Most phones get hacked through the person holding them, not through a technical miracle. A fake delivery text, a banking app installed from a link, a stolen phone with a guessable passcode or a carrier employee talked into moving a number account for far more compromised phones than spyware does. The rare, expensive attacks are real too. Zero-click exploit chains have taken over fully updated iPhones without a single tap, and researchers have shown how a bug in a phone's cellular modem could be reached using only its phone number.
This guide covers both ends. It explains how a smartphone is built and what software it runs, every major way into it, real documented cases with CVE numbers, how to tell if your phone is compromised, and exactly what to change on Android and iPhone. It ends with what businesses, phone makers and app makers can do, because a lot of phone risk starts long before the phone reaches you, in the software supply chain of the apps, SDKs and firmware inside it.
It is a defender's guide written from the attacker's point of view. There is no exploit code here, no tool command lines and nothing that helps anyone get into a phone they do not own.
A note on the law. Testing or accessing a phone you do not own, or have no written permission to test, is illegal in most countries. Security tools mentioned below are named only as context for authorised testing and personal forensics on your own device.
This is part of our device security series, alongside How Wi-Fi Gets Hacked and How to Secure It, How Bluetooth Gets Hacked and How to Secure It and How Smart Home Devices Get Hacked and How to Secure Them.
Contents
- How do phones get hacked?
- What software does a smartphone run?
- What is a phone's attack surface?
- Which phone attacks matter most?
- How do malicious and sideloaded apps get onto phones?
- How do phishing and smishing attacks work on phones?
- What is SIM swapping and how do you stop it?
- What is a zero-click exploit, and how did Pegasus and FORCEDENTRY work?
- Can a phone be hacked through its cellular modem?
- Why do outdated Android phones stay vulnerable?
- How do malicious SDKs end up inside legitimate apps?
- What is stalkerware and how do you find it?
- Is juice jacking real?
- How exposed are phones over Wi-Fi and Bluetooth?
- What happens when a phone is lost or stolen?
- How can you tell if your phone has been hacked?
- What do Lockdown Mode and Advanced Protection actually do?
- How do you secure your phone? A checklist for home users
- How should businesses secure employee phones?
- What should phone makers and app makers do?
- What is inside a phone's software supply chain?
- How Safeguard helps app makers and device makers
- Fix it without waiting for upstream
- Frequently asked questions
How do phones get hacked?
Phones get hacked in five broad ways: the owner is tricked into installing something or typing a password; a supplier ships bad code inside a trusted app; a remote bug is exploited with no interaction; someone nearby abuses a radio, cable or the unlocked device; or the attacker goes around the phone entirely, through the carrier account or cloud account.
Ranked by how many people they hit, the everyday attacks are:
- Phishing and smishing. A text or message leads to a fake login page or a malicious app download.
- Malicious apps. Apps that steal credentials, sign you up for paid services or install more malware, usually from outside official stores.
- Account takeover through the phone number. SIM swapping and SMS code interception.
- Physical theft. A stolen phone, especially one whose passcode the thief watched you type.
- Unpatched software. Known vulnerabilities on phones that no longer get updates, or that get them late.
Ranked by sophistication, the rare attacks are:
- Zero-click exploit chains, such as FORCEDENTRY and BLASTPASS, used to deliver commercial spyware.
- Baseband (modem) exploits that run code on the phone's cellular processor.
- Malicious SDKs and toolchains, where the app is legitimate but a component inside it is not.
The rest of this guide takes each one apart.
What software does a smartphone run?
A smartphone is several computers in one case. The main application processor runs Android or iOS and your apps. A separate secure processor guards keys and biometrics. A separate cellular modem, the baseband, runs its own firmware and talks to the mobile network. Wi-Fi, Bluetooth and NFC chips run firmware of their own.
That separation is the first thing to understand about phone security, because each of those computers is a different target with a different supplier.
Figure 1. The software stack inside a phone. Each box is code from a different team, and often a different company.
The application processor and the operating system
The application processor is the main chip, a system on a chip (SoC) that combines CPU cores, a GPU and other blocks. On Android it runs the Linux kernel with Google's Android framework on top, plus whatever the phone maker adds. On iPhone it runs Apple's XNU kernel and iOS.
Every app runs in a sandbox, a restricted environment that limits what it can see and touch. Android assigns each app its own user ID and uses that to set up a kernel-level application sandbox, reinforced since Android 5.0 by SELinux, a mandatory access control system (Android Open Source Project). On iOS, all third-party apps are sandboxed and each gets a unique, randomly assigned home directory for its files (Apple Platform Security).
Sandboxing is why a single malicious app usually cannot read another app's data directly. It is also why serious attackers chain several bugs: one to get code running, another to break out of the sandbox, a third to take over the kernel.
The secure processor
iPhones have the Secure Enclave, which Apple describes as a dedicated secure subsystem integrated into the SoC and isolated from the main processor, designed to keep sensitive data safe even if the main kernel is compromised (Apple Platform Security). Android phones use a trusted execution environment (TEE) or a separate security chip for the same job. These hold the keys that protect your data at rest and your biometric templates.
The baseband modem
The baseband is the cellular modem: its own processor, running its own real-time operating system and firmware, usually supplied by the chip vendor rather than the phone maker. It parses radio messages from the mobile network before the main OS ever sees them. Google's Android security team describes the baseband as a priority for hardening because of its "unique combination of running in an elevated privilege and parsing untrusted inputs that are remotely delivered into the device" (Google Security Blog).
Wireless chips, the SIM and the boot chain
Wi-Fi, Bluetooth, NFC and ultra-wideband (UWB) radios each run firmware, often from yet another vendor. The SIM or eSIM ties the phone to a carrier account, which matters for SIM swapping. At power-on, a verified boot chain checks that each stage of software is signed before it runs, so a tampered OS should refuse to boot.
Who writes the code?
On a typical Android phone the code comes from Google, the chip vendor, the phone maker, sometimes the carrier, and every app developer and SDK vendor whose code you install. On iPhone, Apple writes far more of the stack itself, but apps and their SDKs still come from thousands of third parties. That mix of suppliers is the phone's software supply chain, and we come back to it near the end.
What is a phone's attack surface?
A phone's attack surface is every place untrusted data or people can reach it: apps and app stores, messages, the browser, the cellular network, Wi-Fi, Bluetooth, USB, the carrier account behind the SIM, the cloud account, and physical access. Each needs something different from the attacker, from a tap to radio range to nothing at all.
Figure 2. Eleven ways into a phone, grouped by what the attacker needs.
It helps to sort the entry points by what the attacker needs, because that tells you which defences matter:
| Entry point | What the attacker needs | Main defence |
|---|---|---|
| Malicious or sideloaded apps | You to install it | Official stores only, Play Protect, permissions review |
| SDKs inside legitimate apps | A foothold in an app's supply chain | App makers' SBOMs and binary scanning |
| Phishing and smishing | You to tap and type | Passkeys, link caution, spam filtering |
| Zero-click messages | Your number or account ID, plus an exploit chain | Updates, Lockdown Mode, Advanced Protection |
| Browser exploits | You to open a page | Updates, HTTPS-only, hardened browsing modes |
| Cellular network and modem | Radio range, or only your number | Updates, 2G off, carrier settings |
| Wi-Fi and Bluetooth | Radio range | Updates, radios off when idle |
| USB and public charging | Your cable in their port | Own adapter, USB data restrictions |
| Carrier account and SIM | A convincing story for the carrier | Carrier PIN, no SMS recovery |
| Physical access | Your phone in their hand | Strong passcode, theft protection |
| Cloud account | Your password or recovery channel | Passkeys, two-factor authentication |
Notice the last two rows. Some of the most damaging "phone hacks" never touch the phone's software at all.
