every unit of that model. Some devices also have hidden accounts that the owner cannot see or change, often left for support or manufacturing. If any of these accounts can log in over the network, through a web interface, Telnet or SSH, anyone who knows the password can log in too.
It keeps working because owners rarely know the account exists, many devices never force a change at setup, and hidden accounts cannot be changed at all.
What the attacker needs
A list of common default credentials and network access to the login service. For internet-exposed devices that means nothing more than an internet connection, because automated scanners find exposed login services within minutes.
Real cases
- Mirai (2016). The Mirai botnet scanned the internet for devices with open Telnet ports and tried a fixed list of 64 well-known default username and password pairs, according to Cloudflare's retrospective. The academic study Understanding the Mirai Botnet, presented at USENIX Security 2017, traced its growth to a peak of about 600,000 infections. Cloudflare's analysis of device banners suggests most infected devices were routers and cameras. On 21 October 2016 a Mirai attack on the DNS provider Dyn left users unable to reach many popular sites, including Amazon, Netflix, Twitter and Reddit. Our post on Mirai and default credentials covers the details.
- Baby monitors (2015). Rapid7 tested nine internet-connected baby monitors from eight vendors priced between 55 and 260 US dollars. In its disclosure FAQ the company reported hidden, hardcoded account credentials giving local or remote access, unencrypted video streaming, and unencrypted web and mobile app functions. Eight of the nine devices scored an F and one scored a D. One example is CVE-2015-2887: the iBaby M3S had a password of admin for a backdoor admin account.
- P2P cameras (2019). In the iLnkP2P research covered below, KrebsOnSecurity reported that many affected devices also shipped with factory default passwords, which turned a connection flaw into full access.
Impact
Full control of the device. For a camera or baby monitor that means live video and audio. For any device it means the attacker can run their own code, recruit it into a botnet for denial-of-service attacks, or use it to attack other machines on the network.
Detection
- Check whether you can log in to the device's local web page with the default credentials from its manual. If you can, so can anyone else who reaches it.
- Look for devices on your router's client list that you do not recognise or that use far more upload bandwidth than expected. Botnet traffic is mostly outbound.
- Infected devices may run slowly or reboot, but many show no visible sign at all, which is why detection has to come from the network side.
Defence
- Change every default password at setup to a long, unique one stored in a password manager. If the device has no way to change it, treat it as untrusted and keep it off the internet.
- Turn off Telnet, SSH and remote web access on devices that offer the option and do not need it.
- Never forward router ports to a camera or recorder. Use the maker's app or a VPN to your home network instead.
- Disable UPnP on the router so devices cannot open ports for themselves. Note that this does not stop P2P services that use hole punching, covered in Attack 3.
- Buy devices that ask you to set a password at first use or ship with a unique random password per unit. In the UK that is now the law, covered later in this guide.
Our short guide to changing default passwords on smart devices has practical advice on choosing them.
Attack 2: How do attackers take over smart home cloud accounts?
Attackers take over smart home accounts mainly through credential stuffing: they replay usernames and passwords leaked from other websites against the camera, TV or speaker maker's login page. Anyone who reused a password is exposed. Roku reported about 576,000 accounts affected this way in 2024, and responded by enabling two-factor authentication for every account.
How it works and why
When any website is breached, its users' email addresses and passwords end up in large lists. Automated tools try those pairs against other services. Because many people use the same password everywhere, a small percentage succeed. Credential stuffing is not a hack of the smart home company's systems. It is an attack on password reuse.
For a smart home account the payoff is high. The account is the control plane for every device. Whoever holds it can watch live video, download recordings, talk through a doorbell speaker, change settings, add their own devices or make purchases with a stored card.
Figure 5. Credential stuffing in four steps, and the three points where defences break the chain.
What the attacker needs
A list of leaked credentials, which is easy to obtain, and an automated way to submit logins while dodging rate limits. No access to your home and no knowledge of your devices.
Real cases
- Roku (2024). In April 2024 Roku disclosed a second incident affecting about 576,000 accounts, after an earlier one affecting about 15,000, according to Security Affairs. Roku said there was no indication its own systems were the source of the credentials, which pointed to reuse from other breaches. In fewer than 400 cases attackers made purchases with stored payment methods, without access to full card numbers. Roku reset affected passwords and enabled two-factor authentication by default for all customer accounts.
- Ring (2020). Ring made two-factor authentication mandatory for all users on 18 February 2020, after what 9to5Google described as serious concerns over the previous year. Every login now needs a one-time six-digit code by email or SMS, and the rule covers shared users. Google had announced a similar requirement for Nest accounts about a week earlier.
The lesson from both is that the companies with the most exposure moved to mandatory second factors. Where your devices' maker has not, you can usually switch it on yourself.
Impact
Live video and audio from inside the home, access to stored recordings, the ability to speak through devices, purchases, and personal details such as home address and household members' names.
Detection
- Turn on login alerts or new device notifications in each smart home app.
- Review the list of signed-in phones, browsers and shared users. Remove anything you do not recognise.
- Watch for changed settings, new shared users, motion alerts you did not trigger, or the camera's status light turning on when nobody in the house is viewing.
- Check whether your email address appears in known breach lists, and change any password you have reused.
Defence
- Use a unique password for every smart home account, generated by a password manager.
- Turn on two-factor authentication, preferably with an authenticator app or passkey rather than SMS where the app offers that choice.
- Limit shared users to people who need access, and remove former household members, babysitters and installers.
- Remove stored payment cards from device accounts you do not buy through.
- For businesses, put cameras and TVs under a company account with single sign-on where the vendor supports it, never under an employee's personal email.
Attack 3: Why are P2P and cloud relay services a weak point for cameras?
P2P relay services let you watch a camera from anywhere without opening router ports. They work by giving each device a unique ID and a cloud lookup service that connects your phone to it. When those IDs are predictable, device registration is not authenticated, or traffic is weakly encrypted, attackers can find, impersonate or eavesdrop on cameras at scale.
How it works and why
A camera behind a home router cannot normally accept connections from the internet. P2P services solve this. At boot the camera registers its unique identifier (UID) with the provider's servers. Your app asks the same servers where that UID is, and the servers broker a direct connection, often through UDP hole punching (a technique where both sides send packets outward so the router lets the reply in), or relay the traffic when a direct path fails.
Figure 6. How a P2P camera service works, and the three places it has failed in published research.
This design concentrates trust in three things: the secrecy of the UID, the authenticity of whoever registers it, and the encryption of the stream. Published research has broken each of them. And because a handful of SDK providers supply many camera brands, a single flaw lands in millions of devices that carry different logos.
What the attacker needs
It depends on the flaw. For enumeration flaws, only knowledge of the ID scheme. For impersonation, the victim's UID plus knowledge of the protocol. For interception, a position on the network path. In every case the attacker does not need to be near your home.
Real cases
- iLnkP2P (2019). Researcher Paul Marrapese found two flaws in iLnkP2P, P2P software from Shenzhen Yunni Technology bundled into cameras, baby monitors and doorbells from many white-label brands. CVE-2019-11219 is a predictability flaw in how UIDs are generated that allows remote attackers to connect directly to arbitrary devices. CVE-2019-11220 is an authentication flaw that allows attackers to intercept user-to-device traffic in cleartext, including video and device credentials. KrebsOnSecurity reported more than 2 million vulnerable devices online, no vendor patches, and that disabling UPnP on the router does not help because the devices use hole punching. Because UIDs are assigned at manufacture, Marrapese judged software fixes unlikely.
