What is tailgating in physical security and why is it dangerous?

Tailgating (also called piggybacking) in physical security is when an unauthorised individual follows an authorised credential holder through a controlled entry point without presenting their own valid credential. It is the most common bypass of physical access control — studies consistently show that 94% of physical security breaches in corporate environments involve tailgating as a component of the intrusion. Tailgating is dangerous because: (1) It completely defeats the credential verification at the entry point — an entire investment in access readers, turnstiles, and credential systems is bypassed by following closely behind an authorised person; (2) It typically triggers no alarm — the access control system logs a valid credential event; the second person's entry is invisible to the PACS; (3) Social engineering facilitates it — the authorised person holds the door out of courtesy, unaware of the security implication; (4) High-traffic entry points at peak hours create a 'rush hour effect' where tailgating is almost impossible to prevent through guard observation alone.

How does AI tailgating detection differ from traditional anti-passback?

Traditional anti-passback is a logical rule in the access control system that prevents a credential from being used twice in the same direction within a time window — if card A enters at Door 1, anti-passback prevents card A from entering at Door 1 again until it has exited. Anti-passback detects credential reuse but cannot detect physical body count. AI tailgating detection uses computer vision to count the number of physical persons passing through the entry zone per credential event. This detects: (1) A valid card holder holding the door for an unauthorised second person — not detectable by anti-passback (two different people, one credential event); (2) A 'piggybacking' intrusion where the unauthorised person follows without touching the door — also invisible to anti-passback. The two systems are complementary: anti-passback prevents credential sharing, AI tailgating detection prevents physical body count violations. Together, they address the full range of entry bypass scenarios.

Does AI tailgating detection work with existing CCTV cameras?

AI tailgating detection can be deployed on existing CCTV cameras if the camera meets the technical requirements for the analytics. Requirements: minimum 1080p resolution at the entry zone; 30fps frame rate; camera position providing clear view of the entry point (overhead or 45-degree angle); minimum 100 lux illumination at the entry zone; NIR or white-light illumination for 24/7 operation. If existing cameras meet these requirements (common for cameras installed after 2018 with IP megapixel cameras), the analytics software can be applied as a server-side analytics layer — processing the existing camera stream without camera replacement. If existing cameras are analogue CCTV, SD resolution, or poorly positioned, camera replacement or addition is required. Platform: Genetec KiwiVision and Irisity AEGIS accept standard ONVIF-compatible RTSP streams from any IP camera, making them compatible with existing multi-vendor camera installations.

What is the false positive rate of AI tailgating detection systems?

False positive rate (incorrectly alerting on legitimate single-person entry) is the key operational metric for AI tailgating detection. Current-generation systems report: Genetec KiwiVision Privacy Protector: false positive rate below 2% in controlled deployment conditions with appropriate camera placement. Irisity AEGIS: reported false positive rate of 1.5–3% depending on entry zone traffic pattern. Avigilon AI tailgating: false positive rate below 2% with recommended camera setup. False positive triggers include: baggage/equipment carried by the authorised person (analytics model must be trained to distinguish person body from carried objects); crowded entry conditions with legitimate single-person entries in close temporal proximity; IR/thermal imaging limitations in certain clothing (thick winter coats can cause adjacent person silhouettes to merge). System accuracy improves over 2–4 weeks of operational deployment as the AI model fine-tunes to the specific entry zone's camera angle, lighting, and pedestrian traffic patterns.

Can AI tailgating detection automatically lock the door when tailgating is detected?

Yes — real-time automated response is a core feature of enterprise AI tailgating detection. Integration path: (1) Tailgating event detected by analytics at sub-500ms latency; (2) Analytics platform sends alert via REST API to access control panel (Lenel, Genetec, Honeywell, C•CURE 9000); (3) Access panel triggers door or turnstile lock-close output relay; (4) Simultaneous alert sent to security operations console and mobile app; (5) Event logged with video clip attached for review. Caution: automated door lock in high-traffic environments can create safety hazards (trapped individuals, emergency egress obstruction) — ASDV recommends alert-only mode (alert security team without automated door action) for first 4–6 weeks of deployment while false positive rate is measured and acceptance threshold calibrated. Automated lock response is appropriate for low-traffic, high-security zones (server room, R&D lab) where false positive impact is lower.

What camera position is required for AI tailgating detection?