Which phone attacks matter most?
For most people the biggest risks are cheap, mass attacks: smishing, malicious apps and malicious SDKs, followed by account takeover through SIM swapping and theft. Zero-click spyware and baseband exploits are far more capable but cost so much that they are aimed at a small number of chosen targets, such as journalists, activists and executives.
Figure 3. A qualitative way to prioritise. Mass attacks sit top left; targeted, expensive attacks bottom right.
The chart is an editorial judgement, not a measurement, and it can mislead if read too literally. A SIM swap affects few people in aggregate, but for the person whose crypto wallet or email is taken it is a disaster. And the high-end attacks matter to everyone indirectly, because the bugs they use are often in code that every phone runs.
The cases below show how the risk has shifted over the past decade.
Figure 4. A decade of public phone attacks. Red: zero-click or remote. Amber: supply chain. Blue: radio or physical proximity.
How do malicious and sideloaded apps get onto phones?
Malicious apps reach phones in two ways: they slip past app store review disguised as something useful, or the user installs them directly from a website or message, which is called sideloading. Sideloading is the bigger problem on Android; Google says over 95 percent of installs from major fraud-linked malware families came from internet sideloading sources.
How it works
The attacker builds an app that looks legitimate: a flashlight, a PDF reader, a parcel tracker, a fake banking update. Once installed, it asks for permissions it can abuse. Accessibility access is a favourite on Android, because accessibility services can read the screen and act on the user's behalf. Some malicious apps start clean and download harmful code later, after passing review.
What the attacker needs: a convincing app and a way to get it installed. That means either beating store review or persuading the user to allow installs from "unknown sources".
Real cases
The Bread family, also known as Joker, is a good example of persistence. Google tracked it from early 2017. It started with SMS fraud (silently texting premium numbers) and moved to WAP billing fraud, charging the user's carrier bill. Google said Play Protect detected and removed 1,700 unique Bread apps from the Play Store before users downloaded them, and that the family had "at some point used just about every cloaking and obfuscation technique under the sun" (Google Security Blog).
The scale of store review is large. Google says it prevented 2.36 million policy-violating apps from being published on Google Play in 2024 and banned more than 158,000 developer accounts. Outside the store, Play Protect's real-time scanning identified more than 13 million new malicious apps in 2024. Google also reported that more than 95 percent of installations from major malware families that exploit sensitive permissions linked to financial fraud came from internet sideloading sources such as web browsers, messaging apps and file managers (Google Security Blog).
That last figure explains Google's developer verification requirement. Under Google's published plan, apps installed on certified Android devices must be registered by verified developers, starting in September 2026 in Brazil, Indonesia, Singapore and Thailand and rolling out globally from 2027 (Android Developers Blog).
On iPhone, sideloading is far more restricted. In the European Union, iOS and iPadOS let authorised developers run alternative app marketplaces that distribute notarized apps, with Apple's authorisation required (Apple Developer). Malicious iOS apps have still reached the official App Store, usually through compromised components, which we cover in the SDK section.
Impact
Stolen banking and email credentials, intercepted one-time codes, fraudulent charges, and in the worst cases full remote control through accessibility abuse.
How to detect it
On Android, run a Play Protect scan, then look for apps you do not remember installing, apps with no icon, and anything holding accessibility or device admin rights that should not. Unexplained carrier charges are a sign of billing fraud families like Joker. On iPhone, look for configuration profiles you did not install (Settings shows them under VPN and Device Management) and apps you do not recognise.
How to defend against it
- Install apps only from Google Play or the App Store, and leave "install unknown apps" off for every browser and messaging app.
- Keep Play Protect on. Read the permissions an app asks for, and refuse accessibility access unless the app genuinely needs it, such as a screen reader.
- Delete apps you no longer use. Fewer apps means fewer updates you depend on.
- If you are high risk, Android's Advanced Protection blocks sideloading entirely (more below).
How do phishing and smishing attacks work on phones?
Smishing is phishing by text message. The attacker sends an SMS or chat message, often about a delivery, a bank alert or a missed payment, that links to a fake login page or a malicious app. Phones make it worse because small screens hide full web addresses and people act on texts quickly.
How it works
The message creates urgency and gives you a link. The link leads to a page that looks like your bank, email provider or company login, or to a download. Modern phishing kits relay what you type to the real site in real time, so they can capture a one-time code too. That defeats most SMS and authenticator app codes.
What the attacker needs: a list of phone numbers, a convincing pretext and a lookalike domain. Nothing about your phone's software.
Real cases
FluBot was an Android banking trojan spread by text messages that masqueraded as missed calls and parcel deliveries and asked recipients to click links that downloaded the malware (Wikipedia).
The 2022 attack on Cloudflare shows the business version. On 20 July 2022, at least 76 Cloudflare employees received text messages within less than a minute, on personal and work phones, pointing to a convincing fake Okta login page. Three employees entered their credentials. The attack still failed, because Cloudflare issues every employee a FIDO2 security key, and the key's origin binding meant the stolen password and a relayed code were useless (Cloudflare). Cloudflare also noted that the three employees were not reprimanded, which is the right call: people who report mistakes quickly are a defence in themselves.
Google's Android security team has also described fake base stations being used to push smishing payloads directly, likely by forcing phones down to 2G (Google Security Blog).
Impact
Account takeover, malware installation, and for businesses, a foothold into internal systems.
How to detect it
Check the real domain before typing anything. Be suspicious of any message that pushes urgency and a link at the same time. On a business fleet, sign-in logs that show a successful password followed by failed second-factor attempts from an unfamiliar location are a classic sign.
How to defend against it
- Use passkeys wherever they are offered. Passkeys are FIDO credentials based on public key cryptography, and the FIDO Alliance describes them as phishing-resistant because they only work on the real site (FIDO Alliance).
- For work accounts, issue hardware security keys. The Cloudflare case is the clearest public evidence that they hold up under a well-run attack.
- Type addresses yourself or use a saved bookmark for banking and email, rather than following links in texts.
- Turn on spam and scam filtering in your messaging app.
We go deeper in what smishing is and how to spot a scam text.
What is SIM swapping and how do you stop it?
SIM swapping is when an attacker convinces your mobile carrier to move your phone number to a SIM or eSIM they control. Your phone loses service, and the attacker starts receiving your calls and texts, including the one-time codes that reset email, bank and crypto accounts. The attack goes through the carrier, not your phone.
Figure 5. A SIM swap never touches the phone. Every stage has a matching protection.
How it works
The attacker collects personal details from breaches and social media, then contacts the carrier pretending to be you, or bribes or tricks an insider, and asks for a new SIM or a number port to another carrier. Once the number moves, they request password resets that send codes by SMS.
What the attacker needs: enough personal information to pass the carrier's checks, and a list of your accounts that allow SMS recovery.
Real cases
Researchers at Princeton tested how five prepaid wireless carriers authenticated customers asking for a SIM change. All five used authentication challenges that attackers could easily subvert. The same team reverse-engineered the authentication policies of more than 140 websites that offer SMS-based authentication and found 17 where an account could be compromised with a SIM swap alone (Lee et al., SOUPS 2020).
High-profile victims include Twitter's then chief executive Jack Dorsey, whose account was hijacked through a SIM swap in 2019 (Wikipedia).
Impact
Loss of email, which is usually the master key to everything else; drained bank and crypto accounts; hijacked social media used for fraud.
How to detect it
The giveaway is sudden loss of mobile service, often with a "SIM not provisioned" or "no SIM" message, while Wi-Fi still works. Emails about password changes you did not request are the second sign. Treat either as an incident: call your carrier from another phone straight away.
How to defend against it
- Ask your carrier for an account PIN, a SIM lock, or port-out protection, whichever it offers.