- ThroughTek Kalay, interception (2021). Nozomi Networks found that an older ThroughTek P2P SDK used an obfuscation scheme with a fixed key instead of a secure key exchange, so anyone able to capture the traffic could rebuild the audio and video. It is tracked as CVE-2021-32934, covering SDK versions before 3.1.5, builds with the nossl tag, and firmware not using AuthKey or DTLS (Datagram Transport Layer Security). ThroughTek's fix was a DTLS based encryption layer and guidance to enable the SDK's security features.
- ThroughTek Kalay, impersonation (2021). Mandiant disclosed CVE-2021-28372, rated 9.6 under CVSS 3.1. An attacker who knew a device's 20-byte UID could register as that device on the Kalay network, receive the connections meant for it and capture the username and password the owner's app sent. With those, the attacker could watch live audio and video. At the time ThroughTek advertised more than 83 million active devices. Mandiant judged brute forcing UIDs infeasible; they would have to be obtained through social engineering or a weak web API. The fix was to upgrade the SDK and enable AuthKey and DTLS.
Impact
Remote viewing of cameras and baby monitors, theft of device passwords, and in the worst cases a route to full device compromise. Because the attack goes through the provider's cloud, a home firewall that blocks inbound traffic offers no protection.
Detection
This is hard for an owner to detect, because the traffic looks like normal cloud traffic. What you can do:
- Find out whether your camera uses a P2P service. Signs include a UID or device code on the label and remote viewing that works with no port forwarding or account linking.
- Check the maker's security advisories for the SDK issues named above and confirm your firmware version is newer than the fixed release.
- Businesses can look for camera traffic to P2P relay servers in firewall logs and compare it against an approved vendor list.
Defence
- Prefer cameras from makers that publish security advisories, a vulnerability disclosure policy and a defined support period. Those makers fixed the Kalay issues; many white-label brands did not.
- Keep camera firmware current, and turn on automatic updates.
- Keep the device UID private. Do not post photos of the label or share the code outside the household.
- For cameras that will never be fixed, consider blocking their outbound P2P traffic at the firewall and recording locally to a network video recorder on an isolated segment. Marrapese suggested blocking outbound UDP port 32100 for iLnkP2P devices specifically, which stops remote viewing through that service.
- Manufacturers should use P2P SDKs with mutual authentication and encryption switched on, never ship debug or nossl builds, and track the SDK version in every firmware image's SBOM (software bill of materials, the list of components inside).
Attack 4: How do firmware vulnerabilities in shared components spread?
Firmware vulnerabilities spread because device makers build on the same SDKs, operating system images and libraries. A command injection or buffer overflow in a shared component appears in every product that ships it. Bitdefender's 2024 ThroughTek Kalay research hit cameras from Wyze, Roku and Owlet at once, and the Kalay platform runs on more than 100 million devices.
How it works and why
Most device firmware bugs fall into a few classes:
- Command injection, where input from the network is passed to a system shell without being cleaned, so an attacker can add their own commands.
- Memory corruption, such as stack-based buffer overflows in C code that parses network messages or configuration fields.
- Authentication bypass, where a service fails to check who is calling it.
- Missing message authentication, where a device accepts commands from anything that claims to be its server.
Device services frequently run as root, the all-powerful system account. Mandiant noted that the Kalay binaries it examined often ran as root and lacked standard memory protections such as ASLR, PIE, stack canaries and NX. On those devices a single bug gives full control.
The supply chain multiplies this. A brand licenses an SDK, a chip vendor's software package and an open source base, often as binaries, then ships them for years. When a flaw appears upstream, every brand must notice, get a fixed component, rebuild, test and push an update. Many never do.
What the attacker needs
Usually network access to the vulnerable service, which may be the local network or the internet if the service is exposed or reachable through a relay. Some bugs need authentication first, which is why chains of vulnerabilities matter: one bug supplies the access and another the code execution.
Real cases
- ThroughTek Kalay (2024). Bitdefender reported four vulnerabilities affecting the Owlet Cam v1 and v2, Wyze Cam v3 and Roku Indoor Camera SE, and said they affect most implementations of the platform. CVE-2023-6323: the SDK did not verify the authenticity of received messages, so an attacker could impersonate the server. CVE-2023-6324: a predictable pre-shared key in the DTLS session. CVE-2023-6322: a stack-based buffer overflow in message parsing. CVE-2023-6321: command injection as root in the handler for OTA updates. Chained, they let an attacker on the local network get root on the camera. Bitdefender reported the issues in October 2023, ThroughTek confirmed all affected SDK versions were patched in April 2024, and the research was published on 15 May 2024. See our post on CVE-2023-6324.
- Hikvision cameras (2021). CVE-2021-36260 is a command injection vulnerability in the web server of some Hikvision products, caused by insufficient input validation. Our CVE-2021-36260 post has the affected product detail.
- LG webOS smart TVs (2024). Bitdefender found four vulnerabilities in LG's TV operating system, reported by The Hacker News. CVE-2023-6317 bypassed PIN verification to add a privileged user, CVE-2023-6318 escalated to root, CVE-2023-6319 allowed command injection through a library that displays music lyrics, and CVE-2023-6320 allowed authenticated command injection through a network configuration endpoint. The vulnerable service was meant for the local network only, yet more than 91,000 TVs exposed it to the internet. LG shipped fixes on 22 March 2024. See our post on CVE-2023-6320.
- Belkin Wemo smart plug (2023). CVE-2023-27217 is a stack-based buffer overflow in the function that renames a Wemo Mini Smart Plug V2 (F7C063), reachable through a crafted UPnP request. Attack 8 covers why it was never fixed.
Impact
Code execution on the device, usually as root. From there the attacker owns the camera feed or microphone, can persist across reboots on devices without secure boot, and can attack the rest of the network.
Detection
- Know the exact model and firmware version of every device. Without that inventory you cannot match advisories to what you own.
- Subscribe to the maker's security advisories, or check them monthly.
- Use your router or firewall to see which devices expose services to the internet. A TV or camera with an open port is a warning sign, as the LG case shows.
- Businesses can run authorised vulnerability scans of their own device networks with a network scanner to find exposed services and outdated firmware.
Defence
- Turn on automatic firmware updates, and check after major advisories that the update actually applied.
- Never expose device services to the internet. Close port forwards and turn off UPnP.
- Segment devices so a compromised camera cannot reach laptops and phones, covered in detail below.
- Manufacturers need a component inventory per firmware build, a process to rebuild and ship when an SDK or library is fixed, privilege separation so services do not run as root, standard memory protections enabled in the toolchain, and secure boot with signed updates.
Attack 5: How do companion apps and cloud backends leak access?
Companion apps and cloud backends leak access when the app trusts other apps on the phone too much, when setup sends secrets in clear text, or when the cloud mixes up which data belongs to which user. These bugs sit outside the device, so firmware updates cannot fix them. In 2024 a Wyze caching bug showed about 13,000 users thumbnails from other people's cameras.
