Most personal safety tools share a weakness: someone has to use them. Pepper spray needs aim, an SOS app needs a free hand and a working connection, and a whistle depends on someone nearby. Lavanya Gupta's research paper, Leveraging AI and IoT Technologies for Personal Security: A Smart Wearable Keychain Framework, asks whether a small device could do the work automatically.

The problem

The paper opens with the scale of the issue. It cites the World Health Organization's estimate that about one in three women worldwide experience physical or sexual violence in their lifetime. It also cites a 2023 India Today Group survey in which 40% of women said they feel unsafe travelling alone at night in their own cities. Existing tools, it argues, are manual, hard to use under panic, and often not discreet.

The device

The prototype is a compact keychain combining several components:

  • Voice and motion triggers: natural language processing listens for distress phrases such as "Help!", while anomaly detection watches for unusual movement such as a fall.
  • Location and alerts: GPS and GSM modules send the wearer's live location to emergency contacts.
  • Deterrence and evidence: a loud 100 dB alarm alerts bystanders, and a concealed microphone records audio that could later serve as evidence.

Early versions were built on a breadboard; a custom printed circuit board then shrank the design to keychain size. A pepper spray module was dropped because of size constraints.

How it was studied

The paper follows a Design Science Research approach and frames the device through Complex Adaptive Systems theory, treating the voice recogniser, motion detector, GPS and microphone as agents whose interaction produces protective behaviour that no single part provides alone. It asks three questions: whether AI-triggered activation can be as dependable as traditional devices, whether a concealed microphone improves the quality of evidence, and whether automation can reduce escalation by an attacker.

What it found

In initial testing the prototype activated three to five seconds after a trigger cue, which the paper describes as far faster than manual activation. It reports tests across fifty prototype trials with different voice profiles and locations. The limitations are stated plainly: false positives from voice activation, dependence on GSM coverage, and the need for more behavioural data to make anomaly detection reliable. Planned improvements include adjustable alertness levels, multilingual commands such as Hindi, and video documentation.

Why it matters beyond the device

The paper also points to questions that technology alone can't answer. As audio evidence improves, the rules on admissibility, chain of custody and the privacy of recorded people have to keep pace. And alerts that reach police in real time raise both opportunities and concerns for smart-city systems.

Sources

  • Lavanya Gupta, Leveraging AI and IoT Technologies for Personal Security: A Smart Wearable Keychain Framework (research paper)
  • World Health Organization, "Violence against women" — who.int
  • Hyndavi, Nikhita and Rakesh, "Smart Wearable Device for Women Safety Using IoT," IEEE, 2020 — ieeexplore.ieee.org
  • Farooq et al., "The Role of IoT in Woman's Safety: A Systematic Literature Review," IEEE Access, 2023 — doi.org