- Remove your phone number as a recovery method on important accounts, and switch to passkeys, hardware keys or an authenticator app.
- Keep your date of birth, address and carrier off public profiles; they are the answers to carrier security questions.
- For businesses, do not use SMS for staff sign-in or password resets on sensitive systems.
Our explainer on phishing, vishing, smishing and SIM swaps covers the social engineering side.
What is a zero-click exploit, and how did Pegasus and FORCEDENTRY work?
A zero-click exploit compromises a phone without the owner tapping anything. The attacker sends data the phone processes automatically, such as a message attachment or a call, and a bug in that processing gives them code execution. Pegasus, commercial spyware sold by NSO Group, has used several such chains against iPhones and WhatsApp.
This is the attack that makes headlines. It is also the one most people will never face, because a working chain against current iOS or Android is extremely expensive to build and burns out once it is discovered and patched.
Figure 6. The conceptual stages of a zero-click chain and the protection aimed at each one.
How it works, conceptually
Phones parse a lot of complex data automatically. When a message arrives, the phone may decode an image to show a preview, render a font, or parse a file attachment, all before you look at it. Each parser is code written in memory-unsafe languages such as C and C++, and a single mistake in bounds checking can let crafted data overwrite memory.
A full chain typically needs several bugs:
- Delivery. Get crafted data to the phone through a channel that processes it automatically.
- Initial code execution inside the limited process that handles that data.
- Sandbox escape to break out of that limited process.
- Kernel privilege escalation to take full control.
- Implant. Install spyware that reads messages, location, microphone and camera.
What the attacker needs: the target's phone number or account ID, and a chain of unpatched vulnerabilities, often several zero-days. Building one takes rare skill; commercial spyware vendors sell access to such chains.
WhatsApp, 2019
In 2019, WhatsApp disclosed that Pegasus had used a vulnerability in its calling feature to install spyware by calling the target, and the spyware installed even if the call was not answered. A court later found NSO Group liable for attacks on about 1,400 WhatsApp users (Wikipedia, Pegasus).
FORCEDENTRY, CVE-2021-30860
In 2021 Citizen Lab analysed the phone of a Saudi activist and found a zero-click iMessage exploit it named FORCEDENTRY. Apple assigned it CVE-2021-30860, described it as "processing a maliciously crafted PDF may lead to arbitrary code execution", and patched it on 13 September 2021 in iOS 14.8. Citizen Lab believed it had been in use since at least February 2021 (Citizen Lab).
Google Project Zero's analysis explains the trick at a conceptual level (Project Zero):
- The attack needed only a phone number or Apple ID.
- Files named with a
.gifextension were processed by an iMessage component outside BlastDoor, the sandbox Apple added in iOS 14 to isolate message parsing. Because the system detected the real format from the content, a "GIF" that was really a PDF reached a much larger set of image decoders. - The PDF used JBIG2, an old compression format for black and white scans, and an integer overflow in the JBIG2 decoder gave the attacker memory corruption.
- JBIG2 combines image regions with logical operations such as AND, OR and XOR. The attackers used tens of thousands of those operations to build a small virtual computer inside the decoder, which then ran the next stage.
Project Zero called it "one of the most technically sophisticated exploits we've ever seen." Apple's follow-up hardening is the defensive lesson: iOS 14.8.1 restricted which image formats that component could reach, and iOS 15 moved GIF decoding inside BlastDoor.
BLASTPASS, CVE-2023-41064 and CVE-2023-41061
In September 2023, Citizen Lab found another zero-click chain, BLASTPASS, on the phone of an employee of a civil society organisation. It used PassKit attachments (the format behind Wallet passes) containing malicious images, sent from an attacker's iMessage account, and could compromise iPhones on iOS 16.6, then the latest version. Apple issued CVE-2023-41064 and CVE-2023-41061. Importantly, Citizen Lab said Apple's security team confirmed that Lockdown Mode blocked this particular attack (Citizen Lab).
Operation Triangulation
Kaspersky found Operation Triangulation in 2023 while monitoring its own corporate Wi-Fi network for mobile devices and noticing suspicious traffic from several iPhones (Securelist). The chain began with an iMessage attachment processed with no sign to the user and used four zero-days (Securelist):
| CVE | Role in the chain |
|---|---|
| CVE-2023-41990 | Code execution through an undocumented TrueType font instruction |
| CVE-2023-32434 | Integer overflow in the XNU kernel's memory mapping |
| CVE-2023-38606 | Undocumented hardware registers used to bypass a kernel memory protection |
| CVE-2023-32435 | Safari exploit used in a later stage |
The chain worked on iOS versions up to 16.2. The hardware piece is the striking part: the attackers wrote to undocumented hardware registers to get around Apple's Page Protection Layer, and Kaspersky could not determine how they learned of them.
Impact
Complete compromise: messages, including those in end-to-end encrypted apps (read on the device, after decryption), call data, photos, location, microphone and camera.
How to detect it
It is hard. Amnesty's forensic research found that Pegasus left traces in system databases such as the logs that record which processes used the network, in Safari history, and in suspicious process names designed to look like legitimate iOS services. Amnesty also believed later versions of Pegasus did not survive a reboot. It noted that investigators find significantly more forensic traces on iOS than on stock Android, which reflects investigator access, not which platform is more secure (Amnesty International). Amnesty released the open source Mobile Verification Toolkit (MVT) for this work (MVT on GitHub); it is meant for checking devices you own or have consent to examine.
Apple also sends threat notifications to users it believes were individually targeted by mercenary spyware. Apple says it has notified users in over 150 countries since 2021 and recommends Lockdown Mode and expert help, such as Access Now's Digital Security Helpline, to recipients (Apple Support).
How to defend against it
- Update immediately when Apple or Google ships an emergency fix. Every chain above was closed by an update.
- Restart your phone regularly. If an implant does not persist, a reboot removes it, and the attacker has to exploit you again.
- If you might be a target, turn on Lockdown Mode (iPhone) or Advanced Protection (Android). Both shrink the parsing surface that zero-click chains need.
- Buy hardware with memory tagging. Apple's Memory Integrity Enforcement, built on Enhanced Memory Tagging Extension, is always on in all iPhone 17 and iPhone Air models and covers key attack surfaces including the kernel (Apple Security Research). On Android, Memory Tagging Extension (MTE) debuted on the Pixel 8 in 2023 and is reaching other phones (EFF). Memory tagging catches many of the memory corruption bugs these chains rely on.
For the supply chain side of commercial spyware, see our piece on Pegasus and NSO Group.
Can a phone be hacked through its cellular modem?
Yes. The baseband modem parses data from the mobile network, and bugs in it can give an attacker code execution on that processor. In 2023 Google Project Zero reported 18 vulnerabilities in Samsung Exynos modems; four allowed internet-to-baseband remote code execution that needed only the victim's phone number.
How it works
The modem runs complex protocol code for 2G through 5G, voice over LTE (VoLTE) and Wi-Fi calling, often without the exploit mitigations that modern operating systems use. Google notes that baseband remote code execution exploits have their own category in exploit marketplaces at a relatively low price, which "suggests baseband bugs may potentially be abundant and/or not too complex to find and exploit" (Google Security Blog).
There are two attacker positions. One is over the network, sending crafted call setup or messaging traffic. The other is over the air, using a fake base station (also called an IMSI catcher) that impersonates a real cell tower. Phones cannot reliably tell a legitimate base station from a fake one, and 2G lacks mutual authentication, which is why attackers try to force phones down to 2G (Google Security Blog).
What the attacker needs: for the worst Exynos bugs, only the phone number. For other baseband attacks, radio proximity with fake base station equipment, or the position of a malicious network operator.