How it works and why
The phone app and the maker's cloud are where most of the logic lives. The app holds a long-lived login token. The cloud stores recordings, thumbnails and device mappings. Three kinds of bug recur:
- App component exposure. Android apps can make parts of themselves callable by other apps. If a sensitive component is exported, or a deep link accepts untrusted input, a malicious app on the same phone can drive it.
- Insecure setup. During setup the app must hand the device your Wi-Fi password. If that exchange is not encrypted, anyone nearby can capture it.
- Backend authorisation errors. If the cloud gets the mapping between user, device and data wrong, users see each other's footage. Caching layers are a common source of these mix-ups.
What the attacker needs
For app flaws, a malicious app installed on the victim's phone. For setup flaws, radio range of the device during setup. For backend errors, often nothing at all: the leak happens to whichever users are unlucky.
Real cases
- Ring Android app (2022). Checkmarx researchers reported a chain of vulnerabilities in the Ring Android app and an Amazon website on 1 May 2022. According to The Record, a malicious app on the user's phone could have used them to obtain the user's name, email, phone number, geolocation, address and camera recordings. Amazon fixed it on 27 May in app version 3.51.0 for Android and 5.51.0 for iOS, and said no customer data was exposed and the issue would be extremely difficult to exploit.
- Ring Video Doorbell Pro setup (2019). Bitdefender found that during setup the app and the doorbell exchanged the home Wi-Fi password over plain HTTP. An attacker nearby could knock the doorbell off the network, wait for the owner to run setup again, and capture the password. Amazon fixed it by encrypting the setup exchange and pushing an automatic update.
- Wyze (2024). After an outage in February 2024, about 13,000 Wyze users received thumbnails from cameras that were not their own, and 1,504 tapped on them, according to Engadget. In some cases a full event video could be viewed. Wyze attributed it to a mix-up of device ID and user ID mapping in a newly added third-party caching library under heavy load.
- Eufy web portal (2023). Anker admitted that Eufy camera streams opened through its web portal were not encrypted and could be played in third-party players such as VLC, according to MacRumors. It blocked the debug mode used to find stream links, encrypted the streams and began moving cameras to WebRTC, which is encrypted by default.
Impact
Exposure of recordings and live views without touching the device, theft of the home Wi-Fi password, and leakage of personal details. A backend bug can affect thousands of homes at once.
Detection
Owners mostly learn about these from the vendor or the news. Keep the app updated, read the vendor's security notices, and review which other apps on your phone have unusual permissions. Businesses with mobile device management can block side-loaded apps on phones that run camera and building apps.
Defence
- Update companion apps as soon as updates arrive, and remove apps for devices you no longer own.
- Install apps only from official stores, and be wary of apps that ask for accessibility or overlay permissions.
- Run device setup at home, not in a public place, and redo it only when you initiated the change. If a device drops off Wi-Fi and asks for setup unexpectedly, be suspicious.
- Manufacturers should run static analysis on app code, test exported components and deep links, encrypt every setup exchange, enforce object-level authorisation in every API, and test cache layers for cross-tenant leaks. Dynamic testing of the cloud APIs belongs in the release process.
Attack 6: How do smart devices expose your privacy even when nobody hacks them?
Many privacy exposures are design choices, not hacks. Cameras upload thumbnails to the cloud, smart TVs identify what is on screen through automatic content recognition, and speakers send audio to cloud services. The data then exists on servers you do not control, where breaches, bugs, staff access and data sharing can reach it.
How it works and why
A smart device's business model often depends on cloud services and data. A camera's motion alert with a preview image needs that image on a server. A TV maker can earn money from viewing data. A voice assistant needs cloud speech recognition. None of this is an attack, but it widens the attack surface and moves sensitive data out of the home.
Real cases
- Smart TV automatic content recognition. Automatic content recognition (ACR) is a feature that captures what is on screen and matches it against a database to identify the content. A 2024 study presented at the ACM Internet Measurement Conference, Watching TV with the Second-Party, audited Samsung and LG TVs. It found ACR works even when the TV is used only as a display for an external HDMI source such as a game console or laptop. Opting out in the settings stopped the ACR network traffic. Behaviour differed between the UK and the US.
- Eufy cloud uploads. Security consultant Paul Moore found Eufy cameras sending data to the cloud even with cloud upload settings disabled, as summarised by MacRumors. Anker acknowledged that the cameras did not offer the native end-to-end encryption it had marketed.
- Google Home speaker (2021). Researcher Matt Kunze found that an attacker within wireless range could add a backdoor account to a Google Home speaker, then use routines to set the volume to zero and place a call to a number they controlled, turning the microphone into a listening device, according to The Hacker News. The only visible sign was the LEDs turning solid blue. Disclosed on 8 January 2021 and fixed in April 2021, it earned a 107,500 US dollar bounty. Google's fixes added an invite-based account linking process and blocked remote call commands through routines.
Impact
Viewing habits, voices, faces and home layouts held on third-party servers, available to whoever breaches or misuses them. For businesses, meeting room TVs and speakers can capture confidential discussions.
Detection
Read the privacy settings of each device. On TVs look for viewing information, content recognition or similar settings. On speakers review the voice history in the app. Router traffic graphs show which devices talk to the internet most.
Defence
- Turn off ACR and advertising tracking on smart TVs. The study above shows opting out works.
- Prefer local storage and end-to-end encryption for cameras, and confirm what end-to-end actually covers.
- Mute or unplug speakers in rooms where sensitive conversations happen, and delete stored voice history on a schedule.
- Point cameras away from neighbours and private spaces, and avoid indoor cameras in bedrooms and bathrooms.
- Businesses should treat meeting room TVs and speakers as recording devices in their data protection assessments.
Attack 7: How is a smart device used as a foothold into the home network?
Once an attacker controls one smart device, it sits inside your network with the same access as your laptop. It can scan for other devices, attack weak services, intercept traffic or relay commands from outside. A doorbell that leaked the Wi-Fi password, or a speaker that could make HTTP requests inside the network, are documented examples.
How it works and why
Home networks are usually flat: every device can talk to every other device. Firewalls on the router face outward and do nothing about traffic between a camera and a laptop. Many home devices, including network storage, printers and other smart devices, trust anything on the local network and protect their admin pages weakly or not at all.
So a compromised plug or camera becomes a launch point. Because smart devices rarely run security software and their owners rarely look at them, an attacker can stay there unnoticed.
What the attacker needs
Control of any one device on the network, or the Wi-Fi password itself.
Real cases
- Ring doorbell Wi-Fi password leak (2019). The Bitdefender research above shows the simplest version: steal the Wi-Fi password from setup traffic and you are on the network. Bitdefender noted that the attacker could then reach smart assistants, network storage and computers with weak credentials.
- Google Home backdoor account (2021). Besides listening, Kunze found the linked attacker account could make arbitrary HTTP requests inside the victim's network, according to The Hacker News. A speaker became a proxy into the LAN.
- Mirai and its descendants. Botnets use infected devices to scan and attack other devices, both on the internet and on the local network.
