The Best Home Security Cameras

We installed 10 security cameras across two test homes — one in Austin, Texas and one in Grand Rapids, Michigan — and ran them continuously for four months through summer heat, thunderstorms, and early winter snow. The performance gap between cameras that cost $50 and cameras that cost $300 is narrower than you would expect in daylight. At night, the gap is enormous. Person detection accuracy ranged from 70% to 97% in our controlled walk tests, and false alarm rates varied from 2% to 34%. A camera that alerts you every time a tree branch sways is worse than no camera at all.

The subscription question matters as much as the hardware. Every camera in this test offered a subscription tier for cloud storage, AI detection features, or both. We evaluated each camera with and without its subscription to determine what you actually get for free versus what requires a monthly fee. Some cameras are essentially paperweights without a plan. Others are fully functional standalone.

#1 Pick

Best home security camera overall

Arlo Pro 5S by Arlo

The Arlo Pro 5S delivered 2K video that stayed sharp at night thanks to its integrated color night vision spotlight. Person detection hit 96% accuracy at 30 feet with only 4% false alarms — the best ratio in our test. The magnetic mount makes repositioning effortless, and dual-band Wi-Fi 6 ensured zero video drops across four months of continuous operation.

How we tested

We evaluated every camera across five categories, each weighted equally in our final score. Video quality was assessed at three lighting levels: full daylight (10,000+ lux), dusk (50 lux), and night (0.1 lux). We captured standardized test scenes — a person walking across a driveway, a license plate at 20 feet, and a package on a porch — and compared resolution, color accuracy, and detail retention across all ten cameras at each light level.

Person detection accuracy was measured using a controlled walk-test protocol. A tester walked past each camera at 15, 30, and 50 feet, repeating the walk ten times at each distance. We recorded detection rate (correct alerts) and false positive rate (alerts triggered by animals, vehicles, or shadows). The same tester, wearing the same clothing, performed all walks to ensure consistency.

False alarm rates were logged over the full 120-day test period. We tracked every notification each camera generated and categorized it as a true positive (person), true negative (no event), false positive (alert but no person), or false negative (person present, no alert). Cameras with high false positive rates received score penalties regardless of detection accuracy — a camera that catches every person but also alerts on every car headlight reflection is unreliable.

Battery life was measured in real-world conditions for wire-free models. We recorded the number of days from full charge to 10% battery with each camera set to maximum sensitivity and highest resolution. For wired cameras, we measured power consumption with a Kill-a-Watt meter. Weather resistance was evaluated by the Michigan installation, which experienced temperatures from 98°F to 14°F, rain, snow, and sustained winds above 30 mph.

Why the Arlo Pro 5S won

The Arlo Pro 5S posted the highest combined score because it had no weak categories. Its 2K HDR video delivered the sharpest detail at night — we could read a license plate at 20 feet in 0.1 lux conditions, something only one other camera (the Reolink Argus 4 Pro) matched. The integrated spotlight activates on motion to provide full-color night footage, eliminating the green-tinted infrared look of cheaper cameras.

Person detection was the Arlo's standout metric. At 30 feet, it correctly identified a person in 96 out of 100 walk-tests and generated only 4 false positives — a 96/4 ratio that was the best in our test. The Ring Stick Up Cam Pro hit 93% detection but with a 12% false positive rate. The Google Nest Cam (battery) reached 91% detection with 8% false positives. The Arlo's on-device AI processing makes the difference here: it analyzes the video feed locally before sending alerts, which reduces both latency and false triggers.

Battery life was good but not exceptional. On maximum settings, the Pro 5S lasted 47 days between charges — compared to 62 days for the Reolink Argus 4 Pro and 38 days for the Ring Stick Up Cam Pro. The magnetic charging cable makes recharging simple, and Arlo offers a $13/month solar panel that maintained the battery at 100% during our Texas summer test.

The competition

Google Nest Cam (battery) — 8.5/10. The cleanest app experience in our test. Google Home integration is seamless, and the familiar-face recognition (with a Nest Aware Plus subscription at $12/month) is genuinely useful for reducing alerts when family members arrive. Video quality is excellent in daylight but drops to 1080p-equivalent detail at night. Person detection scored 91% at 30 feet with 8% false positives. Battery life was the weakest at 31 days on maximum settings.

Ring Stick Up Cam Pro — 8.3/10. The best value if you already have Ring devices. The ecosystem integration with Ring doorbells and alarm systems is unmatched, and the $3.99/month Ring Protect Basic plan is the cheapest subscription in our test. Video quality is solid 1080p with HDR, but the lack of 2K resolution was visible in our license plate test — readable at 15 feet but blurry at 20. Person detection scored 93% with a 12% false positive rate, which means you will get some phantom alerts.