Optimal camera placement for AI tailgating detection: (1) Overhead position (top-down view, 90° nadir) directly above the entry point — provides the clearest person silhouette separation, minimal body occlusion, and most accurate person count. Camera height 2.5–4m AFF. (2) 45-degree overhead angle (above and to the side of entry) — suitable where ceiling mounting directly above is not possible; slight reduction in person separation accuracy at high densities. Required: both an 'entry side' view (1.5m before the door threshold on the approach side) and a 'threshold crossing' view (at the door plane) to count both the credential presenters and any followers. Minimum camera specifications: 1080p (2MP) minimum; 4MP recommended for high-density entries; 30fps; H.265 compression; minimum 100 lux illumination at camera coverage zone; wide dynamic range for backlit lobby environments with glass facades.

Is AI tailgating detection compliant with India's DPDP Act 2023?

AI tailgating detection in anonymised person-counting mode (detecting number of persons without identifying individuals) does not process personal data and is not subject to DPDP Act consent requirements — body count at an entry zone is not biometric data and is not linked to individual identity. However, if the tailgating detection system also captures and retains video of individuals (standard CCTV connected to AI analytics), the video retention is subject to DPDP Act personal data requirements: consent for continuous surveillance (or a legitimate purpose basis), retention period limitations, and data protection measures. If the tailgating detection system is integrated with face recognition (identifying the tailgater as a specific individual), the biometric data processing requires explicit consent and satisfies DPDP sensitive personal data requirements. ASDV recommends anonymised body-count tailgating detection as the DPDP-lowest-risk architecture — no biometric identification, only physical count.

What lighting conditions are required for AI tailgating detection?

AI tailgating detection using visible-light cameras requires minimum 100 lux illumination at the entry zone for reliable person detection. Below 50 lux, shadow-filled body outlines reduce detection accuracy below operational thresholds. For 24/7 operation in environments with variable or low lighting: (1) NIR-illuminated cameras — cameras with integrated IR illuminators (850nm or 940nm) provide consistent illumination regardless of ambient light; NIR-sensitive image sensors provide clear person detection in complete darkness; (2) White-light supplemental illumination — adding dedicated LED luminaires (warm white, 3000K, 200 lux maintained at entry zone) ensures consistent analytics performance; (3) Wide dynamic range (WDR) processing — for lobby environments with bright glass facade backlight, WDR cameras (120dB WDR) prevent blown-out backgrounds that make foreground persons invisible to analytics. ASDV specifies supplemental entry zone illumination as a standard element of AI tailgating detection system design for Indian corporate lobby environments where natural backlight from glass facades is a frequent challenge.

How does AI tailgating detection integrate with Lenel or Genetec access control?

Integration of AI tailgating detection with Lenel OnGuard and Genetec Security Center uses event-driven API architecture: Genetec KiwiVision — native integration within Genetec Security Center; tailgating events appear as native events in the Security Center alarm panel; automatic alarm can trigger barrier/turnstile lock via Genetec Hardware interface. Lenel integration via REST API — Irisity AEGIS, BriefCam, and Motorola Solutions Ava Corridor all provide REST API event output; Lenel OnGuard REST API accepts external event inputs that trigger alarm responses and access control actions; integration requires Lenel's Open Access module and REST API licence. Event data sent per tailgating detection includes: camera ID, timestamp, entry zone location, number of persons detected, access credential event correlated (if PACS event feed is available), and attached video clip. Response actions configured in the ACS: security console alert, email/SMS notification to security supervisor, and optional door/barrier lock relay output.

AI Tailgating Detection: Computer Vision Controlled Entry, Real-Time Anti-Tailgate & Physical Security

The access control system at the server room door validates the credential perfectly. The OSDP v2 encrypted reader communicates with the controller panel, the panel queries the access rights database, finds a valid authorised record, and releases the door. One person enters. One authorised access event is logged. Except two people walked in.

Tailgating defeats every element of the credential stack — the card, the reader, the panel, the policy — by simply walking behind someone who has been correctly authenticated. Traditional anti-passback prevents the same card from being used twice at the same door, but has no knowledge that a second body followed the authorised cardholder. AI tailgating detection brings computer vision into the entry zone — correlating the credential event timestamp with a real-time person count at the door threshold — and acts on the discrepancy before the second person is three steps inside the protected zone.

AI tailgating detection reduces unauthorised entry through controlled access points by 94% compared to traditional anti-passback rules alone — with deep learning person-detection models processing entry events at 30fps with less than 500ms detection latency. Irisity / Genetec AI analytics data, 2025.