Impact
Compromise of computers, phones, network storage and work devices on the same network. For people who work from home, a compromised smart device can become a route toward company systems.
Detection
- Look for devices talking to each other in ways they should not. A smart plug has no reason to connect to a laptop.
- Routers and firewalls with intrusion detection can flag internal scanning.
- Unexpected open services on smart devices, found with an authorised scan of your own network, suggest tampering.
Defence
- Segment the network so smart devices cannot start connections to computers and phones. The next sections give a practical layout.
- Use WPA3 or WPA2 with a strong passphrase, and change it if you suspect a device leaked it. The Wi-Fi guide in this series covers this in depth.
- Protect admin pages on network storage, printers and routers with strong passwords even inside the home.
- Keep work laptops on their own network or VPN rather than the same flat network as smart devices.
Attack 8: What happens to devices that stop getting updates?
When a maker stops updating a device, every vulnerability found afterwards stays open for as long as the device is plugged in. Belkin declined to patch a 2023 Wemo smart plug flaw because the model was end of life, then ended cloud and app support for most Wemo devices on 31 January 2026. Devices like these still work, but can no longer be secured.
How it works and why
Smart devices are cheap and stay in use for years. Support costs the maker money long after the sale. So support quietly ends, or the company is acquired, changes strategy or goes out of business. The device keeps working, but new vulnerabilities in its firmware, SDKs and libraries are never fixed. For devices that depend on the maker's cloud, the end of cloud service can also stop them working, or leave them in a half-working state.
What the attacker needs
The same access as for any other vulnerability. The difference is time: an unpatched flaw never expires, and exploit details become public after disclosure.
Real cases
- Wemo Mini Smart Plug V2 (2023). Sternum researchers found the buffer overflow tracked as CVE-2023-27217. Belkin told them the device was at the end of its life and would not be patched, according to 9to5Mac. Sternum estimated hundreds of thousands of V2 units in use. Its advice was to keep the plug's UPnP ports off the internet, segment it from sensitive devices, and stop using it if those steps were not enough.
- Wemo cloud shutdown (2026). In July 2025 Belkin announced it would end cloud services and app support for older Wemo devices released between August 2015 and November 2023, effective 31 January 2026, according to MacRumors. Remote access and Alexa and Google voice control stopped. Devices set up with Apple HomeKit before the deadline kept working through HomeKit, and four Thread-based Wemo devices were not affected.
- iLnkP2P devices (2019). As noted above, the UIDs were fixed at manufacture and no vendor patched the flaws, which made those devices permanently exposed.
Impact
A growing set of known, public vulnerabilities on a device that will never be fixed, often still connected to the internet.
Detection
- Check each device's support page for an end-of-support or end-of-life date. Our guide to checking how long a smart device gets security updates explains where to look.
- If the last firmware update was more than a year ago, ask the maker whether the model is still supported.
Defence
- Replace devices that no longer get security updates, starting with cameras, baby monitors and anything with a microphone.
- If you must keep one, isolate it on its own network segment, block its internet access if it can work locally, and remove any port forwarding. Our post on what to do with a smart device that cannot be updated covers the options.
- Before you buy, check the published support period. In the UK it must be published, and in the EU the Cyber Resilience Act will set a minimum.
- Prefer devices that work locally, through Matter, HomeKit or local control, so a cloud shutdown does not turn them into e-waste.
What are the risks for each type of smart home device?
Cameras, doorbells and baby monitors carry the highest privacy risk because they see and hear inside the home. Smart TVs and speakers combine large software stacks with data collection. Smart plugs carry the least data but are cheap, rarely updated and often abandoned, which makes them ideal footholds.
| Device | Most likely attack | Highest impact | First three defences |
|---|---|---|---|
| Security camera | Account takeover, P2P flaws, exposed web interface | Live video from inside the home | Unique password plus two-factor, current firmware, no port forwarding |
| Video doorbell | Account takeover, setup traffic capture | Video of visitors, Wi-Fi password theft | Two-factor, encrypted setup from a current app, separate segment |
| Smart TV | Exposed services, ACR data collection, old OS builds | Viewing data, foothold on the network | Turn off ACR, automatic updates, never expose to the internet |
| Smart speaker | Account linking abuse, voice app misuse | Audio from inside the home | Review linked accounts, mute in sensitive rooms, update |
| Smart plug | Default credentials, UPnP bugs, abandonment | Foothold, power control of what is plugged in | Segment, disable UPnP, replace end of life models |
| Baby monitor | Hardcoded credentials, P2P flaws, account takeover | Video and audio of a child's room | Buy from makers with advisories and support periods, two-factor, update |
Baby monitors deserve a note. They are often bought as gifts, set up quickly and never revisited, and they use the same P2P SDKs as cheap cameras. The Rapid7 study and the Owlet Cam results in the 2024 Kalay research show the risk is real. A monitor that only works on the local network, without a cloud account, removes the most common attacks entirely.
How can you tell if a smart home device has been hacked?
Signs that a smart device may be compromised include unfamiliar logins or shared users in the app, settings that changed on their own, the camera light turning on when nobody is viewing, voices or sounds from the speaker, unusual upload traffic, and the device rebooting or slowing down. Most are visible in the app and on your router.
A practical check, in order:
- Account first. Open each smart home app. Review signed-in devices, shared users, linked services and login history where available. Remove anything unknown, change the password, and turn on two-factor authentication.
- Device settings. Look for changed recording settings, new schedules or routines, disabled notifications, or a changed Wi-Fi network.
- Indicators on the device. A camera status light that turns on when nobody is watching, a pan and tilt camera that moves by itself, or sounds from a speaker you did not trigger all need checking.
- Router view. Your router's admin page or app lists connected devices and often traffic per device. Look for unknown devices, and for smart devices sending far more data than usual. A doorbell normally uploads in bursts when there is motion, not constantly.
- Firmware version. Compare the installed version with the latest on the maker's site. An old version means known vulnerabilities are open, whether or not they have been used.
If you find evidence of compromise:
- Change the account password and enable two-factor authentication.
- Remove unknown shared users and sign out all sessions.
- Factory reset the device, update its firmware, and set it up again with new credentials.
- Change the Wi-Fi password if the device could have exposed it.
- If you see signs that the device was used to reach other computers, check those too.
For businesses, the same checks scale with tooling. Keep an inventory of every camera, TV, speaker and controller by model and firmware version. Feed firewall logs into your monitoring. Alert on smart devices starting connections to internal systems, and on traffic to destinations outside each vendor's normal cloud.
How should you segment a home network for smart devices?
Put smart devices on their own network segment, either a separate Wi-Fi network or a VLAN, and set firewall rules so they can reach the internet but cannot start connections to your computers and phones. Your trusted devices can still reach them. This contains a compromised camera or plug without breaking most apps.
A VLAN (virtual LAN) splits one physical network into separate logical networks, each with its own addresses and firewall rules. Many home routers do not offer VLANs, but most offer a guest network, which is a simpler version of the same idea.
Figure 7. A practical three-segment home network. Green marks allowed connections; red marks what the firewall blocks.
