KERALA RISING · FROM EXPORTING PEOPLE TO EXPORTING IDEAS Failure modes and fraud controls HOW IT COULD BE GAMED TH E C O NT RO L TH A T C LOSE S I T Staged photos of the same clean bin, resubmitted. Perceptual hashing against the corpus; timestamp and geofence checks; random human audit of a verified sample. Waste carried into a corridor only to be photographed. A volume-plausibility model per household, cross-checked against the unit’s recorded pickup weights for the route. Reward-point farming by Redemption-velocity caps; partner collusion with a partner reconciliation; anomaly flags on shop. point-to-redemption ratios. Table 7.2, Clean Kerala: gaming vectors and controls. Objections, answered The objections a department secretary will raise first, and the plain answers. Q. Isn’t this just a clean-up campaign with an app bolted on? A. No. A campaign is an event that ends; this is a standing feedback loop that pays the HKS worker, ranks the ward, and reports an audited score every quarter. The app is the least important part; the loop is the point, and the loop persists after the launch noise fades. Q. We already spend on waste and run the HKS. Why add this? 123