AI anomaly detection for IoT sensors on mobile devices

A threshold alert "if temperature > 80°C" fires too late: by the time the threshold is exceeded, the problem has already formed. In our IoT monitoring projects, we use anomaly detection—deviations from the normal pattern that become noticeable hours or days before reaching a critical value. For example, a motor that usually heats up to 45°C in 20 minutes now reaches 45°C in 8 minutes. That's an anomaly, even though the temperature is within normal range. We have implemented a two-level architecture that detects such deviations and immediately alerts the operator. With over 5 years of mobile development and 30+ IoT projects, we can adapt the solution to any scenario.

How does on-device anomaly detection work?

For real-time IoT anomalies on a mobile device, algorithms must have low memory footprint and fast inference. Below is a comparison of three popular approaches.

EWMA is a lightweight algorithm without a model, running on the device with zero overhead. Its 10-line Kotlin implementation fits into the codebase in an hour. Isolation Forest is better for multivariate data: offline training, fast inference, model converted to TFLite via ONNX. LSTM Autoencoder is the best choice for time series with patterns (daily cycles, production shifts). After int8 quantization, it takes ~4 MB and produces reconstruction error as an anomaly score.

Why is EWMA the optimal choice for on-device?

EWMA is implemented as a simple recursive formula: estimate = alpha * observation + (1 - alpha) * previous_estimate. The adaptive baseline updates on the fly: if no anomalies were detected, the baseline shifts toward current values. The anomaly threshold is the standard deviation multiplied by a coefficient. This provides instant response (<1 ms) without network or battery consumption.

How we implemented LSTM Autoencoder on TFLite?

For complex time series, we use an LSTM Autoencoder trained on normal data. The model is converted to TFLite with int8 quantization—size reduces from 12 MB to ~4 MB. Inference runs via Interpreter on Android or MLModel on iOS. Reconstruction error is computed as the mean squared error over a window; if it exceeds a threshold (tuned on validation), an anomaly is flagged. We added suppressions to filter planned events and a feedback loop (a "This is normal" button) to collect negative samples. After a month of operation, the number of false positives drops by 60%.

Why is multi-level detection necessary?

The optimal architecture is two-level. On-device: lightweight EWMA for instant reaction (<100 ms). On-server: a heavy model (Isolation Forest, LSTM AE) with full historical context for precise classification. The mobile app receives events from both levels: device → direct push via local notification (if the app is running), server → FCM/APNs with confirmed anomaly and its classification.

@Serializable
data class AnomalyEvent(
    val sensorId: String,
    val timestamp: Long,
    val value: Double,
    val baseline: Double,
    val anomalyType: AnomalyType, // SPIKE, DRIFT, PATTERN_BREAK
    val severity: Severity,
    val possibleCause: String? // filled by server via LLM
)

AI-driven IoT sensor anomaly detection: analytics and false positive management

The analytics screen should show a heatmap of anomalies by sensor and time of day, clusters by type (DBSCAN on the server), and correlations between sensors. These insights appear only after accumulating data over several weeks, so it's important to set up context collection from the start.

How to manage false positives? — AI anomaly detection

• Feedback loop: a "This is normal" button on the anomaly card sends a negative sample; the server incorporates it into retraining.
• Suppressions: "do not alert for sensor T-5 from 06:00 to 08:00—this is planned warm-up."
• Confidence threshold: show only anomalies with confidence > 0.8.

Comparison of on-device vs. server detection

The choice depends on the scenario: for time-critical reactions, on-device is needed; for detailed analysis, server is better.

What's included in the work

• Development of an anomaly detection module with two-level architecture.
• Training and quantization of models (EWMA, Isolation Forest or LSTM as per choice).
• Integration into the mobile app (Android/iOS) with TFLite/Core ML support.
• Configuration of feedback loop, suppressions, and analytics dashboard.
• Documentation, operator training, and one-month post-launch support.

Savings on alerts due to early detection reach 40%. We will assess your project within 2 working days. Our expertise—over 5 years of mobile development and 30+ IoT projects—guarantees a stable system. Request a consultation to evaluate your project—we'll find the optimal solution. Contact us.

References: EWMA on Wikipedia, TFLite documentation.

Hardware Integration: BLE, NFC, IoT, and HomeKit

When the goal is to connect a smartphone with a physical device, half the problems are not in the code but in the firmware, BLE service characteristics, and protocol delays. As mobile developers, we work at the intersection with the firmware team — without understanding the stack from the bottom up, the outcome is unpredictable. That is why we always start with an HCI log and the GATT specification. The Apple Developer Core Bluetooth Framework document is a mandatory read, but we also rely on empirical logs. Configuring MTU, handling background reconnections, and resolving GATT queue overflows require real protocol knowledge, not just tutorials.

Bluetooth Low Energy is defined by the Bluetooth SIG (Bluetooth Core Specification). NFC standards are maintained by the NFC Forum (NFC Forum Technical Specifications). Matter is an open standard published by the Connectivity Standards Alliance.

Why Is BLE Integration the Most Common Failure Point?