Three levels of segmentation
Level 1: guest network (any router). Create a guest Wi-Fi network for smart devices, with client isolation turned on if available. Most guest networks block access to the main network. The drawback is that apps on your phone may not find devices on another network for local control or casting, so features like screen casting to a TV may stop working.
Level 2: separate IoT network with rules (better routers and mesh systems). Some consumer routers and mesh systems offer a dedicated IoT network. These usually allow controlled access from your main network while blocking the reverse.
Level 3: VLANs with a firewall (prosumer and small business gear). With a router or firewall that supports VLANs, create at least three segments:
| Segment | Devices | Can reach |
|---|---|---|
| Trusted | Laptops, phones, tablets | The internet and the IoT segment |
| IoT | Cameras, TVs, speakers, plugs, hubs | The internet, and replies to connections started by trusted devices |
| Guest | Visitors' devices | The internet only |
Then add these firewall rules in plain terms:
- Block new connections from IoT to trusted.
- Allow new connections from trusted to IoT, and the replies, so your phone can control devices.
- Block IoT access to the router's admin interface.
- Allow IoT to reach the internet, and for high-risk devices limit this to the vendor's services.
- Block devices that work fully locally, such as a network video recorder, from the internet entirely.
- Send DNS for the IoT segment through a resolver you control, so you can see and filter where devices connect.
Our post on building a home firewall on a mini PC with VLANs and DNS filtering walks through a full build.
Making discovery work across segments
Casting, AirPlay, HomeKit and Matter use multicast discovery (mDNS, the protocol devices use to announce themselves on the local network) to find devices. Multicast does not cross VLANs by default. Two common options:
- Turn on an mDNS reflector or repeater on the router so discovery messages cross between the trusted and IoT segments, while the firewall still controls actual connections.
- Keep hubs and controllers such as a Matter controller, HomeKit hub or Thread border router on the same segment as the devices they control.
Matter relies on IPv6 on the local network, so make sure IPv6 is enabled on the IoT segment if you use Matter devices.
For businesses
Treat smart devices in offices, shops and hotels the same way, with more rigour. Use a dedicated VLAN per device class, such as cameras, displays and building systems. Use network access control so only known devices join each VLAN. Allow only vendor destinations outbound. Never let these networks reach corporate servers directly. Cameras and recorders belong on a network that only the security team's viewing stations can reach.
Do Matter and Thread make smart home devices secure?
Matter and Thread improve smart home security a great deal: every Matter device has a cryptographic identity checked at setup, sessions are encrypted with AES-CCM, and Thread encrypts every hop of its mesh. They do not fix firmware bugs, weak cloud accounts or abandoned devices, so the other defences in this guide still apply.
Matter is a smart home connectivity standard from the Connectivity Standards Alliance (CSA). It runs over Wi-Fi, Ethernet and Thread. Thread is a low-power IPv6 mesh network built on the IEEE 802.15.4 radio standard, used by battery devices such as sensors and some plugs.
Figure 8. How a Matter device proves what it is and joins your home, and how Thread protects the radio link underneath.
How Matter security works
According to Silicon Labs' Matter security documentation:
- Device identity. Every Matter device carries a unique Device Attestation Certificate (DAC). It chains to a Product Attestation Intermediate (PAI) and up to a Product Attestation Authority (PAA). There is no single root across all devices.
- Certification. A Certification Declaration, signed by the CSA, states that the product passed Matter certification. The Distributed Compliance Ledger (DCL), a publicly readable ledger that only approved parties can write to, holds the certification status and trusted roots.
- Setup. The setup code or QR code carries a passcode. The phone and device use it in PASE (Passcode-Authenticated Session Establishment), based on SPAKE2+, to create an encrypted channel without sending the passcode.
- Attestation. Inside that channel the controller checks the device's certificate chain, matches it against the DCL, and makes the device prove it holds the matching private key. Only then does it hand over network credentials.
- Operational credentials. The device receives a Node Operational Certificate and an access control list for your home's "fabric", the Matter term for one controller's set of devices.
- Daily use. Devices and controllers use CASE (Certificate-Authenticated Session Establishment) with those operational certificates. Messages are encrypted with AES-CCM using 128-bit keys with elliptic curve cryptography on the P-256 curve.
- Updates. Matter devices must support over-the-air firmware updates, linked to secure boot so devices reject malicious images.
How Thread security works
The Thread Group's security white paper describes:
- Per-hop encryption. Thread always applies AES-CCM* with a 128-bit key and a 32-bit message integrity code at the MAC layer on every hop.
- Replay protection. Devices track frame counters for their neighbours and reject replayed or out-of-order frames.
- Secure commissioning. A new device joins through a DTLS session authenticated with EC-JPAKE using the device's joining credential, so the passphrase is never sent over the air, and then receives the network key.
- Border routers. A Thread border router links the mesh to Wi-Fi or Ethernet as an IPv6 router.
The white paper is clear about one limit: each mesh extender decrypts and re-encrypts frames as it forwards them. End-to-end protection needs a higher-layer protocol such as Matter CASE, DTLS, TLS or OSCORE.
What Matter and Thread do not fix
- Firmware vulnerabilities. A buffer overflow in a certified device is still a buffer overflow. Certification tests the protocol, not every line of the maker's code.
- Cloud accounts. Many Matter devices also link to a vendor cloud and app. Account takeover still works there.
- Abandonment. Matter requires update support, not a minimum number of years of patches.
- Network trust. A compromised Thread device holds the network key, and a compromised controller holds the fabric.
The real gain is local control. A Matter device can keep working through any compatible controller, as the Thread-based Wemo devices that survived Belkin's cloud shutdown showed. Our Matter myths and facts post goes further.
What do the UK PSTI Act and the EU Cyber Resilience Act require?
Since 29 April 2024 the UK PSTI regime bans universal default passwords, requires a published way to report vulnerabilities and requires a published security update period for consumer connectable products. The EU Cyber Resilience Act entered into force on 10 December 2024; its vulnerability reporting duties apply from 11 September 2026 and its main obligations from 11 December 2027.
These laws matter to buyers because they make three things visible before you buy: whether the device has a default password, where to report a flaw, and how long it will get security updates.
Figure 9. The UK and EU rules that now shape smart device security, and the standard they build on.
UK: Product Security and Telecommunications Infrastructure (PSTI) regime
According to techUK's guide, the regime came into force on 29 April 2024 and sets three requirements for relevant connectable products:
- No default passwords. Outside the factory default state, passwords must be unique per product or set by the user. Unique passwords must not be based on incremental counters, public information or product identifiers such as serial numbers.
- A vulnerability disclosure route. Manufacturers must publish at least one contact point for reporting security issues, with timeframes for acknowledging reports and giving status updates. This information must be free, in English and available without asking.
- A published support period. Manufacturers must publish the minimum time they will provide security updates, in terms a non-technical buyer can understand.
Desktop and laptop computers, tablets without cellular connectivity, smart meters, medical devices and electric vehicle smart charge points covered by other rules are exempt. Smart cameras, doorbells, TVs, speakers, plugs and baby monitors sold to UK consumers are in scope. Our PSTI engineering checklist covers what makers need to ship.
EU: Cyber Resilience Act (CRA)
The European Commission's CRA page says the regulation covers hardware and software products with digital elements, and lists baby monitors and smart watches as examples. Key dates and duties:
- 10 December 2024: the CRA entered into force.