Real case: the Exynos modem vulnerabilities
In March 2023 Project Zero disclosed 18 zero-day vulnerabilities in Exynos modems made by Samsung Semiconductor. Four of them, CVE-2023-24033, CVE-2023-26496, CVE-2023-26497 and CVE-2023-26498, allowed remote compromise at the baseband level with no user interaction, and required "only that the attacker know the victim's phone number." Project Zero listed likely affected devices including Samsung Galaxy S22, M33, M13, M12, A71, A53, A33, A21s, A13, A12 and A04 series phones, several Vivo models, Pixel 6 and Pixel 7 series phones, and vehicles using the Exynos Auto T5123 chipset. Pixel phones received fixes for all four severe vulnerabilities in the March 2023 update (Project Zero).
Impact
Code running on the modem can intercept or alter calls and data at the radio level. To control the rest of the phone, an attacker would still need a second bug to cross from the modem into the main OS.
How to detect it
You usually cannot, as a user. Baseband compromise leaves little the main OS can see. Defence depends on updates and on reducing exposure.
How to defend against it
- Install security updates, which carry modem firmware fixes.
- While a known modem bug is unpatched, Project Zero's advice was to turn off Wi-Fi calling and VoLTE, if your carrier lets you.
- Turn off 2G. Android 12 added a setting to disable 2G at the modem level (first on Pixel 6, now on Android devices that support Radio HAL 1.6 or later), and Android 14 lets IT administrators disable 2G across a managed fleet and adds protection against null-ciphered (unencrypted) cellular connections (Google Security Blog). Lockdown Mode on iPhone turns off 2G and 3G.
For phone and chip makers, the fixes are structural, and we list them in the makers' section below.
Why do outdated Android phones stay vulnerable?
Because an Android fix has to pass through several companies before it reaches your phone. Google publishes a monthly security bulletin, but chip vendors and phone makers must then integrate, build, test and ship it for each model. Each hop adds delay, and once a model's support ends, known vulnerabilities stay open for good.
Figure 7. Why Android patches arrive late. The Mali GPU case shows how long a fixed bug can stay open on real phones.
How it works
The Android Security Bulletin is published on the first Monday of each month and lists fixes by security patch level, a date you can see in your phone's settings (Android Open Source Project). But much of a phone's code is not in Android itself. GPU drivers, modem firmware and other chip vendor components come from those vendors, and the phone maker has to merge everything for each model.
The gap between "fixed upstream" and "fixed on your phone" is called the patch gap, and attackers use it. They do not need a zero-day if a published fix has not reached the target yet.
Real cases
Google Project Zero's "Mind the gap" post is the clearest public example. A Project Zero researcher found five exploitable vulnerabilities in the Arm Mali GPU driver and reported them to Arm in June and July 2022. Arm fixed them in July and August 2022. When Project Zero tested its devices months later, every device it tested that used a Mali GPU was still vulnerable, and the fix (tracked as CVE-2022-36449) did not appear in downstream security bulletins (Project Zero).
Exploited bugs do show up in the bulletins too. The September 2025 Android bulletin flagged CVE-2025-38352 (in the upstream Linux kernel) and CVE-2025-48543 (in the Android Runtime) as possibly under "limited, targeted exploitation" (Android Security Bulletin, September 2025).
Support windows are getting longer
Phone makers now promise much longer support for flagships. Google says Pixel 8 and later get seven years of OS and security updates, while Pixel 6 and Pixel 7 series phones get five years from when they first went on sale in the US Google Store (Google Pixel Help). The Galaxy S24 series was the first Galaxy S line with seven years of OS and security updates (Wikipedia). Cheaper phones often get far less, and Apple publishes its fixes on its security releases page (Apple Support) without a fixed support period.
How to detect it
Open your phone's settings and look at the security patch level. If it is several months old and no update is offered, your phone is running with known vulnerabilities. If the model is out of support, it will not get better.
How to defend against it
- Buy phones with a published, long update commitment, and check it before you buy.
- Turn on automatic updates and install them within days.
- Retire phones that are out of support, or at least stop using them for banking, work email and authentication.
- Businesses should enforce a minimum patch level through MDM and block access from phones that fall behind.
How do malicious SDKs end up inside legitimate apps?
A software development kit (SDK) is a packaged library that app developers add for ads, analytics, payments, login or features like text recognition. If an SDK is malicious or compromised, it runs inside a trusted app with every permission that app holds. The user, the store and often the developer never notice.
Figure 8. One bad SDK inherits every permission the user granted the app.
How it works
An SDK is code the app maker did not write and usually does not review line by line. It ships inside the app package (the APK on Android, the IPA on iOS) and runs in the same process. Some SDKs are malicious from the start and offered free or with a revenue share; others are compromised later; some legitimate SDKs fetch remote configuration or code that changes their behaviour after the app passes review.
What the attacker needs: a developer willing to include their SDK, often through attractive ad revenue terms, or a foothold in an SDK vendor's or developer's build process.
Real cases
SpinOk (2023). Doctor Web found an Android module, which it named Android.Spy.SpinOk, presented as a marketing SDK. It could collect information about files on the device, transfer them to attackers, and replace and upload clipboard contents. Doctor Web found it in 101 Google Play apps with at least 421,290,300 cumulative downloads (Doctor Web).
SparkCat (2025). Kaspersky found a malicious SDK in Android and iOS apps, some on Google Play and the App Store, that used an optical character recognition (OCR) plug-in built with Google's ML Kit to scan photos in the gallery for crypto wallet recovery phrases and send matching images to the attackers. The infected Play apps had more than 242,000 downloads, and Kaspersky said it was the first time a stealer had been found in Apple's App Store. Apple and Google removed the apps after the report (Securelist).
CamScanner (2019). CamScanner, a document scanner with more than 100 million downloads on Google Play, was legitimate for a long time. Then later versions shipped with an advertising library containing a malicious module, a trojan dropper that unpacked and ran further malware. Kaspersky said it had also seen that module in some apps preinstalled on Chinese smartphones (Kaspersky).
XcodeGhost (2015). This one attacked the build tools rather than an SDK. Trojanised copies of Apple's Xcode development tool, versions 6.1 to 6.4, were shared through a Chinese cloud file service and downloaded by developers. Apps built with them carried code that collected device information and sent it to command servers, and infected apps passed App Store review (Unit 42).
Impact
Data theft at app scale: files, photos, clipboard contents, contact lists and location, from millions of phones, under the name of an app people trusted.
How to detect it
Users mostly cannot. Detection belongs to app makers and researchers: inventory every SDK in every build, scan the final binaries, and watch for SDK updates that add permissions, network destinations or dynamic code loading.
How to defend against it
For users: keep apps updated (fixed versions remove the SDK), delete apps you do not use, and give photo and file access only when needed, using "selected photos" options where your OS offers them.
For app makers, see the makers' section below. The short version: keep an SBOM (software bill of materials) for every build, scan the APK and IPA you actually ship, pin and verify SDK versions, and treat SDK updates as code changes that need review. Apple now requires privacy manifests for a list of commonly used SDKs, and Xcode validates that a new version of a signed SDK was signed by the same developer (Apple Developer). Google maintains the Google Play SDK Index for SDK information.
We have written more on iOS SDK and ad library risk and third-party library risk in Android apps.
What is stalkerware and how do you find it?
Stalkerware is commercially available monitoring software that lets someone secretly spy on another person's phone: messages, location, calls, photos and browsing. It is usually installed by someone with physical access, often a partner or family member. In 2023 Kaspersky detected it on the phones of 31,031 unique users.
