Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
The Android security architecture is a stack of defences—verified boot, a hardened Linux kernel, SELinux, a per-app sandbox, and runtime permissions—each built to survive the failure of the layer above.

What Is the Android Security Architecture?
Android’s security architecture is a stack of controls that decides what code may run on a device and what that code may touch: a hardware root of trust, a verified boot chain, a hardened Linux kernel with mandatory access control, a per-application sandbox, and a permission model the user can see and revoke. Each layer is designed on the assumption that the one above it will eventually fail.
That assumption is not academic. Android runs on hardware from dozens of manufacturers, with carrier and vendor code in the middle, so the platform cannot rely on a single trusted supplier the way a closed ecosystem can. The architecture compensates by pushing security downwards, into silicon and into the kernel.
Boot begins in immutable ROM, which verifies the bootloader, which in turn verifies the boot image using Android Verified Boot. AVB uses rollback indexes, so a device will refuse to load an older, vulnerable image that an attacker has re-signed with a stale key. A device whose bootloader has been unlocked shows the familiar yellow or orange boot warning; that state is attested to apps later on, which is why root detection is meaningful at all.
Keys live below the operating system. On most modern handsets, cryptographic operations happen inside a Trusted Execution Environment, and devices with a discrete security chip—Google’s Titan M2, Samsung’s Knox Vault—can use StrongBox Keymint, where private keys never enter the application processor at all. Keys can be bound to user authentication, so a decryption operation only succeeds after a fingerprint or PIN check. File-based encryption then gives each user profile its own keys, which is what makes a work profile genuinely separate from a personal one.
Every installed app receives its own Linux user ID, and that UID enforces file ownership inside the app sandbox. Since Android 5.0, SELinux has run in enforcing mode for every domain, which means a compromised system service is confined by policy rather than by whatever privileges it happened to hold. Seccomp filters cut the reachable syscall surface further, and inter-process communication is funnelled through Binder, where the kernel can check the caller’s identity.
These controls matter because kernel bugs keep arriving. Dirty Pipe (CVE-2022-0847) gave a local attacker arbitrary file overwrite on kernel 5.8 and later, which reached shipping devices including the Pixel 6 and Galaxy S22; CVE-2019-2215, a use-after-free in Binder, was exploited in the wild before it was patched. SELinux and seccomp did not prevent those bugs—they limited what the exploit could reach next.
At the application layer the architecture is mostly about consent and scope. Dangerous permissions are requested at runtime rather than at install time, can be granted once only, and are automatically revoked when an app goes unused for months. Scoped storage removed the free-for-all access to shared storage that older apps enjoyed, and Android 14’s selected-photos access lets a user share three images instead of an entire gallery. Our article on mobile app permissions covers that model in detail.
Distribution is controlled separately. Apps must be signed, and Google Play Protect scans both store and sideloaded packages on device. Server side, an app can call the Play Integrity API to ask Google whether the device passes basic and device integrity checks and whether the binary is the one Play distributed—a far stronger signal than any check the app performs on itself.
Google publishes an Android Security Bulletin on the first Tuesday of each month, and Project Mainline lets roughly thirty core modules—media codecs, the Bluetooth stack, the conscrypt TLS provider—update through Google Play without a full firmware release. This was a direct answer to Stagefright (CVE-2015-1538), where a media-parsing bug reachable by MMS took many vendors months to fix. Device support windows still vary: recent Pixel and Galaxy flagships promise seven years of updates, while budget handsets may get two.
For a UK business, the practical questions are patch level and provenance. Cyber Essentials expects mobile devices in scope to be supported by the vendor, to receive high-risk updates within 14 days, and to install software only from approved sources—criteria that quietly rule out a large share of cheap handsets. The NCSC’s device security guidance for Android sets out the enrolment and configuration baselines, which you enforce through mobile device management.
At Privia Security, our Mobile Application Security Services test how your Android apps use these platform controls, and where they quietly work around them.
Yes, under normal conditions. Each app runs under its own Linux user ID, and SELinux policy blocks cross-app file access even for system services. Data leaves the sandbox only through channels the developer opens deliberately, such as exported activities, content providers, or world-readable files in shared storage. Most real cross-app exposure comes from those misconfigured components rather than from a sandbox failure.
It removes a guarantee rather than creating a vulnerability on its own. With the bootloader unlocked, verified boot no longer refuses modified system images, so an attacker with physical access can persist below the operating system. Play Integrity and hardware key attestation both report the state, which is why banking and payment apps commonly refuse to run on unlocked devices.
Cyber Essentials requires updates rated critical or high risk to be installed within 14 days of release, and Android security patch levels are published monthly. Enforce this through your MDM compliance policy: block access to corporate resources from devices more than one or two patch levels behind, and track handsets that have left vendor support entirely.
It returns signed verdicts about the app and the device: whether the binary and its signing certificate match what Google Play distributed, whether the request came from a genuine Play-installed instance, and whether the device passes basic, device, and strong integrity checks. Because the verdict is validated on your server, it resists the tampering that defeats purely on-device checks.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.