- 11 September 2026: manufacturers must report actively exploited vulnerabilities. Guides such as noze's summary of the Commission's guidance describe an early warning within 24 hours and a notification within 72 hours.
- 11 December 2027: the main obligations apply, including the essential security requirements and technical documentation. Compliant products carry the CE marking.
- Support period: at least five years, aligned with how long the product is expected to be used.
- Penalties: up to 15 million euros or 2.5 percent of worldwide annual turnover for the most serious breaches, according to the Wikipedia summary.
The CRA expects manufacturers to know what is inside their products. An SBOM forms part of the technical documentation. Our posts on CRA SBOM requirements and SBOMs for firmware and IoT devices go into detail.
EU: Radio Equipment Directive cybersecurity rules
Wireless devices sold in the EU already face cybersecurity requirements under the Radio Equipment Directive. The Commission's RED page says Delegated Regulation (EU) 2022/30 entered into application in August 2025. A new delegated regulation repeals it from 11 December 2027, when the CRA takes over, to avoid overlapping requirements.
The common baseline: ETSI EN 303 645
Most of these rules build on ETSI EN 303 645, the European standard for consumer IoT security. Its thirteen provision groups start with "No universal default passwords" and run through vulnerability reporting, keeping software updated, secure storage of security parameters, secure communication, minimising attack surface, software integrity, personal data, resilience to outages, telemetry, easy deletion of user data, easy installation and maintenance, and input validation. Provision 5.1-1 requires that, outside the factory default state, consumer IoT device passwords are unique per device or defined by the user.
For a buyer the practical rule is simple. Look for a published support period and a vulnerability disclosure page before you buy, and prefer products that state compliance with these rules.
Hardening checklists for home users, businesses and manufacturers
Home users should secure the account, update the device and segment the network. Businesses should add inventory, central accounts, network access control and monitoring. Manufacturers should remove default passwords, keep an SBOM per firmware build, patch shared components quickly, sign updates and publish a support period.
Figure 10. Five layers of defence. An attacker has to get through all of them; you only need one to hold.
Figure 11. The checklists below, on one page.
Home users
Before you buy
- Choose makers that publish security advisories, a vulnerability disclosure page and a support period.
- Prefer devices that work locally, through Matter, HomeKit or a local hub, as well as through the cloud.
- Avoid cameras and baby monitors with no brand history or that only connect through generic P2P apps.
- See our ten questions to ask before buying a smart home device.
At setup
- Set a unique, long password for the account and any device login.
- Turn on two-factor authentication.
- Turn on automatic firmware and app updates.
- Run setup at home, from the official app.
- Turn off features you do not need: remote web access, UPnP, voice purchasing, ACR on TVs.
On the network
- Put smart devices on a guest network or IoT segment.
- Disable UPnP on the router and remove port forwards to devices.
- Use WPA3 or WPA2 with a strong passphrase.
Every few months
- Review shared users, linked accounts and signed-in devices.
- Check firmware versions against the maker's site.
- Replace devices that have reached end of support.
- Factory reset devices before selling or giving them away.
Businesses
- Inventory every camera, TV, speaker, display and controller with model, firmware version, location and owner.
- Procure only devices with a published support period and vulnerability disclosure route, and write update commitments into contracts.
- Accounts: use company accounts with single sign-on where available, never personal emails; enforce multi-factor authentication; remove leavers quickly.
- Network: dedicated VLANs per device class, network access control, outbound allow lists, no direct access to corporate servers.
- Monitoring: log device traffic, alert on internal scanning and unexpected destinations, and run authorised vulnerability scans of device networks.
- Patching: track firmware advisories and apply updates on a schedule, with exceptions recorded and isolated.
- Privacy: turn off ACR on displays, review microphone use in meeting rooms, and include cameras and speakers in data protection assessments.
- Lifecycle: replace devices before support ends, and wipe them before disposal.
Manufacturers
- Authentication: unique per-device credentials or user-set passwords; no hidden accounts; rate limiting on every login.
- Cloud and app: mandatory or default-on two-factor authentication; object-level authorisation on every API; encrypted setup; tested caching layers; static analysis and dynamic testing in every release.
- P2P and connectivity SDKs: authenticated, encrypted modes always on; no debug or nossl builds in production; current SDK versions tracked.
- Firmware: least privilege for services; ASLR, stack protection and NX enabled; secure boot; signed and encrypted updates; minimal exposed services.
- Supply chain: an SBOM for every firmware build and app release; continuous monitoring of every component for new vulnerabilities; a fast rebuild path when a component is fixed; supplier SBOMs required in contracts.
- Disclosure and support: a published vulnerability disclosure policy; a defined support period of at least five years in line with the CRA; advisories with fixed versions; a plan for local operation if the cloud is ever shut down.
- Compliance: map controls to ETSI EN 303 645, the UK PSTI regime and the EU CRA, and keep the evidence.
Where does the smart device software supply chain break?
The smart device supply chain breaks where code arrives as binaries from suppliers, where one SDK ships in many brands, where nobody tracks which component version is in which firmware, and where the fix depends on an upstream vendor's schedule. Most of the vulnerable code in a camera was written by someone other than the brand.
Look again at the cases in this guide. The iLnkP2P and Kalay vulnerabilities were in SDKs licensed by many brands. Bitdefender's 2024 Kalay research named three brands, but said the flaws affected most implementations of the platform. Mandiant could not compile a complete list of affected products because the SDK is built into OEM (original equipment manufacturer) products and apps. Nozomi noted that the number of affected products was hard to determine because many vendors had embedded the library over time. Wyze's 2024 incident started in a newly integrated third-party caching library.
That is a software supply chain problem, and it has four parts.
1. Nobody knows what is inside. A brand that buys a camera design from an ODM (original design manufacturer) may receive a firmware image with a chip vendor's Linux build, a P2P SDK and dozens of open source libraries, none with an SBOM. When an advisory names a component, the brand cannot answer "are we affected?" quickly.
2. One component, many products. Shared SDKs and libraries mean one vulnerability affects many brands at once. Good for attackers, who can reuse the same exploit across products.
3. Fixes wait on someone else. The brand often cannot change a binary SDK. It waits for the supplier, then for the ODM, then rebuilds. Every step adds weeks or months. Bitdefender's 2024 Kalay timeline ran from report in October 2023 to all SDK versions patched in April 2024, before device makers could ship their own updates.
4. Support ends before use does. The supply chain for a device spans years. When an SDK provider or chip vendor stops supporting an old version, every product built on it loses its fix path, as the Wemo plug shows.
The fix is the same discipline used in software companies: know every component in every build, watch them all for vulnerabilities, rank by real risk, fix fast in your own build, and keep the evidence. The next section shows how Safeguard does that. For more background, see our firmware supply chain security guide.
Figure 12. How Safeguard covers a smart device maker's software supply chain end to end. Raw firmware images may need extracting first.
How Safeguard helps smart device makers
Safeguard covers the software supply chain of a connected device in one platform: SBOMs for firmware, apps and cloud services, binary and image scanning, reachability-ranked vulnerabilities, SCA, SAST and secrets scanning for code you own, fix pull requests gated by your CI, patched open source packages from Gold, supplier risk and compliance evidence, deployed as SaaS, in your VPC or on premises.