How it works
The Coalition Against Stalkerware defines it as software, apps and devices that "enable someone to secretly spy on another person's private life via their mobile device," running hidden in the background (Coalition Against Stalkerware). On Android, the abuser typically installs an app while holding the unlocked phone and grants it broad permissions. On iPhone, Kaspersky notes it is harder and usually requires physical access; an abuser may also jailbreak a new iPhone, install stalkerware and then give it to the victim. Abusers also skip software entirely by signing in to the victim's cloud account and reading synced data.
What the attacker needs: a few minutes with the unlocked phone, or the victim's cloud account password.
Scale
Kaspersky counted 31,031 affected unique users in 2023, up from 29,312 in 2022, across 175 countries, with Russia, Brazil and India the most affected (Securelist). Those are only users of one security product, so the real number is higher.
Impact
Loss of privacy and safety. Stalkerware is a tool of domestic abuse, and its location tracking can put people in physical danger.
How to detect it
Signs include battery drain, unfamiliar apps (especially ones with device admin or accessibility rights), an Android setting allowing unknown apps you did not change, and an abuser knowing things they should not. On iPhone, Safety Check (iOS 16 or later) lets you review who has access to your information and location, review devices signed in to your Apple Account, reset app privacy permissions, and use Emergency Reset to stop sharing with everyone at once (Apple Personal Safety User Guide).
How to defend against it, safely
If you suspect stalkerware, think about safety before removing it. Removal can alert the abuser and escalate risk, so it can help to plan with a domestic abuse support organisation before acting. Then:
- Change your phone passcode and your cloud account password from a device the abuser has never touched, and turn on two-factor authentication.
- Review connected devices and sharing settings on your Apple or Google account.
- Run a reputable security app on Android, and review apps holding accessibility and device admin rights.
- If needed, back up photos and contacts selectively, and factory reset the phone.
Is juice jacking real?
Partly. Juice jacking means a malicious public USB charger stealing data or installing malware. It has been demonstrated since 2011, but reviews have found no credible reported cases against phones outside research. In 2025, though, researchers showed ChoiceJacking, which bypassed the existing protections on every tested phone from eight vendors, so the risk is not zero.
What is real and what is myth
The term was coined by Brian Krebs after a demonstration charging kiosk at DEF CON in 2011. As of May 2023, multiple reviews had found no credible reported cases of juice jacking on mobile operating systems outside research, and experts generally rated the risk to typical users as low (Wikipedia).
The reason it stayed rare is that both iOS and Android added a prompt asking you to trust or allow data access when you connect to a computer. ChoiceJacking showed that prompt can be fooled. Researchers from Graz University of Technology presented it at USENIX Security 2025: a malicious charger spoofs user input to approve its own data connection. In their tests the attacks reached sensitive files on all tested devices from eight vendors, including the six largest by market share, and on two vendors' devices files could be extracted even while the phone was locked. All but one of the vendors notified, including Google, Samsung, Xiaomi and Apple, acknowledged the attacks and were working on mitigations (USENIX; Wikipedia).
What the attacker needs: a modified charger or port in a public place, and you plugging in.
How to defend against it
- Charge from your own adapter plugged into a mains socket, or carry a power bank.
- Use a charge-only cable or USB data blocker when you must use a public USB port.
- Keep the phone locked while charging in public, and decline any "trust this computer" prompt you did not expect.
- Lockdown Mode requires the iPhone to be unlocked to connect to accessories or computers; Android's Advanced Protection turns on USB protection so that new USB connections only charge while the phone is locked.
Our deeper look at what is documented about juice jacking covers the history.
How exposed are phones over Wi-Fi and Bluetooth?
Phones are exposed to anyone within radio range through Wi-Fi and Bluetooth: flaws in the protocols, flaws in the chip firmware, and flaws in the operating system's wireless stacks. Major bugs such as KRACK, FragAttacks, BlueBorne and CVE-2023-45866 were all fixed by updates, which is why updating matters more than avoiding networks.
Wi-Fi
KRACK (2017), found by Mathy Vanhoef, showed weaknesses in WPA2 that let an attacker in range break the protection of a protected Wi-Fi connection by forcing nonce reuse (KRACK attacks). FragAttacks (2021) found design flaws affecting every Wi-Fi security protocol back to WEP in 1997, including WPA3, though the design flaws were hard to abuse and the bigger concern was simpler programming mistakes in Wi-Fi products (FragAttacks). Open networks add a separate risk: phones that auto-join a familiar network name can be lured onto an attacker's access point.
On a phone, the practical defences are updates, turning off auto-join for open networks, forgetting old networks, and relying on HTTPS and encrypted apps rather than the network. Lockdown Mode and Advanced Protection both stop automatic joining of insecure networks. Our companion post, How Wi-Fi Gets Hacked and How to Secure It, covers routers, encryption and network attacks in full, and is public Wi-Fi safe separates myth from fact.
Bluetooth
BlueBorne, disclosed by Armis on 12 September 2017, was a set of eight vulnerabilities in Bluetooth implementations in Android, iOS, Linux and Windows. Armis estimated it could potentially affect over 8.2 billion devices (Wikipedia).
CVE-2023-45866, found by Marc Newlin, was an unauthenticated Bluetooth keystroke-injection flaw: an attacker in range could trick the host into pairing with a fake keyboard without user confirmation and then type on the victim's device. Newlin found related issues in Android, Linux, macOS and iOS, and noted the macOS and iOS variants were exploitable even in Lockdown Mode (Newlin). That is a useful reminder that no single mode covers everything; updates closed it.
What the attacker needs: radio range, typically tens of metres, and a vulnerability your phone has not patched.
How to defend against it: update; turn Bluetooth off when you are not using it, particularly in crowded places; remove old pairings; and do not accept pairing requests you did not start. Our companion post, How Bluetooth Gets Hacked and How to Secure It, covers pairing, Bluetooth Low Energy, earbuds and trackers in depth.
What happens when a phone is lost or stolen?
A stolen phone is only as safe as its lock screen. If the thief has your passcode, often by watching you type it, they can change your account password, turn off tracking and empty your accounts. Stolen Device Protection on iPhone and Theft Detection Lock on Android exist to stop exactly that.
How it works
Thieves who watch a victim type a passcode in a bar or on a train, then steal the phone, can use that passcode for far more than unlocking: changing the Apple or Google account password, removing tracking, reading saved passwords and approving payments.
Defences built for this
On iPhone, Stolen Device Protection requires Face ID or Touch ID, with no passcode fallback, for actions such as using stored passwords and cards when the phone is away from familiar locations. For critical actions such as changing your Apple Account password, it adds a security delay: wait an hour, then authenticate with biometrics again. That hour gives you time to mark the phone as lost. You can set the delay to apply everywhere, not only away from familiar places (Apple Support).
On Android, Google added Theft Detection Lock, which uses on-device AI to sense the motion of a phone being snatched and locks the screen; Offline Device Lock, which locks the screen if a thief keeps the phone offline; and Remote Lock, which lets you lock the phone using only your phone number and a security challenge. Google said Theft Detection Lock and Offline Device Lock would reach Android 10 and later devices through a Google Play services update. Android also tightened factory reset so thieves cannot reset a stolen device without your credentials (Google).
How to defend against it
- Use a six-digit or longer passcode, and shield it in public.
- Turn on Stolen Device Protection or Theft Detection Lock and Offline Device Lock.
- Make sure Find My (iPhone) or Find My Device (Android) is on, and know how to mark the phone lost from another device.
- Keep a list of what to do in the first hour: mark lost, change the account password from a trusted device, call your carrier, and alert your bank.
How can you tell if your phone has been hacked?
Look for things that changed without you: apps, profiles or accessibility permissions you did not add, account sign-in alerts, sudden loss of mobile service, carrier charges, and settings that switched themselves. Battery drain and heat alone prove little. For targeted spyware, use Apple's threat notifications, Android's Intrusion Logging or expert forensics with tools such as MVT.