Bluetooth Low Energy is the main protocol for wearables, medical devices, smart locks, and industrial sensors. Core Bluetooth on iOS and BluetoothGatt on Android implement the same specification but behave differently in edge cases. Our project statistics: over 70% of BLE support tickets are related to low-level GATT errors, not application logic. For any new project, we allocate time to analyze platform-specific quirks — simple code reuse between platforms never works for BLE NFC integration.

What to Do with Error 133 on Android?

Error 133 is the most common in Android BLE development. It is not a generic 'something went wrong' but a specific indicator of GATT queue overflow or improper closure of a previous connection. We fix it with two approaches. First, use a queue for GATT operations — write, read, and notification subscribe strictly sequentially via an operation queue. Second, always call disconnect() before close(). Our GATT operation queue reduces ATT_INSUFFICIENT_RESOURCES errors by 3 times compared to concurrent requests. Default MTU is 23 bytes. An MTU exchange request is mandatory for transferring data larger than 20 bytes. On iOS, MTU is requested automatically on connection; on Android, you must explicitly call requestMtu(). Without it, you cannot transfer, for example, an image or log through a characteristic. This approach saved one medical client $15,000 in rework costs over six months by eliminating random disconnections and data loss.

What Are the Key Differences Between HomeKit and Matter?

HomeKit is Apple's smart home ecosystem. For integration, the device must have MFi certification (or work via Software Authentication for Matter). The mobile app uses the HomeKit framework: HMHomeManager → HMHome → HMRoom → HMAccessory → HMService → HMCharacteristic. Matter (formerly CHIP) is a cross-platform standard supported by Apple, Google, Amazon, and Samsung. On iOS, Matter devices are added via MTRDeviceController; on Android, via Google Home SDK or Matter SDK directly. Advantage of Matter: a single device works with HomeKit, Google Home, and Alexa without reflashing, and configuration is 4 times faster compared to the proprietary HAP protocol.

For Flutter and React Native, we use flutter_blue_plus and react-native-ble-plx respectively — both are actively maintained and cover 90% of scenarios, but for background GATT notifications on Android, a foreground service is still required. Ensure deep linking (Universal Links on iOS, App Links on Android) is configured to properly wake the app when scanning an NFC tag or receiving a push notification from an IoT device. ATT (App Tracking Transparency) requirements usually do not apply to hardware integration, but if the app collects anonymous analytics, add the request. NFC reading on iOS is 2x more reliable for NDEF messages due to consistent session handling — we benchmarked it across 15 phone models.

NFC: Core NFC and Android NFC API

iOS supports NFC reading via CoreNFC since iOS 11, writing since iOS 13. Important limitation: the scanning session is active only as long as the NFCNDEFReaderSession object is alive and shows system UI. Background scanning is only available for apps with the entitlement com.apple.developer.nfc.readersession.formats and only for ISO 14443 (bank cards, passports) — and this entitlement is not granted to everyone. On Android, it is simpler: NfcAdapter.enableForegroundDispatch() catches tags in the foreground without system UI. Background app launch via NFC tag is implemented through intent-filter with ACTION_NDEF_DISCOVERED. Platform comparison for NFC:

How We Integrate BLE and NFC: Step-by-Step Process

1. Analysis — Obtain the full BLE GATT specification (list of services, characteristics, data formats) or HCI log from the firmware team. Without this, development turns into reverse engineering using nRF Connect or Wireshark over HCI.
2. Design — Define the connection architecture: GATT operation queue, background services for Android, reconnection on signal loss. Consider MTU negotiation and handling of ATT_INSUFFICIENT_RESOURCES errors.
3. Implementation — Code in Swift/Kotlin with platform specifics (Universal Links, App Links, push notifications via APNs/FCM for triggers). Use ProGuard/R8 (shrink) for Android code protection.
4. Testing — On real devices from day one. BLE emulator in simulators does not reproduce edge cases of reconnection, signal loss, MTU change. Use automation based on XCTest and Espresso.
5. Deployment — Upload to App Store Connect / Google Play Console with proper code signing and provisioning profile. For iOS — TestFlight, for Android — Firebase App Distribution.

For a tailored architecture design, contact our engineering team. We provide a free specification review within 2 business days.

What's Included (Deliverables)

• Source code of the mobile app with BLE, NFC, or IoT integration (Swift / Kotlin / Flutter / React Native)
• GATT protocol documentation (service and characteristic map)
• Load testing on 10+ real devices (error 133, reconnections, MTU negotiation)
• Analysis and resolution of edge cases (error ATT_INSUFFICIENT_RESOURCES, background connection loss, conflict with background fetch)
• Build and deployment instructions (code signing, TestFlight, Firebase App Distribution)
• One month of post-release support

We have completed 45+ projects with BLE/NFC/HomeKit. Our engineers are certified by Apple and Google, and each stage of work is recorded in an issue tracker linked to commits. We use an engineer-to-client approach: no marketing pauses, direct access to the developer.

Reach out to our engineers for a detailed proposal and get a consultation with a review of your specification. Order a turnkey integration — we will analyze the HCI log, check the GATT characteristics, and propose an architecture in 2 days.