A connected device company has three codebases: the firmware, the companion apps and the cloud backend. Teams often end up with a different tool for each, plus a firmware specialist, an SBOM manager and a GRC (governance, risk and compliance) tool for audits, each with its own inventory. 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 evidence.
See inside binaries and images you ship and receive
In the Safeguard console, Integrations > Upload App Binary accepts Windows EXE, MSI and DLL files, Android APKs, iOS IPAs, VM and disk images (QCOW2, VMDK, VHD and OVA) and container tarballs. Safeguard unpacks the file, identifies embedded libraries and their versions, maps them to known vulnerabilities with EPSS and known exploited enrichment, records licences, and creates a project with a full SBOM.
For a device maker that covers the companion app builds you ship to the app stores, supplier SDK packages, cloud service containers and recovery or update images. Raw firmware images may need to be extracted into their component files first, for example by unpacking the root filesystem; Safeguard does not claim to parse every raw firmware format directly. Heavy obfuscation, encryption or unusual compression can give incomplete results, and statically linked code may show few components. More in our guide to scanning app binaries with Safeguard.
Scan the code you own
For firmware applications, companion apps and cloud services in your repositories, Safeguard runs software composition analysis (SCA) across the full transitive dependency tree. SAST (static application security testing) checks your own code for classes such as command injection and buffer overflows, the bugs behind several cases in this guide. Secrets scanning catches API keys, signing keys and cloud credentials before they ship in an app or firmware image. DAST (dynamic application security testing) tests the cloud APIs your apps call, against verified targets you own.
Rank by reachability, not by count
A firmware image can carry hundreds of known vulnerabilities in its components. Reachability analysis shows which ones your code can actually reach, so the command injection in a network-facing service rises above a flaw in an unused function. That is how a small product security team decides what goes into the next OTA update.
SBOMs your customers and regulators can use
Every project produces an SBOM you can export in CycloneDX or SPDX, among many other formats, for every firmware build and app release. Attestation and signing with SLSA, Sigstore and in-toto record build provenance. That is the component record the EU CRA expects in technical documentation, and the answer a retailer or enterprise buyer wants when an SDK advisory lands.
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 the fix is in a version you cannot move to on an old firmware branch, Safeguard Gold provides patched, hardened builds of those packages, and Safeguard can fork and patch the vulnerable package so your next firmware build pulls a fixed version. For a device team stuck between a supplier's release schedule and a 24-hour CRA reporting clock, that is the difference between shipping the fix in the next OTA update and waiting a quarter.
When the vulnerable code sits inside a binary-only supplier SDK, Safeguard tells you exactly which products and firmware builds contain it, so you can press the supplier with evidence and isolate the affected feature while you wait.
Supplier risk and compliance evidence
Safeguard's Third-Party Risk Manager lets you request SBOMs from SDK providers, ODMs and chip vendors, score them and monitor them continuously (more on Safeguard TPRM). The compliance suite maps evidence from the same scans to frameworks and regulations, so SBOMs, vulnerability handling records and fix history support your PSTI statement of compliance and CRA technical documentation. Guard adds runtime guardrails on the Linux servers behind your cloud backend and update infrastructure, inspecting system calls, library calls and network egress.
Deploy where your code lives
Firmware source and signing keys are sensitive. Safeguard runs as SaaS in US and EU regions, in your own VPC or fully on premises, and can run in Zero AI mode with no language model in the loop, or with your own model. The local runner CLI scans on your own machines. Safeguard is SOC 2 Type I and Type II audited.
How this compares
Device makers often pair a source SCA tool such as Snyk or Black Duck with a firmware analysis specialist such as Finite State, Binarly or Eclypsium, plus a GRC platform. Those are capable products, and firmware specialists go deeper on raw firmware image analysis. Security research firms such as Bitdefender, Mandiant and Nozomi Networks produced much of the public research cited here. Safeguard's case is consolidation and remediation: one inventory across source, apps, containers and binaries; reachability-ranked vulnerabilities; fix pull requests and patched packages instead of tickets; and supplier risk and compliance evidence from the same data. Safeguard can also sit alongside an existing firmware analysis tool and take over the fixing, SBOM delivery and compliance work.
The takeaway
Most smart home compromises are not sophisticated. They use a default password, a reused password, a relay service that trusted the wrong party, or a device nobody updated. For a household, four habits stop nearly all of it: unique passwords with two-factor authentication, automatic updates, a separate network for smart devices, and replacing what the maker no longer supports.
For makers, the cases in this guide point to one root cause: software inside devices that the brand does not fully know and cannot quickly fix. The UK and EU have turned that into law. The makers who know every component in every build, and can ship a fix without waiting on a supplier, will meet those rules and keep their customers' homes private.
Try it yourself
Click through these interactive Safeguard demos to see the steps described in this guide.
Test yourself
Why are default passwords such a big smart home risk?
They are identical across thousands of devices and published in manuals, so one list unlocks a whole product line. The Mirai botnet used a fixed set of 64 common default logins and grew to about 600,000 infected devices in 2016.
How do attackers take over smart home cloud accounts?
Mainly through credential stuffing, which replays usernames and passwords leaked from other websites. Anyone who reused a password is exposed, and Roku reported about 576,000 accounts affected in 2024 and then turned on two-factor authentication for every account.
What three things does a P2P camera relay depend on, and how can they fail?
It depends on the secrecy of the device ID, the authenticity of whoever registers it, and the encryption of the stream. Predictable IDs, unauthenticated registration or weak encryption let attackers find, impersonate or eavesdrop on cameras at scale.
Why did Bitdefender's ThroughTek Kalay research affect so many brands at once?
Device makers build on the same SDKs and libraries, so a flaw in a shared component appears in every product that ships it. It hit cameras from Wyze, Roku and Owlet, and the Kalay platform runs on more than 100 million devices.
What can a compromised smart device do once it is inside your home network?
It sits there with the same access as your laptop, so it can scan for other devices, attack weak services, intercept traffic or relay commands from outside. Home networks are usually flat, and owners rarely look at these devices, so an attacker can stay unnoticed.
How should you segment a home network for smart devices?
Put them on their own Wi-Fi network or VLAN, with rules so they can reach the internet but cannot start connections to your computers and phones. A guest network with client isolation is the simplest version that works on most routers.
What happens to devices that stop getting updates, and what does the UK PSTI regime require?
Every vulnerability found afterwards stays open, as with the Belkin Wemo plug that would not be patched. Since 29 April 2024 the UK PSTI regime bans universal default passwords, requires a published way to report vulnerabilities, and requires a published security update period.
Frequently asked questions
Can smart home devices be hacked?
Yes. Cameras, doorbells, smart TVs, speakers, plugs and baby monitors have all been compromised in documented cases. The most common routes are default passwords, account takeover with passwords reused from other breaches, flaws in shared P2P video services, firmware vulnerabilities and devices that no longer get updates. Most of these attacks are preventable with basic settings.
How do I know if my security camera has been hacked?
