First, remember that an estimate is highly unlikely to be 100% correct, so do the best you can with the information that is available to you.
Second, break down the tasks as far as you reasonably can, and get figures for similar low level tasks from other projects that have been done in the past. Don't worry if they are not identical, as long as they are broadly similar.
Third, get a consensus. Don't just ask one person. Get a few people in a room and ask them for their best estimates. Debate, discuss, argue, and then choose a value that most people will accept.
Fourth, ask questions. If it feels wrong, you either don't understand the problem, or someone else doesn't. Ask them to explain why they gave their estimate, and do this (nicely) in public, so that others can also provide their input and challenge the estimate.
Fifth, learn from what has gone before. Learn whose judgement you can trust, and whose you cannot. Learn from your mistakes, and also from your successes.
Sixth, Refine your estimates as the project goes on. You can change estimates. Just communicate the changes openly and honestly.