Everyday signs worth checking
| Sign | What it can mean | What to do |
|---|---|---|
| Unknown app, or one with no icon | Malicious app or stalkerware | Check its permissions, then uninstall |
| Accessibility or device admin granted to an unexpected app | Android banking malware or stalkerware | Revoke, uninstall, run Play Protect |
| Configuration profile you did not install (iPhone) | Monitoring or traffic interception | Remove it, change Apple Account password |
| Sudden "no SIM" or no service | SIM swap | Call carrier from another phone at once |
| Password reset emails you did not request | Account takeover attempt | Secure email first, then other accounts |
| Unknown premium charges on your bill | Billing fraud malware | Check apps, contact carrier |
| Old security patch level, no update offered | Phone out of support | Replace, or reduce what it is used for |
Battery drain, heat and slowness have many innocent causes. On their own they are weak evidence, though they are worth a check if they appear suddenly alongside other signs.
Signs of targeted spyware
Sophisticated spyware is designed to leave no visible trace. The realistic options are:
- Apple threat notifications, shown on the iPhone's Lock Screen and in Settings, sent by email and displayed on the Apple Account website when Apple detects mercenary spyware activity (Apple Support).
- Intrusion Logging on Android, part of Advanced Protection, which securely backs up device logs in a privacy-preserving, tamper-resistant way that only the user can access, so they can be analysed if compromise is suspected (Google Security Blog).
- Expert forensics. Amnesty's MVT checks backups and system data against published indicators of compromise (MVT). Use it only on your own phone or with the owner's consent, and work with an organisation that does this regularly.
- Network monitoring for organisations. Operation Triangulation was found because Kaspersky watched traffic on the Wi-Fi network its staff phones used (Securelist). Unusual connections from phones to unfamiliar domains are worth investigating.
If you think you are compromised
- Disconnect from the network if you are in immediate danger of data loss, but keep the phone powered if you want a forensic analysis; then seek expert help.
- From a different, trusted device, change the password for your Apple or Google account and your email, and sign out other sessions.
- Update the phone, then restart it.
- If the evidence is strong, back up only what you need and factory reset, then restore apps from the store rather than from a full backup.
What do Lockdown Mode and Advanced Protection actually do?
Lockdown Mode (iPhone, iOS 16 and later) and Advanced Protection (Android 16 and later) are opt-in modes for people at high risk. They switch off or restrict features that attackers use as entry points, such as complex message attachments, some web technologies, unknown apps, 2G and USB data, in exchange for some convenience.
Figure 9. What each high-risk mode changes. Feature sets evolve with updates and some depend on hardware.
Lockdown Mode on iPhone
Apple's support page lists what Lockdown Mode changes (Apple Support):
- Messages: most attachment types are blocked, other than certain images, video and audio.
- Web browsing: certain complex web technologies are blocked.
- FaceTime: incoming calls are blocked unless you have called that person or contact in the past 30 days.
- Photos: location is removed when you share photos, and shared albums are removed.
- Device connections: the iPhone must be unlocked to connect to an accessory or computer.
- Wireless: the phone will not automatically join non-secure Wi-Fi networks, and 2G and 3G are turned off.
- Configuration profiles and MDM: profiles cannot be installed and the phone cannot be enrolled in mobile device management while Lockdown Mode is on.
Lockdown Mode is available in iOS 16 and later, with additional protections from iOS 17. The best public evidence that it works is BLASTPASS, which Apple confirmed it blocked.
Advanced Protection on Android
Google introduced Advanced Protection as a device-level setting in Android 16. Once on, it enables a set of protections and prevents them from being turned off individually (Google Security Blog). According to Google and the EFF's review (EFF), it:
- turns on Android's theft detection features;
- forces HTTPS for websites in Chrome and turns off Chrome's JavaScript optimizer;
- enables scam and spam protection in Google Messages;
- disables 2G;
- keeps Google Play Protect and Safe Browsing on, and blocks installs from unknown sources;
- restarts the phone if it stays locked for 72 hours, returning data to its fully encrypted state;
- turns on USB protection, so new USB connections only charge while the phone is locked;
- stops automatic reconnection to insecure Wi-Fi networks;
- uses Memory Tagging Extension where the hardware supports it;
- adds Intrusion Logging.
Google said some of these, including Intrusion Logging, USB protection, the Wi-Fi reconnect option and Scam Detection in Phone by Google, would arrive after the initial Android 16 release.
Should you turn them on?
If you are a journalist, activist, lawyer, executive, public figure, or anyone who has received a threat notification, yes. For everyone else, the trade-offs are real (some websites break, attachments are restricted, sideloading is gone) but small for many people. Try it for a week. Our guide on who should turn on Lockdown Mode goes deeper.
Neither mode is a guarantee. CVE-2023-45866 worked against Apple devices in Lockdown Mode, and both modes depend on you also installing updates.
How do you secure your phone? A checklist for home users
Keep the phone updated and supported, use a strong passcode and theft protection, switch important accounts to passkeys and off SMS recovery, add a carrier PIN, install apps only from official stores, review permissions, and turn off radios you do not use. People at high risk should add Lockdown Mode or Advanced Protection.
Figure 10. The 12-point checklist. The first six close the most common routes in.
Android
- Updates: turn on automatic system and app updates; check the security patch level is recent.
- Model support: confirm your phone is still within its update window.
- Lock screen: a six-digit or longer PIN or a password; fingerprint or face unlock for convenience.
- Theft protection: turn on Theft Detection Lock, Offline Device Lock and Find My Device.
- Accounts: use passkeys for your Google account and others that support them; remove the phone number as a recovery method where you can.
- Carrier: set an account PIN or port-out protection.
- Apps: install from Google Play only, and keep "install unknown apps" off.
- Play Protect: keep it on and run a scan now and then.
- Permissions: review accessibility, device admin, location, photos and notification access; revoke what you do not need.
- Charging: use your own adapter.
- Radios: turn off 2G if your phone offers it; switch Bluetooth off when idle; stop auto-joining open Wi-Fi.
- High risk: turn on Advanced Protection.
iPhone
- Updates: turn on automatic updates, including security responses.
- Model support: check your model still receives the current iOS.
- Passcode: six digits or longer, or an alphanumeric code.
- Theft protection: turn on Stolen Device Protection, and Find My.
- Accounts: use passkeys and two-factor authentication on your Apple Account; keep recovery contacts current.
- Carrier: set an account PIN or port-out protection.
- Apps: delete apps you do not use; review App Privacy Reports.
- Profiles: make sure there are no configuration profiles you did not install.
- Permissions: limit photo access to selected photos, and location to "while using".
- Charging: use your own adapter and decline unexpected "trust this computer" prompts.
- Sharing: run Safety Check if you share devices or accounts with anyone.
- High risk: turn on Lockdown Mode.
And for both: restart your phone regularly. It costs nothing and removes any implant that does not survive a reboot.
How should businesses secure employee phones?
Businesses should manage phones with mobile device management (MDM), separate work data from personal data, enforce a minimum OS and patch level, move staff to phishing-resistant sign-in, disable 2G where the platform allows, and plan for lost and stolen devices. High-risk staff should use Lockdown Mode or Advanced Protection.
Figure 11. Defence in layers for people and organisations.
Manage the fleet
MDM lets an organisation enrol phones and apply policy: passcode rules, minimum OS versions, app allow lists, encryption, remote lock and wipe. Apple's platform deployment guide covers enrolment methods including Automated Device Enrollment and User Enrollment (Apple Platform Deployment). On Android, work profiles keep work apps and data in a separate, managed space on the same phone (Android Developers).
For personal phones used for work (bring your own device, or BYOD), separating data is usually more acceptable to staff than full device control. Our guides on what MDM can control and MDM versus MAM for BYOD explain the trade-offs.