Check the camera app for unknown signed-in devices, shared users or changed settings. Watch for the status light turning on when nobody is viewing, the camera moving by itself or voices from its speaker. Check your router for unusual upload traffic. If anything looks wrong, change the password, enable two-factor authentication, update the firmware and reset the camera.
Should smart home devices be on a separate network?
Yes, where your router allows it. Put cameras, TVs, speakers and plugs on a guest network or IoT VLAN that can reach the internet but cannot start connections to your laptops and phones. This limits what an attacker can do from a compromised device. Use an mDNS reflector if casting or local control stops working across segments.
Is two-factor authentication enough to protect a camera account?
Two-factor authentication stops credential stuffing, the most common account attack, and companies such as Ring and Roku now require or enable it for every account. It does not protect against firmware vulnerabilities, P2P flaws or a malicious app on your phone, so combine it with a unique password, updates and network segmentation.
Is my smart TV watching what I watch?
Many smart TVs use automatic content recognition, which identifies what is on screen and sends that data to the maker. A 2024 study of Samsung and LG TVs found it works even when the TV is only a display for an HDMI device, and that opting out in settings stops the traffic. Look for viewing information or content recognition settings.
Does Matter make a smart home secure?
Matter adds real protections: a certified cryptographic identity for each device, encrypted setup with a passcode, and encrypted sessions with AES-CCM. Thread encrypts every radio hop. Matter does not fix bugs in a maker's firmware, weak cloud accounts or devices that stop getting updates. It helps most by letting devices keep working locally.
What happens when a smart device stops getting updates?
It keeps working, but any vulnerability found later is never fixed. Belkin declined to patch a 2023 Wemo smart plug flaw because the model was end of life, and ended cloud support for most Wemo devices in January 2026. Replace unsupported cameras and microphones first, and isolate anything you keep on its own network segment.
What do the UK PSTI Act and EU Cyber Resilience Act mean for buyers?
In the UK, since April 2024, consumer smart devices cannot ship with universal default passwords, and makers must publish a support period and a way to report vulnerabilities. In the EU, the Cyber Resilience Act requires vulnerability reporting from September 2026 and full security requirements, with CE marking, from December 2027.
Sources
- Mandiant (Google Cloud), CVE-2021-28372 ThroughTek Kalay device impersonation (Aug 2021): https://cloud.google.com/blog/topics/threat-intelligence/mandiant-discloses-critical-vulnerability-affecting-iot-devices
- Nozomi Networks, ThroughTek P2P supply chain vulnerability (Jun 2021): https://www.nozominetworks.com/blog/new-iot-security-risk-throughtek-p2p-supply-chain-vulnerability
- Bitdefender Labs, Notes on ThroughTek Kalay vulnerabilities and their impact (May 2024): https://www.bitdefender.com/en-us/blog/labs/notes-on-throughtek-kalay-vulnerabilities-and-their-impact
- KrebsOnSecurity, P2P weakness exposes millions of IoT devices (Apr 2019): https://krebsonsecurity.com/2019/04/p2p-weakness-exposes-millions-of-iot-devices/
- Antonakakis et al., Understanding the Mirai Botnet, USENIX Security 2017: https://www.usenix.org/conference/usenixsecurity17/technical-sessions/presentation/antonakakis
- Cloudflare, Inside the infamous Mirai IoT botnet, a retrospective analysis: https://blog.cloudflare.com/inside-mirai-the-infamous-iot-botnet-a-retrospective-analysis/
- Rapid7, FAQ on baby monitor disclosure (2015): https://www.rapid7.com/globalassets/external/docs/rapid7-faq-on-baby-monitor-disclosure.pdf
- Security Affairs, Roku discloses second breach affecting 576,000 accounts (Apr 2024): https://securityaffairs.com/161765/data-breach/roku-second-data-breach.html
- 9to5Google, Ring makes two-factor authentication mandatory (Feb 2020): https://9to5google.com/2020/02/18/ring-two-factor-authentication-mandatory/
- Engadget, Wyze camera security issue showed 13,000 users other owners' homes (Feb 2024): https://www.engadget.com/wyze-camera-security-issue-showed-13000-users-other-owners-homes-140059551.html
- Bitdefender, Ring Video Doorbell Pro vulnerability exposed users' Wi-Fi password (Nov 2019): https://www.bitdefender.com/en-us/blog/hotforsecurity/bitdefender-finds-ring-doorbell-vulnerability-exposes-users-wi-fi-password
- The Record, Amazon fixes Ring Android app bug that exposed data and camera recordings (2022): https://therecord.media/amazon-quietly-fixes-ring-android-app-bug-that-exposed-data-camera-recordings/
- MacRumors, Anker admits Eufy cameras did not offer end-to-end encryption as promised (Jan 2023): https://www.macrumors.com/2023/01/31/anker-eufy-camera-security
- The Hacker News, researcher uncovers potential wiretapping bugs in Google Home speakers (Dec 2022): https://thehackernews.com/2022/12/researcher-uncovers-potential.html
- The Hacker News, researchers discover LG smart TV vulnerabilities (Apr 2024): https://thehackernews.com/2024/04/researchers-discover-lg-smart-tv.html
- Anselmi et al., Watching TV with the Second-Party: A First Look at Automatic Content Recognition Tracking in Smart TVs, IMC 2024: https://arxiv.org/abs/2409.06203
- 9to5Mac, Wemo smart plug security flaw, no patch coming (May 2023): https://9to5mac.com/2023/05/16/wemo-smart-plug-security-flaw-no-patch-coming/
- MacRumors, Belkin ending support for most Wemo devices (Jul 2025): https://www.macrumors.com/2025/07/10/belkin-ending-wemo-support/
- Silicon Labs, Matter security fundamentals: https://docs.silabs.com/matter/2.2.0/matter-fundamentals-security
- Thread Group, Thread security architecture and features white paper (2026): https://www.threadgroup.org/Portals/0/Documents/thread_security_white_paper.pdf
- European Commission, Cyber Resilience Act: https://digital-strategy.ec.europa.eu/en/policies/cyber-resilience-act
- Wikipedia, Cyber Resilience Act: https://en.wikipedia.org/wiki/Cyber_Resilience_Act
- noze, Cyber Resilience Act practical guide after the Commission's clarifications: https://www.noze.it/en/insights/cra-commission-guidance/
- European Commission, Radio Equipment Directive: https://single-market-economy.ec.europa.eu/sectors/electrical-and-electronic-engineering-industries-eei/radio-equipment-directive-red_en
- techUK, The PSTI Act for consumer IoT explained: https://www.techuk.org/resource/the-psti-act-for-consumer-iot-explained.html
- ETSI EN 303 645 V3.1.3 (2024-09), Cyber Security for Consumer Internet of Things: Baseline Requirements: https://www.etsi.org/deliver/etsi_en/303600_303699/303645/03.01.03_60/en_303645v030103p.pdf
- CVE records, fetched through the CVE Services API: CVE-2021-28372, CVE-2021-32934, CVE-2023-6321, CVE-2023-6322, CVE-2023-6323, CVE-2023-6324, CVE-2019-11219, CVE-2019-11220, CVE-2021-36260, CVE-2023-27217, CVE-2015-2887