AI Tailgating Detection Platform Comparison

Platform	Detection Method	False Positive Rate	Integration	Hardware Required	Alert Speed
Genetec KiwiVision	Person bounding box + zone crossing count	<2%	Native Genetec SC	Existing IP cameras (1080p+)	<500ms
Irisity AEGIS	Deep learning person detection + event correlation	1.5–3%	REST API (Lenel, Genetec)	Existing ONVIF cameras	<400ms
Motorola Ava Corridor	AI body count + credential correlation	<2.5%	REST API	Ava cameras preferred	<500ms
BriefCam Tailgating	Video synopsis + person tracking	<3%	Lenel, Genetec, Milestone	Existing cameras (2MP+)	<600ms
Avigilon AI Tailgating	Avigilon AI + body count	<2%	Avigilon ACC + Lenel	Avigilon cameras preferred	<500ms

Technical Design: AI Tailgating Detection Architecture

Person detection model (YOLOv8/Faster R-CNN): Person bounding box detection at entry frame at 30fps; persons are counted in entry zone (1.5m approach side) and threshold zone (door plane) simultaneously
Credential event correlation: PACS credential event (timestamped Wiegand/OSDP read from reader) correlated with video frame person count at entry zone; 1 credential swipe = 1 person rule — second person in threshold zone within credential event window triggers tailgating alert
Virtual line counting zones: Configurable virtual lines drawn at door threshold and approach zone in the camera view; count of persons crossing each virtual line per credential event is the detection logic
Anti-passback comparison: Anti-passback detects credential reuse (same card twice); AI tailgating detects physical body count excess per credential event — addressing the primary bypass scenario anti-passback cannot detect (authorised person holding door for unauthorised follower)
REST API integration (Lenel/Genetec): Tailgating event → REST API → ACS event input → alarm panel alert + optional door lock relay output; Genetec KiwiVision provides native integration within Security Center without API development
Camera requirements: Minimum 1080p, 30fps; overhead or 45-degree angle above entry point; minimum 100 lux at entry zone; WDR for backlit lobby; NIR for 24/7 operation — upgrade or addition may be required for existing analogue or SD CCTV installations
DPDP Act 2023 compliance: Anonymised person-count mode (body count only, no identity) avoids biometric data classification; video retention subject to DPDP personal data requirements — 90-day retention maximum for access control purpose; event clips may be retained longer for security investigation purposes
India applications: Corporate campus lobby turnstiles (GCC, IT parks, manufacturing); data centre server room mantrap; financial institution branch vault corridors; government and defence perimeter entry

Future Outlook: 2028–2032

Identity-Linked Tailgating: Real-Time Intruder Identification

The next generation of AI tailgating detection will combine body count detection with real-time face recognition — not only detecting that a second person entered on a single credential event, but identifying who the second person is within the facility's visitor and employee database. If the second person is an unregistered visitor, an automatic security alert with their photo is dispatched to the security console within 3 seconds of detection. If they are an employee who should have used their own credential, a policy violation is logged and their manager is notified. The combination of tailgating detection + identity makes the physical entry point as accountable as a logical network login — every person's presence in a controlled zone is attributed to a verified identity, continuously.

Frequently Asked Questions

Tailgating is when an unauthorised person physically follows an authorised cardholder through a door without presenting their own credential. Anti-passback prevents the same credential being used twice without an exit event — it detects credential reuse. But it cannot detect two different physical persons using one credential event (cardholder holds door for follower). AI tailgating detection counts physical persons per credential event using computer vision — addressing the bypass scenario that anti-passback cannot detect. The two systems are complementary: anti-passback + AI tailgating detection addresses the full range of entry bypass scenarios.

Yes, if existing cameras meet requirements: minimum 1080p resolution, 30fps, correct position (overhead or 45-degree above entry), 100 lux minimum illumination at entry zone. If cameras are 2018+ IP megapixel cameras in good positions, server-side analytics (Genetec KiwiVision, Irisity AEGIS) processes existing ONVIF RTSP streams without camera replacement. Older analogue CCTV, SD resolution, or poorly positioned cameras require replacement or addition of a dedicated tailgating detection camera. ASDV assesses existing camera inventory as part of the AI tailgating detection design process.

Anonymised person-counting (body count only, no identity) does not process personal data and is DPDP-exempt. If the system also captures and retains video of individuals, DPDP personal data requirements apply: consent or legitimate purpose basis, retention period limits (90 days for access control purpose), and data protection measures. Identity-linked tailgating detection (face recognition) requires explicit DPDP consent as biometric data processing. ASDV recommends anonymised body-count tailgating detection as the DPDP-lowest-risk architecture for most deployments.

AI Tailgating Detection: Computer Vision at Controlled Entry Points, Real-Time Anti-Tailgate & Genetec/Lenel Integration

AI Tailgating Detection Platform Comparison

Technical Design: AI Tailgating Detection Architecture

AI Tailgating Detection Design

Identity-Linked Tailgating: Real-Time Intruder Identification

Frequently Asked Questions

Design Your AI Tailgating Detection System