One conflict to plan for: iPhones in Lockdown Mode cannot be enrolled in MDM while the mode is on (Apple Support). Decide in advance how you will handle high-risk staff who need both.
A business phone policy checklist
- Enrol every phone that touches company data, with work profiles or User Enrollment for personal phones.
- Enforce patch level. Block access from phones whose security patch level is older than a set number of days, and from models out of vendor support.
- Require a strong passcode and encryption, and turn on theft protection features.
- Phishing-resistant sign-in. Issue FIDO2 security keys or passkeys for staff accounts; remove SMS as a factor and as a recovery path for administrators.
- Apps. Allow only store apps, block unknown sources on Android, and review apps that request accessibility access.
- Cellular. Disable 2G on Android 14 and later managed devices, which Android supports for enterprise fleets (Google Security Blog).
- High-risk staff. Executives, finance, legal, security and communications staff are the people commercial spyware and SIM swappers target; offer them Lockdown Mode or Advanced Protection and a direct line to the security team.
- Carrier accounts. Put port-out protection on company numbers and train the people who manage the carrier account.
- Lost device playbook. Remote lock and wipe, account password reset, token revocation, and carrier contact, all in the first hour.
- Monitor. Watch sign-in logs and, where possible, network traffic from mobile devices. Kaspersky's discovery of Triangulation is the case for this.
- Report without blame. Cloudflare did not reprimand the employees who entered credentials. That is how you get reports in minutes rather than days.
What should phone makers and app makers do?
Phone makers should shorten the patch gap, harden modem and driver firmware with memory safety tools, publish long update commitments and keep an SBOM for every firmware build. App makers should treat every SDK as supply chain risk: inventory it, scan the binaries they ship, pin and verify versions, and fix vulnerable components quickly.
For phone and chip makers
- Shrink the patch gap. Project Zero's advice to vendors is to "follow upstream sources closely, and do their best to provide complete patches to users as soon as possible" (Project Zero). Track upstream fixes for every component you ship, including GPU, modem and Wi-Fi firmware, not only the Android bulletin.
- Harden firmware. Google recommends enabling Clang's Integer Overflow Sanitizer (IntSan) and BoundsSanitizer (BoundSan) in baseband code, because they work in bare-metal environments and catch classes of bugs that baseband attackers exploit. Google also rates baseband remote code execution as critical severity in its vulnerability rewards program (Google Security Blog).
- Adopt memory tagging in hardware and turn it on for the most exposed code.
- Publish support periods and meet them. Google and Samsung's seven-year commitments set the bar for flagships.
- Keep an SBOM per firmware build for the OS image, modem, Wi-Fi and Bluetooth firmware and preinstalled apps, so that when a component vulnerability is disclosed you know within hours which models ship it.
- Vet preinstalled apps. The CamScanner dropper had already appeared in some preinstalled apps on phones. Anything you put in the system image is your responsibility.
- Run a disclosure program and a bug bounty that covers firmware.
For app makers
- Know what is in every build. Generate an SBOM for each release that lists every library and SDK with its version, from both your dependency manifests and the final binary. Our guide to SBOMs for mobile applications explains formats and tooling.
- Scan what you ship, not only what you declare. SDKs can pull in their own dependencies, and build steps can add code. Scanning the final APK and IPA catches what manifests miss, as XcodeGhost showed for build tools.
- Pin and verify SDKs. Lock SDK versions, verify checksums or signatures, and use Apple's SDK signature validation in Xcode. Review the privacy manifests Apple requires for common SDKs.
- Review SDK updates like code changes. Look for new permissions, new network destinations and dynamic code loading.
- Test against a standard. The OWASP Mobile Application Security Verification Standard (MASVS) is the industry standard for mobile app security (OWASP MAS). Our MASVS testing guide maps it to practice.
- Protect your build pipeline and signing keys. A stolen signing key or a poisoned build machine turns your app into the delivery vehicle.
- Keep secrets out of the app. API keys and tokens embedded in an APK or IPA can be extracted by anyone who downloads it.
- Fix fast. When a component in your app has a known vulnerability, ship the fix in the next release, not the next quarter.
What is inside a phone's software supply chain?
A phone's software supply chain is every company and component whose code ends up on it: the OS vendor, the chip vendor's drivers and modem firmware, the phone maker's software, carrier and partner apps, every app developer, every SDK inside those apps, and the build tools that compiled them. A weakness at any link reaches the phone.
Most of the attacks above are supply chain problems in disguise:
- The Mali patch gap was a chip vendor driver fix that did not flow downstream.
- The Exynos modem vulnerabilities were in firmware from a chip supplier, shipped by several phone brands and even cars.
- SpinOk, SparkCat and CamScanner were SDKs inside legitimate apps.
- XcodeGhost was a compromised build tool.
- Even FORCEDENTRY ran through a JBIG2 decoder that Project Zero said was derived from Xpdf, open source code from long before the iPhone.
Each layer has its own owners and its own release schedule, which is why a single unfixed component can sit on millions of phones. The same pattern runs through every connected device in a home, as our guide to how smart home devices get hacked shows; a phone is often the remote control for all of them. The chips themselves, and the firmware they carry, have a supply chain that our post on fabless chip designers and the software supply chain explores, and our laptop teardown follows the same pattern through a PC. For firmware specifically, see our firmware supply chain security guide.
The practical answer for anyone who builds software for phones is the same at every layer: know exactly what is in each build, know which of those components have vulnerabilities that matter, and be able to ship a fix fast.
How Safeguard helps app makers and device makers
Safeguard covers the software supply chain of a mobile app or device in one platform: source and binary inventory including APK and IPA uploads, SBOMs in CycloneDX and SPDX, vulnerabilities ranked by reachability, SAST and secrets scanning on your own code, fix pull requests gated by your CI, Gold patched packages, supplier risk and compliance evidence.
Mobile teams often assemble this from several tools: a source composition analysis (SCA) tool, a mobile app security testing product, a separate binary scanner, an SBOM manager and a GRC platform for audits. Each keeps its own inventory. Safeguard does the whole job end to end, from the first inventory to the merged fix and the audit evidence, from one inventory and one vulnerability model.
Figure 12. From APK and IPA to a fixed build, in one platform.
Scan the APKs and IPAs you actually ship
In the Safeguard console, Integrations > Upload App Binary accepts compiled software including Android APKs and iOS IPAs, as well as Windows EXE, MSI and DLL files, VM and disk images (QCOW2, VMDK, VHD and OVA) and container tarballs. Safeguard unpacks the file, identifies embedded libraries, SDKs and versions, maps them to known vulnerabilities with EPSS (Exploit Prediction Scoring System) and known exploited enrichment, records licences, and turns the result into a project with a full SBOM. That covers SDKs you never declared in a manifest and code added during the build. Firmware images may need to be extracted into their component files first.
Binary analysis has limits, and it is better to say so plainly: heavy obfuscation, encryption or unusual packing can hide components from any binary scanner, and statically linked code may show few identifiable libraries. Pair binary scans with SBOMs from your SDK suppliers. More detail is in our guide to scanning app binaries with Safeguard and our explainer on binary composition analysis.
SDK SBOMs your customers and auditors can use
Every project produces an SBOM you can export in CycloneDX or SPDX, among many other formats. For an app maker, that means one SBOM per release that lists every SDK and its version, ready for an enterprise customer's security review. Attestation and signing with SLSA, Sigstore and in-toto give the SBOM and the build provenance behind it.
Vulnerabilities ranked by what your code actually reaches
A mobile app with dozens of SDKs can map to hundreds of known vulnerabilities, many in code paths the app never calls. Safeguard's reachability analysis ranks vulnerabilities by whether your code can reach them, combined with exploitation signals, so a small mobile team works on the handful that matter first.
