How can I tell if my alarm has been "down" for a period of time, assuming I don't believe the records of the alarm company in such cases?
There is a plethora of various devices suitable for an alarm system, both off-the-shelf and homemadeable. You can cheaply roll out a camera system with a cheap PC with Linux and a TV-input card with a 4051 analog-multiplexer-based parallel-port-controlled switch for selecting cameras for slow-motion video. Frame rate sucks as there is no means of syncing the cams, so you have to give the card couple frames to lock after every cam change, but you should be able to get about four switches per second (which gives you one frame every 2 seconds for all 8-cams cycle). You can also switch them irregularly, so you get better framerate on the cams whose signal indicates something's happening. The photos may be recorded to a suitably big hard drive in a circular buffer, with suspicious frames optimally flagged for more permanent storage. This, together with a Net connection (and a backup phone line and eventually a cellphone) gives you a relatively very cheap system. Don't forget the UPS. Don't have this system as the only one; the adversary can switch off the power to the house and starve the computer. A new standard for GSM phones is appearing here: MMS - multimedia SMS messages. With suitably equipped phone you can send/receive short sound and video clips, and - which is important for us now - pictures. The alarm system then can send you a picture of the neighbor's cat getting into the house and tripping the sensors instead of just a panic-inducing generic message about activated PIRs and IR gates. I'd advise against relying on 802.11 cameras; they are too easy to both detect and jam. They can add some finishing touch though, and an attempt to jam the cams can be a possible alarm (or at least suspicion) source on its own. A powerful and reliable device may be a door/window-opening logger. A small chip with a little switch (or a reed-relay magnetic switch, which allows it to get completely sealed in the doorframe, or an IR beam gate, or just about anything). A microcontroller running with slow clock in low-power mode, a serial EEPROM for storing the circular-buffer log of the times when the contact was opened/closed, a RTC chip with I2C bus, a small battery for making the device independent on the power. (It can even log the times of power blackouts; a sequence of blackout - door open - door close - power-up when nobody is home is VERY suspicious.) If you use a wireless interface there (eg, use AT24RF08 for the EEPROM), you may read the door open/close times without any direct contact with the device, which makes its concealmentability much easier. It can then reside in the doorframe, a little magnet in the door triggering its reed contact, read/write being done with a handheld reader held over a certain point of the doorframe. Scatter couple such devices in the object, and you have good idea about what was happening there. There is no authorization there; they will just log visits, including you - but you KNOW when you were there and can ignore these entries on your own. This is suitable as an audit device, just to make sure. Be aware that in the highest-threat situations the adversary may enter by unconventional means; there were cases of entry through the roof, the floor, even the wall. The cellphone-based uplink can be strengthened by a trick. There are cellular jammers that can be employed together with the plain ol' cutting of the lines. However, if you have good visibility to some cell base stations, you may employ a directional antenna pointing at it, which makes the setup much more resistant against localized jamming from directions other than the antenna's one. For high-budget people there are also satellite phones. The system can also report to you once every interval, telling you about status changes and if there was anything that could look suspicious but wasn't worth of an asynchronous alarm. Consider using an anonymous phone with a prepaid card, for both sending and receiving the alarms and status messages. The adversary could othervise visit the phone company and ask for cancelling your service for the given time, so you'd miss your alarm. If the neighbourhood situation is suitable, you can combine the efforts of several people in the area, and form a sort of Neighbourhood Watch - a decentralized P2P web of alarm systems talking with each other by any suitable means - from an over-the-street IrDA to 802.11b to Ronja-grade optical links for longer distances. (The network can also double as a community network for TCP/IP communication and Net connection.) This can make an eventual attack very unlikely to pass unnoticed. An interesting low-tech backup system could be a cheap one-time film camera triggered by eg. a fishing line tied to the door or lead over the floor. Even if everything electrical goes down, there is still a chance to get a picture of the intruder. Use different technologies in overlapping layers. Employ your fantasy. Be aware about what dependencies you have in the system (eg, camera system can run for no more than two hours of mains blackout, certain sensors depend on the computer, some alert modes depend on the telephone or cellphone), be prepared for different failure modes - both natural mishaps and intentional sabotage, compensate for them, design the system to fail gracefully and to keep at least some functions (at least the autonomous audit switch loggers) even in the reasonably worst case. This makes the adversary's job less pleasant.