Reolink Argus 4 Pro — 8.1/10. The best no-subscription option. This camera stores footage locally on a microSD card and offers free basic cloud storage with no monthly fee. The 4K resolution is the highest in our test, though diminishing returns are real — the extra pixels did not meaningfully improve our license plate test compared to the Arlo's 2K. Battery life was the best at 62 days. The trade-off is the app, which feels dated and lacks the polish of Arlo, Google, or Ring.

What to look for in a security camera

Resolution matters less than night vision quality. A 4K camera with poor night vision is less useful than a 2K camera with a good spotlight. Most security events happen in low light — prioritize cameras that perform well below 1 lux.

Calculate the total cost of ownership. A $50 camera with a $10/month subscription costs $170 in the first year and $120 every year after. A $200 camera with no subscription costs $200 total. Over three years, the cheaper camera costs $290 versus $200. Factor subscriptions into your budget before buying.

Local storage is your backup plan. Cloud storage is convenient, but if your internet goes down during a break-in, cloud-only cameras capture nothing. Look for cameras with a microSD slot or local NVR support as a fallback.

Person detection accuracy is non-negotiable. A camera that cannot distinguish a person from a car or a shadow will train you to ignore its alerts. Look for on-device AI processing — it reduces false positives and delivers faster notifications than cloud-only analysis.

Check your Wi-Fi before buying. Most security camera issues are actually Wi-Fi issues. If the camera's mounting location is far from your router, choose a camera with Wi-Fi 6 support or consider a wired PoE option. We saw zero dropped frames from cameras with Wi-Fi 6 versus 2-3% from Wi-Fi 5 models at the same distance.

Gold Tested Score: Video quality: 9.5/10, Night vision: 9.4/10, Person detection: 9.6/10, Battery: 8.8/10, Weather resistance: 9.3/10. Overall: 9.3 (/ 10)

The bottom line

The Arlo Pro 5S earned our Gold Certified pick with a 9.3 overall score, more than two points above the category average. Its combination of sharp 2K night video, industry-leading person detection accuracy, and reliable wire-free operation makes it the camera we recommend to most homeowners. The subscription requirement is its only real weakness — if you want to avoid monthly fees entirely, the Reolink Argus 4 Pro is a strong alternative with local storage and no recurring costs.

Resolution, field of view, and night vision: what you actually need

Resolution: 1080p (2 megapixels) is sufficient for general security monitoring — identifying whether a person is present, which direction they went, and their approximate size and clothing. 2K (4 megapixels) provides enough detail to identify facial features at distances up to 20 to 25 feet — adequate for doorbell cameras and entry-point monitoring. 4K (8 megapixels) provides identification-quality detail at distances up to 35 to 40 feet and is useful for covering large areas (driveways, backyards, parking areas) where subjects may be far from the camera. Higher resolution produces larger video files, consuming more storage and bandwidth — a 4K camera recording continuously generates 10 to 15 GB of data per day versus 3 to 5 GB for a 1080p camera.

Field of view: Measured in degrees, this indicates how wide the camera's view is. A 90-degree field of view (standard for most security cameras) covers a focused area directly in front of the camera — appropriate for monitoring a specific entry point, window, or corridor. A 130-to-180-degree wide-angle view covers a broader area (an entire driveway, a full room, or a wide porch) but with reduced detail at the edges due to lens distortion. The trade-off between field of view and detail is constant: wider views show more area but with less per-pixel detail, meaning a wide-angle camera needs higher resolution to maintain the same identification quality as a narrow-angle camera at the same distance.

Night vision: Infrared (IR) night vision illuminates the scene with invisible IR light and captures the reflected light with the camera sensor. Most security cameras include IR LEDs that illuminate a distance of 20 to 30 feet in complete darkness. The resulting image is grayscale (IR sensors cannot distinguish color). Color night vision (used by cameras with ambient-light sensors or supplemental white LED illumination) provides color images in low light by using larger sensor pixels that capture more light — useful for identifying vehicle and clothing colors, which IR grayscale cannot distinguish. Color night vision requires some ambient light (streetlights, porch lights) to function; in complete darkness, it falls back to IR grayscale.

Optimal placement and setup

Camera placement has a larger impact on detection accuracy than most buyers realize. During our 120-day test, we experimented with mounting heights and angles, and the data was unambiguous: cameras mounted between 8 and 10 feet high, angled 15-20 degrees downward, produced the highest person detection rates and the fewest false positives across all models. Mounting too low (below 6 feet) increases false triggers from pets and passing vehicles because the camera's field of view captures more ground-level motion. Mounting too high (above 12 feet) reduces facial detail and makes person-versus-object classification harder for the on-device AI. For the Arlo Pro 5S specifically, we recommend enabling the activity zone feature to exclude areas with frequent non-human motion — trees, busy sidewalks, or flag poles. In our Austin test home, configuring a single activity zone that excluded the street reduced false positives from 4% to under 1% without affecting person detection rates. Also ensure the camera's spotlight is not aimed at reflective surfaces like windows or car windshields, which can create glare loops that trigger repeated false alerts at night.