SCA, SAST and secrets scanning for the code you own
For the code in your repositories, including Gradle and CocoaPods dependency manifests, Safeguard runs SCA across the full transitive dependency tree, SAST on your own source, and secrets scanning that catches API keys and tokens before they are compiled into an app anyone can download and unpack (securing CocoaPods and Gradle dependencies).
Supplier risk and compliance evidence
Safeguard's Third-Party Risk Manager lets you request SBOMs from SDK vendors and other suppliers, score them and monitor them continuously. The compliance suite collects evidence for frameworks such as SOC 2 and ISO 27001 from the same scans and fixes, so the work your mobile team already does shows up in the audit. Safeguard itself is SOC 2 Type I and Type II audited.
Deploy where your code lives
Safeguard runs as SaaS in US and EU regions, in your own VPC, or fully on premises. It can run in Zero AI mode with no language model in the loop, or with your own model.
How this compares
Mobile teams commonly pair a source SCA tool such as Snyk or Black Duck with a mobile app security testing product such as NowSecure, or the open source MobSF, plus a GRC platform for audits. Those are capable tools, and dynamic mobile testing products go deeper on runtime behaviour of the app itself. Safeguard's case is consolidation and remediation: one inventory across source and the binaries you ship, one vulnerability model with reachability, fix pull requests and patched packages rather than tickets, and supplier risk and compliance evidence from the same data. Safeguard can also sit alongside an existing mobile testing tool and take over SBOM delivery and fixing.
Fix it without waiting for upstream
When a vulnerable component sits inside an SDK or library you depend on, you often cannot wait for its maintainer. Safeguard opens a fix pull request that your own CI builds and tests, and Safeguard Gold supplies patched builds of open source packages when upstream has no fix yet.
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 upgrades or patches the vulnerable dependency. Your CI builds and tests it, and your policies decide what can merge, so nothing changes in your app without passing your own gates.
For open source components where upstream has not shipped a fix, or the fix is in a version you cannot move to without breaking the app, 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 mobile team waiting on an SDK vendor or an abandoned library, that is the difference between shipping the fix in the next release and carrying a known vulnerability for months.
The same logic applies to device makers. The Mali case is what a patch gap looks like when every link waits for the one above it. A team that can patch a vulnerable open source component in its own build closes its part of that gap in days.
If you build or resell mobile apps, devices or managed mobility services and want to offer SBOMs, vulnerability management and fixes to your customers, you can partner with Safeguard.
The takeaway
Phones are the most personal computers people own, and the attacks against them follow the money and the effort. Most compromises are cheap: a text, an app, a carrier phone call, a passcode seen over a shoulder. The defences for those are equally cheap, and the checklist above covers them. The expensive attacks, the zero-click chains and modem exploits, are real and well documented, and the best answers to them are prompt updates, regular restarts and the high-risk modes Apple and Google now ship.
Underneath all of it is a supply chain problem. A phone runs code from dozens of companies, and every app carries code its own developer did not write. The fewer unknown components in that chain, and the faster each known vulnerability is fixed, the less there is for any attacker to use.
Try it yourself
Click through these interactive Safeguard demos to see the steps described in this guide.
Test yourself
How do most phones actually get hacked?
Mostly through the person holding them. A fake delivery text, an app installed from a link, a stolen phone with a guessable passcode, or a carrier employee talked into moving a number account for far more compromised phones than spyware does.
What is SIM swapping, and why does it work?
An attacker convinces your carrier to move your number to a SIM or eSIM they control, then receives your calls and texts, including one-time codes that reset email, bank and crypto accounts. The attack goes through the carrier, not your phone, and needs enough personal information to pass the carrier's checks.
What is a zero-click exploit?
It compromises a phone without the owner tapping anything. The attacker sends data the phone processes automatically, such as a message attachment or a call, and a bug in that processing gives them code execution. Pegasus, sold by NSO Group, has used such chains against iPhones and WhatsApp.
How can a malicious SDK put a trusted app at risk? Give a case.
An SDK runs inside the app with every permission the app holds, so the user, the store and often the developer never notice. In 2023 Doctor Web found SpinOk in 101 Google Play apps with at least 421,290,300 cumulative downloads; it could collect file information and replace clipboard contents.
Why do outdated Android phones stay vulnerable?
A fix has to pass through several companies, including chip vendors and phone makers, who each integrate, build, test and ship it per model. Each hop adds delay, and once a model's support ends, known vulnerabilities stay open for good.
Is juice jacking real?
Partly. It has been demonstrated since 2011, but reviews found no credible reported cases outside research. In 2025 researchers showed ChoiceJacking, which bypassed existing protections on every tested phone from eight vendors, so the risk is not zero.
What do Lockdown Mode and Advanced Protection do?
They are opt-in modes for people at high risk. They switch off or restrict features attackers use as entry points, such as complex message attachments, some web technologies, unknown apps, 2G and USB data, in exchange for some convenience.
Frequently asked questions
Can a phone be hacked without clicking anything?
Yes. Zero-click exploits such as FORCEDENTRY (CVE-2021-30860) and BLASTPASS (CVE-2023-41064) compromised iPhones through iMessage with no interaction, and Project Zero found Exynos modem bugs that needed only a phone number. These attacks are expensive and targeted. Prompt updates, regular restarts, and Lockdown Mode or Advanced Protection are the main defences.
How can I tell if my phone has been hacked?
Look for changes you did not make: unfamiliar apps, accessibility or device admin permissions granted to unexpected apps, configuration profiles on iPhone, sign-in alerts, password reset emails, sudden loss of mobile service and unexplained carrier charges. Battery drain alone is weak evidence. For suspected spyware, watch for Apple threat notifications and seek expert forensic help.
Is an iPhone more secure than an Android phone?
Both are strong when updated, and both have been hit by serious exploit chains. iPhones get updates directly from Apple, while Android fixes pass through chip vendors and phone makers, so update speed and support length vary by brand. A recent, supported Android flagship with long update commitments is far safer than an old phone of either kind.
Does a factory reset remove spyware?
It removes most malware and stalkerware that lives in apps and user data. Amnesty believed later Pegasus versions did not survive a reboot at all, while threats that compromise firmware are rarer and may survive. After a reset, restore apps from the store rather than a full backup, and change passwords for your main accounts from a trusted device.
Is it safe to charge my phone at a public USB port?
The risk is low but not zero. Reviews found no credible juice jacking cases against phones outside research, but the 2025 ChoiceJacking study bypassed the data-access prompts on phones from eight vendors. Use your own adapter in a mains socket or a power bank, use a charge-only cable on public USB, and keep the phone locked while charging.
How do I protect myself from SIM swapping?
Ask your carrier for an account PIN, SIM lock or port-out protection. Remove your phone number as a recovery option on email, banking and crypto accounts, and switch to passkeys, security keys or an authenticator app. Keep your date of birth and address off public profiles. If your phone suddenly loses service, call your carrier from another phone immediately.
Should I turn on Lockdown Mode or Advanced Protection?
Turn it on if you are a journalist, activist, executive, public figure or anyone who has received a threat notification. Lockdown Mode blocked the BLASTPASS zero-click chain, and both modes cut the features attackers use most. Everyone else can try it; the cost is some broken websites, restricted attachments and, on Android, no sideloading.
How can app makers stop malicious SDKs getting into their apps?
Keep an SBOM for every build that lists every SDK and version, scan the final APK and IPA rather than only the manifests, pin and signature-check SDK versions, and review SDK updates that add permissions, network destinations or dynamic code loading. Apple's SDK privacy manifests and signatures help on iOS. Fix vulnerable components in the next release, patching them yourself if upstream is slow.